University of Southern California Dissertations and Theses

Development, validation and testing of a new sensor array for intra-articular pressure measurement: in-vitro human lumbar spine intra-articular facet testing

(USC Thesis Other)

Development, validation and testing of a new sensor array for intra-articular pressure measurement: in-vitro human lumbar spine intra-articular facet testing

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

DEVELOPMENT, VALIDATION AND TESTING OF A NEW SENSOR ARRAY FOR
INTRA-ARTRICULAR PRESSURE MEASUREMENT: IN-VITRO HUMAN LUMBAR
SPINE INTRA-ARTICULAR FACET TESTING
by
Judson B. Welcher

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BIOMEDICAL ENGINEERING)

August 2011

Copyright 2011 Judson B. Welcher
ii

DEDICATION

I dedicate this dissertation to my wonderful family. Particularly to my wife who has put up with
many years of research and without whose help I would not have finished, to my children for
understanding why I had to work and for giving me a reason to finish, and finally to my mother
and my father who passed away during my research. My loving parents provided me the
nurturing, fortitude and constitution that has made me what I am today. Their belief in science,
the pursuit of academic excellence, and constant betterment of oneself instilled in me the love of
learning which drove me to start, and ultimately complete this journey in the first place.

iii

ACKNOWLEDGEMENTS

I would like to thank all those people that made this dissertation possible.
First, I wish to thank my advisor, Dr. Thomas Hedman. Despite my many hardships,
setbacks and seemingly insurmountable problems, Dr. Hedman always gave me unwavering
guidance, support, and encouragement. His patience and continued support of my research is
beyond comparison. His mentoring including his willingness to defend me when I needed
defending and push me when I needed pushing will serve as a benchmark from which I will judge
others in the future. His in depth understanding of spine biomechanics and critical and detailed
analysis of my research undoubtedly elevated the quality of my work.
My dissertation committee members provided me with an outstanding experience and
their time, energy and expertise in making sure my dissertation research was at its best is greatly
appreciated. Specifically, Dr. Michael Khoo, not only served as the co-chair of my committee,
but as head of the department during many years of my dissertation, he unfortunately routinely
defended and supported me in my dealings with the University Administration. His guidance,
support, and understanding that I worked full time and went to school part time were essential to
me completing this research. Drs. Jill McNitt-Gray and Satwindar Sadhal gave graciously of
their time and provide invaluable comments and questions.
I would also like to thank Mischal Diasanta, the Biomedical Engineering Graduate
Student Advisor, for helping me navigate the sea of paperwork, running interference between me
and the administration, answering my dumb questions and for generally helping me whenever I
needed help.
I would like to thank the many researchers, MRI technicians and orthopedic surgeons at
Cedars-Sinai Medical Center that assisted in this research. Specifically, I would like to thank Dr.
iv

Wafa Tawackoli who, in spite of a very difficult transition, maintained true to his word and
allowed me to finish my research in the Spine Biomechanics Research Lab. Dr. Tawackoli also
made many technical contributions to my research. Additionally, Dr. Melodie Metzger provided
me assistance in imaging and getting my specimens graded.
The best “non-engineer” engineer, Dr. John Popovich led by example. His unwavering
dedication and devotion to his research gave me motivation to finish. His attention to detail and
ability to jump in with both feet into topics he had no past experience with was inspirational
beyond this work. Dr. Popovich also gave enumerable hours of his time and his insight that were
critical to the completion of this research. To his future student, you are fortunate to have
someone as dedicated to a complete understanding of his craft as he.
This would not have been possible without the constant support of my family. My wife,
Jennifer, put up with many years of research, many weekends lost to testing, and many years
without a vacation. She has been the backbone of my support and has spent considerable time
with my children acting as both parents so that I could have time to complete my research. She
unselfishly frequently put my needs ahead of hers to help me complete this. To my children Cole
Welcher and Kylie Welcher, you provided the final motivation for me to complete this work so
that I would have more time to spend with my family. I hope you can forgive me and understand
why dad had to work every weekend to finish this.
Finally, to the scum bags that broke into my car and stole most of my research equipment
and lab books half way through my work, you nearly caused me to abandon my research but in
the process also reminded me that anything worth having is worth persevering to achieve.
“Adversity has the effect of eliciting talents which, in prosperous circumstances, would have lain
dormant.” – Horace.

v

TABLE OF CONTENTS
DEDICATION ................................................................................................................................ II

ACKNOWLEDGEMENTS ........................................................................................................... III

TABLE OF CONTENTS ................................................................................................................ V

LIST OF FIGURES ....................................................................................................................... IX

ABSTRACT ................................................................................................................................ XIV

CHAPTER 1: INTRODUCTION ................................................................................................... 1
1.1 SPECIFIC AIMS ............................................................................................................. 2
1.2 ORGANIZATION ............................................................................................................ 3

CHAPTER 2: BACKGROUND AND SIGNIFICANCE ............................................................... 4
2.1 THE EPIDEMIC OF BACK PAIN ....................................................................................... 4
2.2 THE FACET JOINT AS A SOURCE OF PAIN ...................................................................... 5
2.3 THE FACET AS A LOAD BEARING ELEMENT .................................................................. 7
2.4 REVIEW OF LUMBAR SPINE ANATOMY AND BASIC TERMINOLOGY ............................ 10
2.4.1 Lumbar Articular Facet Width and Height Constraints ............................................. 13
2.4.2 Lumbar Articular Facet Curvature ............................................................................ 15
2.4.3 Lumbar Articular Facet Thickness Limitations ......................................................... 17
2.4.4 Environmental Constraints ........................................................................................ 20

CHAPTER 3: LITERATURE REVIEW AND PRELIMINARY EVALUATION OF
EXISTING TECHNOLOGY ......................................................................................................... 21
3.1 RATIONALE FOR AN EXHAUSTIVE LITERATURE REVIEW ........................................... 21
3.2 BIOLOGICAL JOINT FORCE AND AREA MEASUREMENT METHODS - OVERVIEW ......... 21
3.3 STRAIN GAUGES ......................................................................................................... 23
3.4 PRESSURE SENSITIVE FILMS – INK BASED AND TRANSFERS...................................... 27
3.5 FORCE SENSITIVE/SENSING RESISTORS (FSR) .......................................................... 36
3.5.1 Interlink FSR-I ........................................................................................................... 37
3.5.2 Tekscan FSR-T .......................................................................................................... 48
3.5.3 Other FSR Technology ............................................................................................ 101
3.6 CAPACITIVE BASED SENSORS AND SENSOR ARRAYS ............................................... 102
3.6.1 Novel FSC-N (Force Sensitive Capacitor-Novel) ................................................... 102
3.6.2 XSENSOR FSC-X (Force Sensitive Capacitor-XSENSOR) .................................. 131
3.7 MATHEMATICAL AND FINITE ELEMENT MODELS ................................................... 131
3.8 PRELIMINARY EVALUATION OF EXISTING TECHNOLOGY ....................................... 132
3.8.1 Sensor Products Inc. ................................................................................................ 134
3.8.2 Vista Medical/FSA/Verg Inc. .................................................................................. 134
3.8.3 Novel ....................................................................................................................... 134
3.8.4 Pressure Profile Systems ......................................................................................... 135
3.8.5 Tekscan .................................................................................................................... 135
3.8.6 Interlink/Force Sensitive Resistors (FSR) ............................................................... 137

vi

CHAPTER 4: SENSOR PERFORMANCE CHARACTERISTICS .......................................... 141
4.1 INTRODUCTION ........................................................................................................ 141
4.2 METHODS ................................................................................................................. 144
4.2.1 Description of new sensor array construction. ........................................................ 144
4.2.2 Test apparatus .......................................................................................................... 145
4.2.3 Specifications .......................................................................................................... 148
4.3 RESULTS .................................................................................................................. 153
4.3.1 Thickness – Total sensor array less than or equal to 0.8 mm thick. ........................ 153
4.3.2 Scalable geometry - Down to less than 12.0 mm in width and 15.1 mm in
height. ................................................................................................................................. 153
4.3.3 Flexibility in sensor spacing - Spatial resolution including individual element
location, number of sensors and effects of sensor spacing on adjacent sensors (cross-talk). 154
4.3.4 Curvature affects - Less than 1% zero load offset and less than 1% change in
sensitivity on curved surfaces down to a 9 mm radius sensor array curvature. ..................... 155
4.3.5 Frequency response – Less than 2% loss of signal power (less than 0.09 dB) at
frequencies from steady state (0 Hz) up to 5 Hz. .................................................................. 156
4.3.6 Non-linearity – Less than 1% of full scale (FS). ..................................................... 156
4.3.7 Drift – Less than 1% of full scale out to at least 700 seconds. ................................ 157
4.3.8 Hysteresis – Less than 1%. ...................................................................................... 157
4.3.9 Repeatability – Coefficient of variation less than 2%. ............................................ 158
4.3.10 Total cost of utilization including hardware, software, and sensors ........................ 159
4.4 DISCUSSION ............................................................................................................. 159

CHAPTER 5: PRELIMINARY IN-VITRO RESULTS ............................................................. 162
5.1 INTRODUCTION ........................................................................................................ 162
5.2 METHODS ................................................................................................................. 162
5.2.1 Sensor construction ................................................................................................. 162
5.2.2 Facet geometry and sensor geometry constraints .................................................... 163
5.2.3 Preliminary in vitro lumbar spine facet testing ........................................................ 168
5.2.4 Total cost of utilization - hardware, software, sensors and in vitro durability. ....... 173
5.3 RESULTS .................................................................................................................. 173
5.3.1 Facet Geometry and Sensor Geometry Constraints ................................................. 173
5.3.2 Preliminary In vitro lumbar spine facet testing ....................................................... 173
5.3.4 Total cost of utilization - hardware, software, sensors and in vitro durability. ....... 176
5.4 DISCUSSION ............................................................................................................. 176

CHAPTER 6: LUMBAR FACET SPATIAL AND TEMPORAL MEASUREMENTS
UNDER PURE FLEXION-EXTENSION LOADING ................................................................ 181
6.1 INTRODUCTION ........................................................................................................ 181
6.2 METHODS ................................................................................................................. 186
6.3 RESULTS .................................................................................................................. 191
6.4 DISCUSSION ............................................................................................................. 196
6.5 CONCLUSIONS .......................................................................................................... 200

CHAPTER 7: SUMMARY AND CONCLUSIONS .................................................................. 202

REFERENCES ............................................................................................................................ 206
vii

LIST OF TABLES

Table 1: Means (and Standard Errors of the Mean) of the Inferior Articular Facet
Dimensions ........................................................................................................................... 15
Table 2: Means (and Standard Errors of the Mean) of the Superior Articular Facet
Dimensions ............................................................................................................................ 15
Table 3: Zygapophyseal Joint Thickness (PT#=subject number, RL# and LL#=right and left
lumbar spinal level, O#=observer number, SD=standard deviation) ..................................... 18
Table 4: L3-L4 facet loads with EAS technique and 7.5Nm of bending about each axis ........... 27
Table 5: Fujifilm results for L4-5 facets ...................................................................................... 31
Table 6: Facet loads for intact and implanted specimens ............................................................. 32
Table 7: FSR performance results ................................................................................................ 48
Table 8: Tekscan typical system performance ............................................................................. 50
Table 9: Intraclass correlation coefficients for dynamic testing .................................................. 57
Table 10: Facet loads for flexion and extension and lateral bending ........................................... 83
Table 11: Facet contact forces [Average+/-S.D.] ........................................................................ 93
Table 12: Comparison of facet loads ........................................................................................... 94
Table 13: Contact force in the facet joint measured by the FSR-T under 7.5 Nm moments
(negative is compressive force) ............................................................................................ 94
Table 14: Intact facet loads reported in the literature ................................................................... 95
Table 15: Technical Data for the EMED platforms ................................................................... 104
Table 16: Technical Data for the Pedar system ......................................................................... 104
Table 17: Technical Data for the Pliance system ........................................................................ 105
Table 18: Intraclass correlation coefficients .............................................................................. 109
Table 19: Comparison of Means ................................................................................................ 113
Table 20: Sensor attachments sites on trans-tibial stump mould [note correction to site
3 and 4] ................................................................................................................................ 115
Table 21: Summary of results [mean+/- SD, n=sample size] .................................................... 116
viii

Table 22: Dynamic measurement error of the FSC-N ................................................................ 122
Table 23: In-shoe repeatability results ....................................................................................... 124
Table 24: Characteristics of the five tested PMD’s .................................................................... 129
Table 25: Results of PMD testing ................................................................................................ 131
Table 26: Existing Technology Evaluation (* refers to unverified claims in product
literature, n/a – no information was located, # combined thickness of the sensor,
Fujifilm and tape) ................................................................................................................. 139
Table 27: Proximity results ......................................................................................................... 154
Table 28: Affects of sensor array curvature ................................................................................ 155
Table 29: Sensor array drift response ......................................................................................... 157
Table 30: Repeatability Testing .................................................................................................. 158
Table 31: Lumbar facet height, width and curvature data .......................................................... 168
Table 32: Description of spine segments utilized ....................................................................... 169
Table 33: Human lumbosacral specimens (n=6) ......................................................................... 188
Table 34: Average maximum L4-5 pressures under various loading conditions. ....................... 195
Table 35: Center of pressure location under various loading conditions ..................................... 195
Table 36: Comparison of facet loads to prior research ............................................................... 198

ix

LIST OF FIGURES

Figure 1: Relative load sharing in resisting (a) shear(S), (b) compression (C), (c) torsion (T),
and (d) bending moment (M). The proportion of resistance of each structure is
represented by the size of the solid arrowhead ....................................................................... 8
Figure 2: Effects of disc degeneration on facet loads .................................................................... 9
Figure 3: Illustrations of the Anterior (A), Middle (M) and Posterior (P) Columns of the
Human Spine ........................................................................................................................ 11
Figure 4: Lumbar spine ligaments ............................................................................................... 11
Figure 5: Lumbar Spine: Facets Joint and Articular Capsule ..................................................... 12
Figure 6: Lumbar Spine Osseous Anatomy ................................................................................. 13
Figure 7: Mnemonics for Each Facet Dimension ........................................................................ 14
Figure 8: Shape and Inclination of the Lumbar Facet Joints ....................................................... 16
Figure 9: Histological Cross Section of Lumbar Spine ............................................................... 16
Figure 10: Transverse section of L3-4 Joint ................................................................................. 17
Figure 11: Lumbar Zygapophyseal Joint Thickness “A” ............................................................. 18
Figure 12: Effects of Sensor Thickness ....................................................................................... 19
Figure 13: Strain gauges on exterior of lumbar facet ................................................................... 23
Figure 14: Strain coupling in right facet from loading of the left facet ....................................... 25
Figure 15: L3-L4 Facet forces under a 7.5 Nm axial moment ...................................................... 26
Figure 16: Fuji Prescale Pressure Sensitive Film system.............................................................. 28
Figure 17: Fujifilm Low Pressure LW film calibration charts (Left is continuous pressure
and right is momentary) ......................................................................................................... 29
Figure 18: Peak facet pressure at various postures ...................................................................... 31
Figure 19: Effects of loading rate on color density ....................................................................... 34
Figure 20: FSR-I construction ..................................................................................................... 37
Figure 21: FSR-I calibration curve .............................................................................................. 38
x

Figure 22: Interlink Electronics FSR-I sensor construction ......................................................... 39
Figure 23: Area affects on FSR-I resistance ................................................................................ 39
Figure 24: Variation in FSR-I output with contact area ............................................................... 40
Figure 25: Domed FSR-I ............................................................................................................. 41
Figure 26: (A) Response of sensor with an epoxy dome for input forces applied using
different size surface areas, including a curved surface. (B) Response of a sensor
without an epoxy dome for forces applied using different size surface areas ...................... 42
Figure 27: Induced facet force with sensor thickness .................................................................. 44
Figure 28: FSR-I response variation with sensor warm up (note the up then down values on
the x-axis) .............................................................................................................................. 45
Figure 29: Modified FSR-I configuration .................................................................................... 46
Figure 30: Modified FSR-I calibration results ............................................................................. 47
Figure 31: Grid based and single load cell sensors from Tekscan ............................................... 49
Figure 32: Cross sectional dimensions (mm) of one ink-covered grid conductor strip ............... 49
Figure 33: Tekscan variation in resistance with applied load ...................................................... 51
Figure 34: Absolute error with increasing pressure ..................................................................... 56
Figure 35: Maximum pressure variation under cyclic load to constant peak force ..................... 58
Figure 36: Change in readings during cycle and between cycles ................................................ 59
Figure 37: Effects of indentor shape on F-Scan FSR-T response ................................................ 61
Figure 38: Temperature effects on F-Scan output ....................................................................... 63
Figure 39: FSR-T dynamic response time ................................................................................... 64
Figure 40: Phase and magnitude lags of F-Scan (FSR-T) ............................................................ 65
Figure 41: Sensor drift response (a) long term continuous and intermittent loading, (b) short
term continuous loading ........................................................................................................ 67
Figure 42: Erroneous time constant results ................................................................................... 68
Figure 43: Magnitude ratios ......................................................................................................... 69
Figure 44: FSR-T sensitivity change ........................................................................................... 70
xi

Figure 45: Effects of surface conditions on FSR-T sensitivity .................................................... 71
Figure 46: Measured area change with constant area loading ..................................................... 72
Figure 47: Curvature effects on FSR-T output ............................................................................ 74
Figure 48: Offset and sensitivity changes in FSR-T with curvature ............................................ 75
Figure 49: MRI of patellofemoral joints (A) 0º, (B) 45º, (C) 90º, and (D) 135º knee flexion ..... 80
Figure 50: Curvatures of the knee joint. Lateral x-ray(left) and anterior view (right) ................ 81
Figure 51: Facet loads intact and modified L5-S1 ....................................................................... 82
Figure 52: “Injured” L3-4 fact loads measured by FSR-T under 7.5 Nm (1.3º/sec) extension
loading .................................................................................................................................. 85
Figure 53: FSR-T loading configuration ...................................................................................... 86
Figure 54: Radius of curvature on superior Talus surface ........................................................... 87
Figure 55: The flat (A), spherical (B) and cylindrical (C) load application fixtures ................... 88
Figure 56: Pin Method calibration relative to average raw sensel response (+/- 1 SD) ............... 91
Figure 57: Tekscan 6900 FSR-T sensor [1 of 4 fingers] ............................................................... 92
Figure 58: Facet load at L3-4 with 7.5 Nm flexion and extension moment ................................ 93
Figure 59: Force measurement errors of FSR-T in lumbar face joint .......................................... 97
Figure 60: FSR-T measured L4-5 facet forces (E-Extension, F-Flexion, RLB and LLB –
Right and left lateral bending, CAR and AAR – Clockwise and Anticlockwise axial
rotation) ................................................................................................................................. 98
Figure 61: Percentage load on conductor intersection ................................................................. 99
Figure 62: FSR-T loading interface ........................................................................................... 100
Figure 63: Coefficient of reliability as a function of the number of2 walks onto platform at
medium speed ..................................................................................................................... 107
Figure 64: Trublu Novel Air Bladder calibration set up ............................................................ 110
Figure 65: Effects of FSC-N sensor position ............................................................................. 111
Figure 66: Gait parameters for comparison ............................................................................... 112
Figure 67: Pressure mean error +/- 1 SD at each stump location................................................ 117
xii

Figure 68: FSC-N drift measured during swing phase of five subjects ..................................... 119
Figure 69: Static FSC-N sensor drift .......................................................................................... 121
Figure 70: Bladder calibrator and test fixture ............................................................................. 126
Figure 71: Static pressure measurement results measured over the whole surface ................... 130
Figure 72: Tekscan Error Relative to Tekscan Measured Load .................................................. 136
Figure 73: Normalized performance of the Tekscan, Enduratec and new pressure sensors. ....... 137
Figure 74: Calibration Plots for FSR’s. ...................................................................................... 138
Figure 75: Pressure sensor element geometry (left) and sample sensor array configuration
(right) ................................................................................................................................... 145
Figure 76: Calibration pressure vessel set-up (prior to sealing). ................................................ 147
Figure 77: Position of two sensor elements for proximity affects testing (small hash marks
are millimeters) .................................................................................................................... 149
Figure 78: Sensor array for curvature tests (left) placed in cylindrical collars (right) of
various radii ......................................................................................................................... 150
Figure 79: A typical PSD results for the reference transducer and the new sensor array ........... 156
Figure 80: Typical sensor hysteresis curve ................................................................................. 158
Figure 81: Pressure sensor element geometry (left) and sample seven sensor array
configuration (right) ............................................................................................................. 163
Figure 82: Sensor positions in the array (to scale) ...................................................................... 165
Figure 83: Typical L5 Vertebrae with Metal Measurement Rod “K” in Solid Black ................. 166
Figure 84: Basic Circle Descriptors (r=radius, c=chord length, h=chord height,
s=arc length) ........................................................................................................................ 167
Figure 85: Optotrak coordinate system and target location ........................................................ 171
Figure 86: Anterior view of lumbosacral specimen setup in Bose Kinematic
Spine Simulator ................................................................................................................... 172
Figure 87 Specimen moment loading and response .................................................................... 174
Figure 88: Sensor at 0 mm,-5mm location response to all cycles ............................................... 175
Figure 89: Facet pressures at 50% extension .............................................................................. 175
xiii

Figure 90: Facet pressures at 100% (10 Nm) extension spines 1 and 2 ...................................... 176
Figure 91: Seven sensor array (left) and sensor position and coordinate system (right). ............ 187
Figure 92: Measurement sensor in right L4-5 facet and dummy array in left facet. ................... 190
Figure 93: Typical system response to applied bending moment (positive moments and
angles are flexion). ............................................................................................................. 192
Figure 94: Relationship of applied moment to L4-5 rotation and facet pressure ....................... 193
Figure 95: Spatial pressure distribution at 50% of maximum L4-5 extension............................. 194
Figure 96: Spatial pressure distribution at 100% of maximum L4-5 extension. .......................... 194
Figure 97: Mean center of pressure migration under extension loading from neutral
position to 50% extension, 7.5 Nm and 100% extension (10 Nm). ..................................... 196

xiv

ABSTRACT

There is a very high and increasing frequency of back pain in modern society with an
estimate annual cost of $100 billion dollars just in the United States. The precise etiology of low
back pain lacks a general consensus. However, the facet joints have been shown to be a
significant source of spinal pain, and specifically low back pain. Despite being a significant
source of pain, very little accurate data is available on the loads within the facet joints.
This study’s primary goal was to provide improved data, thus facilitating a better
understanding of lumbar spine intra-articular facet loads during movement. This was ultimately
accomplished by the development, validation and implementation of a new method for direct
measurement of intra-articular load. It was initially thought that a better understanding of lumbar
spine intra-articular facet loads during movement could be accomplished with the application of
existing technology. The preliminary phases of this study involved evaluating a large volume of
literature related to previously published methods and techniques purporting to quantify intra-
articular load, meeting and discussing design requirements with various manufacturers of
potential technology, and finally bench top testing and evaluating the more plausible
technologies. Despite a substantial volume of published literature utilizing many of these
technologies for this task and similar measurements in other articular joints (both human and non-
human), none of the existing technologies were found to have the necessary geometry and/or
were capable of taking accurate measurements when curved to physiologically prevalent radii.
These deficiencies in the existing technologies necessitated the development of an appropriate
method to accurately measure intra-articular pressures.
A new sensor array intended to accurately and directly measure both the spatial and
temporal distributions of pressures within a highly curved intra-articular joint was developed and
xv

tested. To evaluate performance of the new sensor array for application within intra-articular
joints generally, and specifically to fit within the relatively restrictive space of the lumbar spine
facet joint, geometric constraints of length, width, thickness and sensor spatial resolution were
evaluated. Additionally, the effects of sensor array curvature, frequency response, linearity, drift,
hysteresis, repeatability, and total system cost were assessed.
The new sensor array was approximately 0.6mm in thickness, scalable to below the
nominal 12 mm wide by 15 mm high lumbar spine facet joint size, offered no inherent limitations
on the number or spacing of the sensors with less than 1.7% cross talk with the sensors positioned
immediately adjacent to one another. No difference was observed in sensor performance down to
a radius of curvature of 7 mm and a 0.66±0.97% change in sensor sensitivity was observed at a
radius of 5.5 mm. The sensor array had less than 0.07 dB signal loss up to a loading frequency of
5.5 Hz, linearity was 0.58±0.13% full scale (FS), drift was less than 0.2% FS at 250 s and less
than 0.6% FS at 700 s, and hysteresis was 0.78±0.18%. Repeatability was excellent with a
coefficient of variation less than 2% at pressures between 0 and 1.000 MPa. Total system cost
was relatively small as standard commercially available data acquisition systems could be
utilized, with no specialized software, and individual sensors within an array could be replaced as
needed. The new sensor array had small and scalable geometry and very acceptable intrinsic
performance including minimal to no alteration in performance at physiologically relevant ranges
of joint curvature.
The most limiting attribute of the sensor array is durability. Although this was improved
during the study with judicious selection of sensor orientation and modifications of the individual
sensor construction via discussions with the manufacture, durability remained less than ideal.
This limitation was tempered by the fact that the componentized nature of the sensor array
allowed easy replacement of the individual sensor elements within a given array if one or more
xvi

sensors failed. This characteristic is beneficial especially when compared to other commonly
commercially available arrays such as those from Tekscan (FSR-T) and Novel (FSC-N) which
are constructed by the manufacturer in fixed configurations and do not allow either redistribution
or replacement of individual sensor elements within an array.
Preliminary in vitro data is presented demonstrating the utility of the new sensor array in
quantifying temporal and spatial distributions of pressure within the L4-5 facet joint. Preliminary
results are generally in line with singular peak pressure measurements from pressure sensitive
film testing with peak pressures measured in the current study at between 1,210 kPa and 3,059
kPa. Additionally, the distribution of pressure matched prior studies in that the measured facet
pressure increased in extension, decreased in flexion, and the center of pressure migrated
inferiorly and medially under increasing extension moments. Initially, in vitro durability was
problematic with very high initial sensor failure rates. Manufacturing changes and orientation
optimization of each individual sensor relative to load direction improved durability. Durability
was ultimately considered acceptable in light of the ease and relatively low cost of individual
sensor replacement.
The utility of the sensor in more accurately quantifying spatial and temporal changes in
lumbar spine facet intra-articular pressure was demonstrated with testing six fresh-frozen human
cadaveric lumbosacral specimens under pure moment bending (±10Nm). The new sensor was
inserted in the L4-5 facet joints. L4-5 facet contact pressures were continuously measured at
seven locations within the facet. Center of pressure at various phases of loading was calculated.
The data demonstrated an increase in facet pressure with increasing extension moments and
displacements. Facet contact pressure increased relatively linearly in proportion to applied
bending moment up to approximately 2-3 degrees of L4-5 extension and approximately 7-8 Nm
of extension moment. The highest average maximum pressures of 1,087 kPa were found in the
xvii

midline sensor 2 mm medial of the midpoint and 973 kPa in the most inferior midline sensor.
The most superior midline sensor always had the lowest average peak pressures during extension.
The center of pressure started very near the anatomical center of the facet and migrated medially
and inferior under increasing extension moments.
The demonstrated functionality of the new sensors in the relatively small and sharply curved
human lumbar spine facet joint should ensure viability and utility of the sensor array in other less
geometrically demanding joints and surface interfaces such as the hip, knee and ankle joints. The
sensors could also be used as a source of tactile feedback in prosthetic designs or for external
measurement of a portion of the body, such as the foot interacting with the ground or other
objects.

1

CHAPTER 1: INTRODUCTION

Many studies attest to the high frequency of back pain complaints in modern society,
with 60–85% of all people experiencing back pain at some time in their lives (Andersson, 1997,
1999, Bierung-Sorensson, 1984, Frymoyer et al., 1983, Manchikanti et al., 2009, Svensson et al.,
1983). Back pain results in approximately 40% of absences from work (Guo et al., 1999), is
second to the common cold as the most frequent cause of sick leave (LaBar, 1992) , the third
most common cause of surgical procedures (LaBar, 1992, Hart et al., 1995), and is the leading
cause of disability in people under 45 (Pope et al., 1991). The total cost of back pain is estimated
to exceed $100 billion annually (Frymoyer, 1991, Luo, 2003, Katz, 2006). The precise etiology
of back pain lacks a general consensus. However, a recently published literature review utilizing
criteria established by the International Association for the Study of Pain found the facet joints
were found to be a possible source of chronic pain in 15% to 45% of patients with chronic low
back pain (Dryer & Dreyfuss, 1996).
A more complete understanding of facet joint mechanics could improve treatment and
intervention methods and help reduce back pain related liabilities. A quantitative understanding
of spinal intra-articular joint mechanics is fundamental to a complete understanding of spine joint
mechanics. A significant component of spine joint mechanics involves understanding the
temporal and spatial changes in intra-articular loads associated with movement. One of the three
main load bearing elements in the human spine is the facet joints. Quantifying the temporal and
spatial changes of human lumbar spine intra-articular facet loads will assist development of
motion sparing prosthetics that better mimic the in vivo condition, will provide data for validation
of finite element models, and will aid in development of other preventative measures.
2

The current sensor design requirements are quite rigorous because the human lumbar
spine facet joints are relatively small, are highly curved and frequently very little true intra-
articular space exists. Accurate functioning under these more restrictive requirements provides
justification for utilization of these sensors in large body joints such as the knee or ankle.
1.1 Specific Aims
The current study was performed at various locations including the Orthopedic
Biomechanics Laboratory at USC, the Spine Biomechanical Laboratory at Cedars-Sinai Medical
Center, and Biomechanical Research and Testing. This work will update and improve with new
technology the original pioneering works conducted by Prasad et al., (1974), Prasad and King
(1974) and Hedman (1992, 1995, 1997). The specific aims of the current study were as follows:
1. Critically review the applicability and accuracy of all currently published methods
and techniques for measurement of intra-articular force, pressure and/or contact area.
2. Build, validate and utilize a bench top test apparatus for performance evaluation of
existing and potential new technologies intended to measure the spatial and temporal
pressure distributions within human lumbar facets joint under load.
3. Bench top performance validation of the new sensor array with physiologically
relevant test criteria including: geometric constraints of length, width, thickness and
sensor spatial resolution; and the effects of sensor array curvature, frequency
response, linearity, drift, hysteresis, repeatability, and total system cost.
4. Perform preliminary in vitro testing to identify any performance issues associated
with in vitro testing compared to bench top testing.
5. Provide more accurate and more robust data on lumbar spine mechanics by directly
measuring the in vitro intra-articular temporal and spatial distributions of lumbar
3

facet joint contact pressures when spine specimens are subjected to pure flexion-
extension bending moments.
1.2 Organization
This research is effectively presented in a similar order outlined in the Specific Aims
above. Chapter 1 provides a very cursory statement of the problem and the proposed aims of this
study. A more thorough evaluation of the literature is presented as the background in Chapter 2.
This chapter also provides a brief presentation of spinal anatomy with specific details relative to
the complicated geometry and constraints of the lumbar spine facet joints. Chapter 3 begins with
justification for what follows as a rather lengthy and rigorous review of the existing technology in
the literature. Chapter 4 describes the construction of the bench top test apparatus, test criteria
such as curvature, and the results of the bench top testing. Chapter is 4 a longer version of a
paper that has already been accepted for publication in the Journal of Medical Engineering and
Physics (Welcher JB, et al. Development of a versatile intra-articular pressure sensing array. Med
Eng Phys (2011), doi:10.1016/j.medengphy.2011.03.004). Similarly, Chapter 5 is a longer
version of a paper that has been submitted to the Journal of Biomechanics and describes the
preliminary in vitro testing, associated problems and solutions. Chapter 6 represents the
utilization with quantification of the temporal and spatial intra-articular facet pressure
distributions under pure-moment bending. The data from Chapter 6 has been submitted to the
Spine journal. The last chapter, Chapter 7, consists of a summary of the current study and overall
conclusions.

4

CHAPTER 2: BACKGROUND AND SIGNIFICANCE

2.1 The epidemic of back pain
Many studies attest to the high frequency of back complaints in modern society, with 60–
85% of all people experiencing back pain at some time in their lives (Andersson, 1997, Bierung-
Sorensson, 1984, Frymoyer et al., 1983, Manchikanti et al., 2009, Svensson et al., 1983). The
annual prevalence of back pain ranges from 15% to 45%, with a point prevalence averaging 30%
(Andersson, 1997, Manchikanti et al., 2009, Svensson et al., 1983). In the United States, back
pain is the most common cause of activity limitation in people younger than 45 years (Pope et al.,
1991). After the common cold, low back pain is the second most frequent reason for visits to the
physician and the fifth-ranking cause of admission to hospitals (Labar, 1992, Hart et al., 1995).
Back pain results in about 40% of the absences from work and is the second most frequent cause
of sick leave (Guo et al., 1999, Labar, 1992). Low back pain is the third most common cause of
surgical procedures (Palmer et al., 2000; Taylor et al., 1994; Hart et al., 1995, Andersson, 1999).
In 1988, back and spine impairments resulted in over 185 million days of restricted activity (21
days per impairment), which included 83 million days confined to bed (Praemer el al., 1999). In
the United States between the periods of 1971 to 1986, the number of people disabled from back
pain increased 168%, a rate 14 times the rate of population growth (Haldeman, 1990). There was
an average of 158 low back surgeries per 100,000 adults in 1990 (Goldberg et al., 2001). Recent
studies (Deyo et al, 2005, 2009, Manchikanti et al., 2009) document a 629% increase in Medicare
expenditures for epidural steroid injections; a 423% increase in expenditures for opioids for back
pain; a 307% increase in the number of lumbar magnetic resonance images among Medicare
beneficiaries; and a 220% increase in spinal fusion surgery rates from 1990 to 2001 in the United
States. In the United States, the total cost of back pain is estimated to exceed $100 billion
5

annually (Frymoyer, 1991, Luo, 2003, Katz, 2006). The high prevalence of back pain is not
unique to the United States. A study in England (Harkness et al., 2005) comparing the prevalence
of low back pain in 1950 to that in 1995 found an increase in males from 8.1% to 17.8% and an
increase from 9.1% 18.2% in females.
2.2 The facet joint as a source of pain
The precise etiology of low back pain lacks a general consensus. However, facet
arthrosis is a common radiographic finding and has been suggested as a common cause of low
back pain since as early as 1911 (Goldthwait, 1911; Badgley, 1941). Putti (1927) and Goldthwait
(1911) studied x-rays and noted “the diseased (facet) joint, by its swelling and deformity, changes
the shape and capacity of the foramen, thus irritating and compressing the nerve within it. Hence
the neuralgia, or “nevrodocites” (nevrodocites was defined as nerve pain due to stimulation of the
nerve by the bony foramen or canal which contains it (Sicard, 1918)). A study in 1933
(Ghormley, 1933) first published the term “facet syndrome” and suggested that hypertrophic
changes secondary to osteoarthritis of this joint led to nerve root entrapment which caused low
back pain.
Mooney and Robertson (1976) were able to elicit pain in normal volunteers by injecting h
hypertonic saline into the facet joint. The pain then was relieved with local anesthetic injection
into the facet joint. A later randomized clinical trial (Lilius et al., 1989) with three groups of
unilateral low back pain patients examined injections of cortisone and steroids into the joint and
around the joint and injection of saline into the joint. They found significant improvement in all
groups, independent of the treatment given. The Lilius study has been criticized for using 8 ml of
fluid for intra-articular injections when other studies have demonstrated the joint has a typical
capacity of only 2 ml (Maldjian et al., 1998). Using facet injections Griffiths et al. (1993)
achieved long-term pain relief in 50% of their population with immediate relief in 90%. Studying
6

chronic spinal pain patients (Manchikanti et al., 2004), there was found to be a 31% prevalence of
facet joint pain in patients with chronic lumbar spine pain. Other studies have found that
percutaneous facet joint block is a useful diagnostic and therapeutic procedure in the management
of lumbar facet syndrome if used in a meticulously selected patient population (Maldjian et al.,
1998). In an extensive literature review by Boswell et al. (2007), consistent with criteria
established by the International Association for the Study of Pain (Merskey & Bogduk, 1994),
facet joints were found to be a possible source of chronic pain in 15% to 45% of patients with
chronic low back pain; 36% to 60% of the patients with chronic neck pain; and 34% to 48% of
the patients with thoracic pain. A lower prevalence of facet pain was reported in individuals
under 65 years of age (30%) compared with those greater than 65 years of age (52%)
(Manchikanti et al, 2001). Studies of anesthetized rabbits and goats (Avramov et al., 1992, Lu et
al., 2005, Cavanaugh et al., 2006) showed the facet capsules in their animal models were
innervated by various types of nerve fibers and they provide a electrophysiologic based
hypothesis for sources of facet syndrome pain and a possible rational for the often poor
localization of pain in patients with facet mediated low back pain. Excessive stretching of the
joint capsule has been hypothesized as a source of facet pain as the capsule surrounding the facet
joint has been shown to be well innervated by afferent nociceptive fibers, which are activated by
mechanical stresses (Yang & King, 1984, Yamashita, et al., 1990, 1996, Cavanaugh et al., 1996,
1997, 2006, El-Bohy, 1988, 1990, El-Bohy et al1989).
Research (Vernon-Roberts & Pirie, 1977, Gotfried et al., 1986, Butler et al., 1990)
examining the evolution of facet and disc degeneration has consistently shown that disc
degeneration precedes facet degeneration. Butler et al. (1990) concluded that disc degeneration
occurs before facet joint osteoarthritis (degeneration), which “may be” secondary to mechanical
7

changes in the loading of the facet joint. The sequence of degeneration is concluded to be that the
intervertebral disc is first affected and then for mechanical reasons the facet joints degenerate.
2.3 The facet as a load bearing element
The facets have been shown to be significant load bearing elements in the spine. Early
testing examining vertical spine loading (Ewing et al., 1972, King et al, 1975, Prasad et al., 1974,
Prasad & King, 1974) showed the L3-4 level facets carried between 35% and 50% of the total
spine load when the spine was subjected to 2.5 to 9 G’s of vertical accelerations. The authors
noted “the complexity of the posterior structures and limitation of space in and around the joints
of the facets precluded a direct measurement of facet load” and this was “still beyond state of the
art.” The facet loads were determined by assuming the measured seat pan loads were
proportional to the spine loads. An intervertebral load cell (IVLC) measured the load carried by
the vertebral body and the difference between the IVLC values and the seat pan load cell was a
measure of the loads carried by the facets.
The increasing clinical and biomechanical data indicating the importance of the posterior
elements, including the facet joint, led to a symposium in 1983 at the International Society for the
Study of the Lumbar Spine completely devoted to “The Biomechanics of the Posterior Elements
of the Lumbar Spine” (Anderson, 1983, Adams & Hutton, 1983, Jayson, 1983, Miller et al.,
1983). A paper at the conference by Adams and Hutton (1983) noted the facet joints and joint
capsule resisted most of the intervertebral shear force, share in resisting compressive force in
extension, and prevent torsional and flexural damage to the disc annulus. The relative
contribution to each direction of loading is shown in Figure 1.

8

Figure 1: Relative load sharing in resisting (a) shear(S), (b) compression (C), (c) torsion
(T), and (d) bending moment (M). The proportion of resistance of each structure is
represented by the size of the solid arrowhead (Adams & Hutton, 1983)

Finite element models with extension testing and a 1000 N preload found the L2-3 facets carried
approximately 30% of the compressive preload when extended to 6 degrees (Shirazi & Drouin,
1987). Externally mounted strain gauges (Schendel, 1990, Schendel et al., 1993) demonstrated
the L1-2 facets carried no load in flexion, 205 N at a 10 Nm extension moment and 190 N axial
load, 65 N at a 10 Nm torsion moment and 150 N axial load, and 78 N at a 3 Nm lateral bending
moment and 160 N axial load. The facets have also been found to be a substantial contributor to
lumbar shear stiffness with a 77.7% reduction in anterior shear stiffness and 79% reduction in
posterior shear stiffness with removal of the posterior elements (Lu et al., 2005). The facet and
9

capsule alone were found to contribute 54% to anterior shear stiffness and 67% to posterior shear
stiffness (Skipor et al., 1985). Pollintine et al. (2004) showed that the percentage of compressive
load carried by the facet joints (posterior arch) increased from 0 to 25% with disc degenerated
graded 1-2 (scale of 1 “no degeneration” to 4 “severe degeneration”) to between 0 and 100% with
disc degeneration graded 3-4 as shown in Figure 2. They noted neural arch load bearing
remained below 20% up to age 50 yrs, but then increased to 40-90% in old age.

Figure 2: Effects of disc degeneration on facet loads (Pollintine et al., 2004)

A complete table summarizing prior facet load research is provided in Chapter 6, Table
36.
Despite the prevalence of facet mediated spinal pain and the facet joints serving as
significant load bearing elements in the spine, none of the current standards or published testing
protocols specify requirements for measurement or quantification of facet loads (Wilke et al.,
1998, McNally, 2002, Goel et al., 2006, ASTM F2077-03, ASTM F2346-05, ASTM F2423-05,
ASTM F2624-07, ASTM F2790-10, ISO 18192-1, ISO 18192-1). One standard noted that a
10

purpose of an artificial IVD (intervertebral disc prosthesis) is long-term restoration of function
(ASTM F2423-05) yet this standard has no requirements for quantification of facet intra-articular
loading, static or dynamic. A standard specifically intended to provide guidance for the static and
dynamic testing of Lumbar Total Facet Prostheses (FP) (ASTM F2790-10) likewise does not
address measurement of facet loads or pressures. A set of test protocols for evaluation of spinal
implants published by a group of highly regarded spine researchers (Goel et al., 2006)
acknowledged a need for a basic understanding of the facet joints and facet arthritis, and that a
focus of testing is quantification of load-sharing between the device and functional spinal unit
structures. However, these authors do not mention measurement of the load or load sharing on
the facets. A book on spinal implants subtitled “Are we evaluating them appropriately?”
(Melkerson et al., 2003) contains no data on facet loads or how to evaluate them and the word
facet is not present in the subject index. These failures to specify, measure or otherwise quantify
either the static or dynamic load or pressures within the facet joints may account for some of the
reported clinical failures of motion-sparing/preserving prosthetics.
2.4 Review of lumbar spine anatomy and basic terminology
Contemporary descriptions of the human lumbar spine indicate it is composed of three
columns (Denis, 1983, 1984) as shown in Figure 3. The posterior column is formed by the
posterior bony complex (posterior arch) alternating with the posterior ligamentous complex
which is composed of the supraspinous ligament, interspinous ligament, facet capsule and
ligamentum flavum (Figure 4). The middle column is formed by the posterior longitudinal
ligament, posterior annulus fibrosus, and the posterior wall of the vertebral body. The anterior
column is formed by the anterior longitudinal ligament, the anterior annulus fibrosus, and the
anterior part of the vertebral body (Denis, 1983).

11

Figure 3: Illustrations of the Anterior (A), Middle (M) and Posterior (P) Columns of the
Human Spine (Denis, 1983).

Figure 4: Lumbar spine ligaments (White & Panjabi, 1990)
12

The articular facets are involved in the articulation between adjacent vertebrae. The
superior (upper) vertebra articulates with the superior articular facet of the vertebra beneath it
forming a facet joint as shown in Figure 5.

Figure 5: Lumbar Spine: Facets Joint and Articular Capsule (White & Panjabi, 1990,
Adams & Hutton, 1983)

The lumbar vertebral bodies are the major load bearing elements in the lumbar spine
anterior column. But the facet joints carry a significant proportion of load in extension and
compression. The percentage of load carried by the facets has also been shown to increase with
increasing degeneration (Dunlop et al., 1984).
As will be demonstrated, the lumbar facet joints exhibit challenging geometric properties
that make appropriate sensor design difficult. The facets are relatively small in size, curved and
in some specimens irregularly and sharply curved, and are separated by a virtual space occupied
by a thin layer of synovial fluid that has been shown to be compatible with only the thinnest of
sensors due to artifactual pre-loading (Hedman, 1992, 1995).
The general osseous anatomy of typical lumbar vertebrae is as shown in Figure 6. The
lumbar spine facet joints, also known as zygapophyseal joints (an apophyseal joint in the spine),
13

are true diarthrodial joints complete with cartilage and synovial lining on each surface and a joint
capsule around them as shown in Figure 5 (Giles & Taylor, 1984, Woodburne, 1988).

Figure 6: Lumbar Spine Osseous Anatomy (Netter, 1993, Woodburne, 1988)

The superior and inferior articular processes project respectively upward and downward
from the junctions of the pedicles and lamina. The facets on the superior processes are concave,
and look backward and medial while those on the inferior processes are convex, and are directed
forward and lateral. The superior are wider apart than the inferior, since in the articulated column
the inferior articular processes are medial to the superior processes of the lower (caudal) vertebra.
2.4.1 Lumbar Articular Facet Width and Height Constraints
Many articles have been written on the generalized lumbar vertebral geometry, but
relatively few have been written specifically detailing the unique geometry of the lumbar facet
joints. The most comprehensive paper in the published literature specifically dealing with spinal
facet geometry was by Panjabi et al. (1993). The nomenclature used by Panjabi is shown in
Figure 7.

14

Figure 7: Mnemonics for Each Facet Dimension (Panjabi et al., 1993)

As previously discussed, on the bony vertebrae the inferior facet is convex and articulates
within the concave superior facet of the vertebrae immediately below it. A sensor designed to
capture facet loading must fit within the width (FCW) and height (FCH) of the facet. As
demonstrated in Table 1 and Table 2, both the facet widths and heights generally get larger as one
moves down the spine. A sensor designed to fit within the L2-3 through L5-S1 facet joints must
maintain the active sensing elements within a width (FCW) of approximately 12.0 mm and a
height (FCH) of 15.1 mm. The width was calculated by taking the smallest of the average of the
left and right widths at the L2-3 articulation (the L2 inferior facet) minus one average standard
deviation of the error. The height was calculated with the same method applied to the respective
heights (L3 superior facet).

15

Table 1: Means (and Standard Errors of the Mean) of the Inferior Articular Facet
Dimensions (Panjabi et al., 1993)

Table 2: Means (and Standard Errors of the Mean) of the Superior Articular Facet
Dimensions (Panjabi et al., 1993)

2.4.2 Lumbar Articular Facet Curvature
The facet joints exhibit a curved, and in some specimens, sharply curved geometry.
Although these joints are generally described as planar, the joints are most often either
hemicylindrical or boomerang shaped in horizontal section (Taylor & Twomey, 1986). As one
progresses down the lumbar spine, the angulations of the facets relative to the sagittal plane (a
vertical plane which passes from front to rear dividing the body into right and left sections)
become steeper allowing more relative flexion and extension motion and less rotation. These are
illustrated graphically in Figure 8 and histologically in Figure 9.

16

Figure 8: Shape and Inclination of the Lumbar Facet Joints (Taylor & Twomey, 1986;
White & Panjabi, 1990; Van Schaik et al., 1985, 1997)

Figure 9: Histological Cross Section of Lumbar Spine (Giles & Taylor, 1984)

17

Investigation of existing sensing and measurement technology as discussed in Chapter 3
revealed that all contemporary grid or array based pressure or force measurement systems
performed extremely poorly when curved to physiologically relevant radii. It was therefore
deemed important to try and determine some descriptive measure of facet curvature. This
methodology and the results are discussed in Chapter 5.
2.4.3 Lumbar Articular Facet Thickness Limitations
The lumbar articular facet joints are essentially a virtual space in that they are a concave
and a mated convex structure separated by only a thin layer of synovial fluid. There is no true
empty or open space between the two sides. This is shown on the left side of Figure 9 and below
in Figure 10.

Figure 10: Transverse section of L3-4 Joint (Taylor & Twomey, 1986)

Relative to this point, it should be noted that histological specimens such as the one
shown in Figure 10 are not loaded with the nearly 40-60% of body weight above the lumbar spine
that would be seen in vivo. This physiological axial compressive loading would further compress
the joint. As also noted in the right side of Figure 9, Figure 11 and Table 3, the articular cartilage
18

(convex and concave sides) and the synovial fluid combined are typically no more than
approximately 3 mm thick (Giles & Taylor, 1984; Cramer et al., 2000).

Figure 11: Lumbar Zygapophyseal Joint Thickness “A” (Cramer et al., 2000)

Table 3: Zygapophyseal Joint Thickness (PT#=subject number, RL# and LL#=right and
left lumbar spinal level, O#=observer number, SD=standard deviation)

19

Given the extremely small space of the facet joints, it is desirable to have a sensor that is
as thin as reasonably possible. A sensor that is too thick will artificially distract the joint. The
only research that was located in the literature that quantified the effect of intra facet sensor
thickness has been the work by Hedman (1992, 1995). Figure 12 from Hedman (1992) shows
that as the thickness of the sensing element is increased, an exponential increase in the induced
load in the facet joint occurs. Based on this work and personal discussions with the author, it was
deemed desirable to have a sensor with a total thickness of no more that 0.8 mm. At this
thickness, and with some specimen variability, the induced sensor error will be on the order of
15-30 Newtons. Although any error is undesirable, a space for the sensor is being created where
one did not exist before and some amount of induced force (error) is to be expected.

Figure 12: Effects of Sensor Thickness (Hedman, 1992)

20

2.4.4 Environmental Constraints
The proposed testing involves utilization of cadaveric specimens. To attempt to maintain
the tissue integrity of the specimens, the specimens were kept moist with a warm saline solution.
Any new sensor design has to be able to withstand at least a mist-type salt water spray
environment.

21

CHAPTER 3: LITERATURE REVIEW AND
PRELIMINARY EVALUATION OF EXISTING
TECHNOLOGY

3.1 Rationale for an exhaustive literature review
The first stated aim of this research was to review the applicability and accuracy of all
currently published methods and techniques for measurement of intra-articular force, pressure
and/or contact area. Given the relatively large number of companies making thin film sensors
and a 40+ year history of published papers discussing measuring facet loads, it was anticipated
that current technologies could be applied to accurately measure both the temporal and spatial
distributions of loads with the facet joints. This was not case.
Relative to using various sensors to measure intra-articular loads (knee, hip, elbow, ankle,
canine, bovine, porcine, etc.), a very large amount of research has been published in highly
regarded (high impact factor) peer reviewed publications that is outright inaccurate and in some
instances, so fatally flawed in its methodology as to be at best totally useless and at worst
misleading. This bold and sweeping statement is supported by an inordinate amount of evidence
as documented in the below exhaustive and methodical review of the related literature. The lack
of viable existing technology, compounded by inaccurate information in the published literature,
was a driving force in developing a new method for accurately measuring intra-articular loads.
3.2 Biological joint force and area measurement methods - Overview
As early as 1904, loading on the facet joints was being considered as a source of lumbar
spine pain. Early research on vertical spine loading determined the facets carried no load (Fick &
von Bardeleben, 1904, Spurling, 1953), were “important force-bearing structures” (Kelly, 1958),
or they carried around 15 % (Nachemson 1960, Nachemson, 1963) of the total spinal load.
Further testing by Nachemson (1963) examining the contribution of each element of the lumbar
22

spine to load distribution drew a contradictory conclusion to his prior research and concluded the
“arches and facets do not have any significant effect on the tensile strains in the annulus fibrosus
from a forward, backward, or sideward tilt of 5 degrees” (Nachemson, 1963). As recently as the
early 1970s, the facets were stated in text books to carry no load (Basmajian, 1971). Individual
functional spinal unit (FSU – superior vertebrae-disc-inferior vertebrae and ligaments) testing
under normal body weight showed the facets carried approximately 16% of the joint load in the
normal standing posture, but this was reduced to zero with a slight flexed posture replicating an
erect sitting posture (Adams & Hutton, 1980). Eccentric compressive loading of the lumbar spine
by Yang and King (1984) found that normal facets carried between 3 and 25% of the spinal load,
increasing to as much as 47% for arthritic facets.
These analytical and indirect methods ultimately gave way to more empirical attempts to
directly measure the facet loads and contact areas. As early as 1970, pressure sensitive films
were being used in biological joints (Frisina & Lehneis, 1970). Early attempts to quantify
articular joint contact area included using cold-curing acrylic polymers poured into the joint
which were allowed to set under joint load (Simon, 1970), injecting dyes into the loaded joint
which stained the areas of non-contact (Goodfellow et al., 1976, Townsend et al., 1977), micro-
indenting thin plastic materials (Ahmed, 1983), drilling holes into the joint and monitoring the
pressure (Adams, 1978; Indaba, 1989), and drilling and implanting thin 0.38 mm (0.015 in) thick
piezoresistive based sensors into the articular surface of the disarticulated joint (Brown, 1982)
and re-approximating the pieces. The predominant methods have involved the use of strain
gauges on the exterior surface of the joint (another indirect method), pressure sensitive film
within the joint, resistive transducer arrays (such as Tekscan) or capacitive transducer arrays
(such as Novel). Each of these more prevalent methods will be discussed.
23

Many force or pressure measurement techniques in biomechanics were driven by early
attempts at analysis of force and pressure distributions under the foot (gait) and torso (bed sores).
A number of excellent reviews of pressure and force measurement techniques predominantly
related to foot pressures or loads have been published (Lord, 1981, Lord & Reynolds, 1986,
Alexander et al., 1990, Cavanagh et al, 1997, Ferguson-Pell et al., 1992, Ferguson-Pell & Cardi,
1993, Graf, 1993, Rosenbaum & Becker, 1997, Urry, 1999, Orlin & McPoil, 2000). The reviews
by Cobb and Claremont (1995) and Urry (1999) provided the most detailed reviews of
information on various sensors with tables of data for each method including number of sensors,
sensor size, full scale range, non-linearity, hysteresis, sensitivity and frequency response.
3.3 Strain gauges
One of the first studies to empirically measure the loads on the posterior spinal elements
(Ewing et al., 1972) did so by attaching strain gages to the vertebral bodies in a manner similar to
that shown in Figure 13.

Figure 13: Strain gauges on exterior of lumbar facet (Sawa & Crawford, 2008)

These early studies were driven by a need to reduce aircraft ejection seat related lumbar
vertebral body fractures and led researchers to discover what they termed a dual load path in the
spine when subject to vertical acceleration. Ewing et al. (1972) found the spine was able to take
24

50% greater force prior to fracturing when extended compared to when it was flexed. Although
they did not know the exact mechanism for the observed reduction in fractures, they did note “it
was possible that some of this load was being transmitted via the articular facets.” The tolerance
of vertebrae fracture was dependent on the orientation of the spine, with the extended position
having lower forces in the vertebral body. It was thought that the posterior facets acted as motion
limiters, preventing the posterior elements from displacing as much as the anterior elements.
Research two years later (Prasad et al., 1974, Prasad & King, 1974, King et al., 1975)
further delineated and expanded on the findings of Ewing. Prasad also affixed strain gauges on
the vertebral bodies and neural arches of cadavers undergoing vertical accelerations under various
degrees of spinal flexion or extension. This method of inferential measurement was necessary as
the authors recognized the difficulty in measuring loads within such a small and irregular joint
and stated “several techniques were used to deduce the load-bearing capability of the articular
facet, since direct measurement of force in a limited space environment is still beyond the state of
the art.” The load in the posterior neural arch was found to increase by a factor of four when the
spine was extended under the same loading condition. This suggested “a shift in load from the
vertebral body to the lamina and facets.” The authors concluded the articular facets are capable
of bearing compressive and tensile load. Further strain gauge type measurements were conducted
(Hakim & King, 1976, Suezawa et al., 1980) confirming that the posterior elements, including the
facets, were significant load bearing elements in the spine, especially in extension and
compression, carrying as much as 25% of the load. The strain gauge method of indirectly
measuring the facet load has also been employed during in vivo canine spine testing (Butterman
et al., 1991a, 1991b, Nagata et al., 1993, Luo et al., 1996).
Utilizing strain gauges to quantify facet forces remains in current use (Schendel et al.,
1993, Sawa & Crawford, 2008, Zhu et al., 2008). Although these authors extolled the virtue of
25

the strain gauge method allowing the facet joint capsule to remain intact, they calibrated their
strain gauges by cutting the facet capsule, completely disarticulating the specimen and then
applied known loads to the facet. This would necessarily, as it is not being evaluated in the
calibration, imply the facet capsule contribution is negligible.
Zhu et al., (2008) developed a more detailed calibration method that considered the
effects of strain coupling at the L3-L4 facet joint. They referred to their set up as the EAS (extra-
articular strain) method. Even properly done, this method is inherently limited in that it is an
inferential measurement where the loads on the inside of the facet are calculated by strains
measured on the outside of the facet. These external loads could be affected by the joint capsule
which attaches to -- and may also strain the bone to which the gauges are affixed. From their
direct calibration, they did find strain coupling between the facets on both sides of the facet joint
when a compressive force was applied on a unilateral facet joint surface as seen in Figure 14. A
linear relationship between the strain and facet load existed in all the specimens. Accuracy
assessment of the EAS technique demonstrated errors in the peak facet joint contact forces from 4
to 26%.

Figure 14: Strain coupling in right facet from loading of the left facet (Zhu et al., 2008)
26

As seen in Table 4, no significant difference was found between the EAS facet contact
force of the two sides in all directions except for lateral bending where the right facet load was
greater than the left facet load. Using the EAS technique, a clear pattern of facet contact force in
axial rotation was observed in all specimens. In right axial rotation, the left facet joint sustained a
compressive load, and the right facet sustained a tensile load. In a similar way, in left axial
rotation, the left and right facets sustained tensile and compressive joint contact forces
respectively as shown in Figure 15 and Table 4 (negative force is compression and positive
tension). The facet load patterns were variable or inconsistent in flexion, extension, or lateral
bending.

Figure 15: L3-L4 Facet forces under a 7.5 Nm axial moment (Zhu et al., 2008)

27

Table 4: L3-L4 facet loads with EAS technique and 7.5Nm of bending about each axis (Zhu
et al., 2008)

The strain gauge method, depending on the calibration method, may also require post test
destructive testing of the specimen to determine the Young’s modulus to convert the measured
strain to a more meaningful stress value. The strain gauge method does have the advantage of
allowing the facet joint capsule to remain intact, but the relative contribution of each facet
element to the facet load still remains unknown.
3.4 Pressure sensitive films – Ink Based and Transfers
Pressure sensitive films have been used to quantify both the joint contact forces or
pressures and joint contact area. One of the first descriptions of a pressure sensitive film in a
biomechanical application was by Brand and Ebner in 1969. They used microcapsules with a
range of shell strengths on a sheet of polyurethane foam to examine hand and foot pressures.
Despite discussing the advantages and disadvantages, very little technical data is included.
Frisina & Lehneis in 1970 used a sheet with a mild acid placed against a second sheet with an
acid indicator. The authors noted that when pressure is applied, the reactants combine; the color
change that occurs is a function of pressure and can be sensed visually for immediate qualitative
28

information or recorded with a filtered densitometer for quantitative data. The response range
and sensitivity of the transducer varied with the time interval the composite was pressurized,
reactant type and concentration, along with type, thickness and number of layers of suspension
material. Testing was conducted in ranges between 2 and 1200 psi (Frisina & Lehneis, 1970).
The biomechanical literature was relatively silent on thin film pressure transducers from
1970 until 1980 when Fukubayashi and Kurosawa (1980) used commercially available pressure
sensitive film called Fuji Prescale Pressure Sensitive Film (Fujifilm Corp., Tokyo, Japan) to
examine the pressure distribution in the human knee. This thin film system became commonly
referred to as “Fujifilm.” It is a polyester based film that contains a layer of tiny microcapsules.
The application of force causes the microcapsules (2-26 µm diameter) (Liggins & Finlay, 1997)
to rupture, producing an instantaneous and permanent "topographical" image of pressure variation
across the contact area. The system consists of two pieces of film, one called the Transfer Sheet
(A-film) with the microcapsules on it and the other called Developer Sheet (C-film) with a
combined thickness of approximately 0.2 mm (Hale & Brown, 1992, Singerman et al., 1987) as
shown in Figure 16. The transferred image is preserved like litmus paper, where the color
intensity of film is directly related to the amount of pressure applied.

Figure 16: Fuji Prescale Pressure Sensitive Film system
29

However, the film has a specified range of pressures: below a lower threshold value of
pressure no stain is produced and above the upper threshold pressure a saturated stain results
(Liggins & Finlay, 1997). The Fujifilm instruction manual provides a series of calibration curves
for each film grade under two loading protocols, designated as “momentary” and “continuous”.
The continuous protocol consists of two 2 min intervals. The Continuous Pressure calibration
chart for Low Pressure LW Film is on the left side of Figure 17. The momentary protocol
consists of an increase in pressure up to the desired level, over 5 seconds, followed by a further 5
second hold at that pressure value. The Momentary Pressure calibration chart for the Low
Pressure LW film variant is show on the right side of Figure 17. The optical density at any
position on a test stain is compared to a set of eight color samples. Pressure values are obtained
using one of four calibration curves A through D, each of which corresponds to a range of
temperature and humidity (Liggins et al., 1995a, Liggins and Finlay, 1997, “FujiFilm Features-
Specifications,” 2010). However, the color is essentially only various shades of pink of variable
optical density (Liggins & Finlay, 1997, Liau et al., 2001)

Figure 17: Fujifilm Low Pressure LW film calibration charts (Left is continuous pressure
and right is momentary)
30

The quantitative data of contact area and contact pressure distribution can be estimated
by recovering the spatial information from the Developer Sheet using digital image analysis. It
should be noted that most contemporary users have abandoned the factory recommended
calibration method in favor of a pixel-by-pixel scan of the resultant stain and then constructing a
polynomial relationship, typically fifth-order, between pixel-value and applied pressure (Liggins
et al., 1995a, 1995b, Liggins & Finlay, 1997, Hale & Brown, 1992, Caldwell et al., 1993).
Fujifilm has a long and continuing history of utilization in biomechanical testing. This is
likely attributable to it being easy to use, quick to set up, relatively unobtrusive (thin),
customizable (it comes in roles that can be cut to any size or shape), comes in a range of
sensitivities and is relatively inexpensive. Fujifilm has been used in nearly every major joint of
the body including measuring the loads inside of knees including human (Fukubayashi &
Kurosawa, 1980, Huberti & Hayes, 1984, Haut et al., 1995, Hsu, 1989, Collier, 1991, Marder,
1993, Lee, 1994, Szivek, 1996, McDermott, 2008), rabbit (Haut et al., 1995) and cat knees
(Ronsky et al., 1995). The film has also been used in testing external loading of the foot (Aritomi
et al., 1983), and internal loads of the ankle joint (Sangeorzan et al., 1990, 1992, Wagner et al,
1992, Wang et al., 1995), hip (Afoke et al., 1987, Bay et al., 1997, von Eisenhart et al., 1997,
1999), wrist (Tencer et al, 1988) and elbow (Stormont et al, 1985).
Quantification of facet loads has also been done with Fujifilm since at least 1983 (Lorenz
et al., 1983). These authors noted that direct measurements of facet loads had not been reported
up to that point. They placed 12x25 mm rectangular pieces of Fujifilm in “moisture proof
packets”, the posterior portion of the facet capsule was incised to the height of the joint at either
the L2-3 or L4-5 levels, the packet film was inserted, and the spine was subjected to neutral
position (end plates parallel) and extension to between 6-8 degrees. In each position, tests were
performed at compressive loads of 196.2, 392.4, 686.7, and 1373.4 Newtons. Fujifilm was used
31

“to record the facet pressures and contact area during the compressive loading.” The authors
noted “the film was checked for accuracy and reproducibility both in an out of the moisture proof
packets.” Post test stains were scanned in progressive 3 mm diameter circles. Average pressure,
total facet load, peak pressure and contact area were reported. Examples of the results for the L4-
5 level are shown in Table 5.
Table 5: Fujifilm results for L4-5 facets (Lorenz et al., 1983)

In 1984, researchers (Dunlop et al., 1984) investigated the relationship between disc
height, posture and facet loading, using the Fujifilm technique describe by Lorenz et al. (1983).
Their Fujifilm was encapsulated in “Sellotape to prevent damage by moisture.” Their tape-
Fujifilm-tape had a thickness of 0.3 mm. The authors noted film was cut to fit each facet, but
they do not specify the sizes. Some of their results are shown in Figure 18.

Figure 18: Peak facet pressure at various postures (Dunlop et al., 1984)
32

Hedman (1992) used Fujifilm in conjunction with Force Sensitive Resistors (FSR) to
quantify facet loads (this will be further discussed in the subsequent FSR section). However, the
film in this study was only used to establish the contact area or “foot print of the loaded surface.”
Calibration and testing was constructed with the packaged Fujifilm and FSR. As recently as
2005, Fujifilm was being used to quantify facet loads (Wiseman et al., 2005) in the intact spine
and in spines implanted with a device designed to prevent narrowing the spinal canal and neural
foramina during extension. However, beyond indicating they used Fujifilm, no description of the
size of film, packaging or specifics of the film calibration were provided. Some of their results
are shown in Table 6.

Table 6: Facet loads for intact and implanted specimens (Wiseman et al., 2005)

Despite the many advantages of using Fujifilm, there are a number of significant factors
that limit its utility. First and foremost, Fujifilm has no temporal sensitivity and limited spatial
sensitivity because is only capable of capturing the total aggregate load and distribution response
during any load applications. If the location of the load shifts or moves during loading, the
Fujifilm will effectively sum these effects, reporting one distribution which is the aggregate of all
temporal and spatial shifts in loading up until that point. Additionally, if there are multiple
33

loading cycles on one location, the film will only record the highest load it was subjected to at
that location.
As this system involves essentially transferring stain onto a developer sheet, any liquid
such as blood, saline, etc. that is allowed on or in between the unsealed two-sheet system may
distort or wash out the stain transfer. To help reduce this, most researchers incorporate some
form of sealing or packaging of the two film system. However, research has shown (Liggins et
al., 1995) that sealing the film with even very thin Tegaderm (60 µm) film had a “significant
effect on the film’s response under one of three experimental ambient conditions.” They did note
that using sealed film for producing calibration data may only be required when higher pressure
resolution is required. At lower resolution, the effect of any errors caused by using unsealed film
may be less than those incurred by the rendering procedure.
The manufacturer’s calibration curves in Figure 17 indicate the film is sensitive to
loading rate with the noted 5 second “momentary” calibration curves or the two minute
“continuous” calibration curves. However, testing by Rudert et al., (1987) showed “the response
is highly time dependent, especially in the range of zero to five seconds.” Analysis of their
graphical results shows an approximately 14% increase (from 0.7 to 0.8) in color density for the
same loading pressure when the time to reach peak pressure was decreased from 5 seconds to 1
second as shown in Figure 19. Loading rate was noted to be of “critical importance in the
evaluation of experimental results” and individual calibration curves must be produced
corresponding to the specific loading times being studied. This would only be possible with some
form of a priori knowledge of the loading rate from some other source, as Fujifilm itself does not
transduce information relative to loading rate.

34

Figure 19: Effects of loading rate on color density (Rudert et al., 1987)

As also noted by the manufacturer’s calibration curves in Figure 17, the film is sensitive
to both temperature and humidity. As such, the calibration and testing much be conducted at
similar ambient conditions for accurate measurement. One group (Liggins et al., 1995) has
already suggested that packaging may trap the local relative humidity at the time of packaging,
potentially altering the calibration. It is either not addressed at all or it is not clear from many
studies that this “packaging” effect is being accounted for in the data. For example, Lorenz et al.
(1983) described “the packets containing the film were then removed from the facets and
inspected for moisture damage or artifacts due to the insertion or removal.” It is not clear if
Lorenz et al. (1983) used calibration data applicable to packaged film. Subsequent researchers
(Dunlop et al., 1984) encapsulated their film in “Sellotape to prevent damage by moisture.” They
noted the film was calibrated for the laboratory, but they did not specify if this was done on the
encapsulated film. Singerman et al. (1987) noted that packets were created and sealed to avoid
“contamination caused by the wet environment which most biomechanical experiments involve.”
They explicitly noted the packets were used for both testing and calibration. Hedman (1992) who
was only using the film to establish the contact area, did indicate all his calibrations were done
35

with the packaged film. Sterilization has also been found to alter the calibration values, but in a
predictable and repeatable way (Liggins et al., 1995).
Liggins & Finlay (1992) showed that Fujifilm exhibits a “development period” of
approximately 50 hours following a stain production. During that 50 hour window, stain density
was observed to fluctuate +/- 5% of the original immediate post test value. Subsequent testing by
Liggins et al. (1994, 1995a, 1995b, 1997) always allowed for the 50 hours development period.
The film’s performance is recognized as being problematic when the contact surface
possesses two principal planes of substantial curvature, a feature of virtually all biological joints.
Because the paired film sheets can freely bend only in one curvature plane, conformation to a
bicurvilinear surface requires that the sheets fold or “crinkle” locally. This gives rise to high local
stresses at or near the fold edges, and results in discontinuous artifactual streaks in what otherwise
would usually be relatively smooth gradations of staining intensity (Caldwell et al., 1993).
Fujifilm was found to have similar calibration values when used in constant radius curved
surfaces down to a radius of 7 mm compared to calibration values obtained on flat surfaces
(Hasler et al., 1996).
Tests which evaluated the ability of Fujifilm to detect contact stress gradients using a
bench top protocol, described by the authors as nominally representative of typical in-vitro
articular contact measurement, found the film was able to transduce contact stress gradients up to
about 9 or 10 MPa/mm (Hale & Brown, 1992).
The manufacturer of Fujifilm quotes the accuracy as +/- 10% at 23 C and 65% relative
humidity (www.fujifilm.com/products/prescale/guide/feature/spec.html). In the literature,
Fujifilm has a reported load accuracy of 10-15% under ideal conditions (Fukubayashi &
Kurosawa, 1980, Lorenz et al., 1983, Singerman et al., 1987, Liggins & Finlay, 1992, Hale &
Brown, 1992, Muriuki et al., 2009) and this “bench top performance probably represents the
36

upper bound on accuracy of contact stress gradient detection for in situ measurement” (Hale &
Brown, 1992). Other researchers (Wu et al., 1998) have reported that the measured joint contact
pressures may contain errors as large as 14-28 percent using Fujifilm.
3.5 Force Sensitive/Sensing Resistors (FSR)
In the biomechanical literature, predominantly two types of force sensing (or sensitive)
resistors (FSR’s) have been used - FSR’s produced by Interlink Electronics (Interlink Electronics,
Camarillo, CA, USA), hereafter referred to FSR-I, and those produced by Tekscan (Tekscan,
South Boston, MA, USA) hereafter referred to as FSR-T. Both these companies produce FSR’s,
but with different underlying technology. Therefore, each system is discussed separately.
Other early forms of force sensitive sensors where primarily designed to examine the
body segment to surface interface such as the foot-to-ground or hand-tool interactions. Schwartz
and Heath (1947) were one of the first to use an individual pressure sensor on the undersurface of
the foot for gait analysis. Their sensors were poorly described but consisted of “paper discs” also
called or possibly in addition to, “carbon discs” sandwiched between phosphor bronze electrodes.
The whole sensor was indicated to be 0.090 inches (2.29 mm) thick. Brand & Ebner (1969)
reported work on a sensor “based on two flexible conductive membranes separated by a third
flexible membrane which is compressible and has an electrical resistance which varies with the
degree of compression” that was used for foot and hand-tool pressure measurements. Collis &
Jayson (1972) developed a strain gauge based sensor to measure pedal pressure that was
approximately 2 mm thick. Schwartz and Heath (1947), Brand & Ebner (1969) and Collis &
Jayson (1972) did not provide any calibration or performance data on their sensors. Although
Collis & Jayson (1972) noted “calibration using known pressures showed that over the required
range, a linear relationship existed between the applied pressure and the deflection of the recorder
trace.”
37

3.5.1 Interlink FSR-I
Interlink’s FSR’s, FSR-I, are polymer thick film (PTF) devices that exhibit a decrease in
resistance when increased pressure is applied to the surface of the sensor. A typical FSR-I sensor
construction is shown in Figure 20.

Figure 20: FSR-I construction (www.interlinkelectronics.com)

The manufacturer notes “they are more appropriate for qualitative rather than precision
measurements.” The manufacturer indicates device thickness between 0.20 to 1.25 mm, part-to-
part force repeatability of +-15% to +- 25% of established nominal resistance, single part force
repeatability of +- 2% to +- 5% of established nominal resistance, a force resolution of better than
0.5% full scale, and a mechanical device rise time of 1-2 msec. When plotted on a log resistance
versus log force graph, the sensor gives an approximately linear response as shown in Figure 21
and is stated to approximately follow an inverse power-law characteristic.
(http://www.interlinkelectronics.com/force_sensors/products/forcesensingresistors/standardsensor
s.html).

38

Figure 21: FSR-I calibration curve (www.interlinkelectronics.com)

FSR-I have been used extensively in hand grasp application such as instrumented gloves
to measure finger contact forces (Jensen et al., 1991, Park & Yun, 1996, Yun, 1997, Coronado et
al., 1999, Slingsby, 2001, Nikonovas, 2004, Kong, 2004), prosthetic hand development (Tura et
al., 1998), in foot contact pressure measurements (Stone, 1990, Baumhauer, 1997, Maalej et al.,,
1989, Smith, 2002), and even in the developing of an electronic larynx (Choi et al., 2001).
Maalej et al., (1988) noted the sensor was constructed of conductive polymer film
composed of molybdenum disulphide, a resin and possibly powdered carbon thinned with resin
all suspended on a Mylar (bioaxially-oriented polyethylene terephthalate) film. The bottom layer
is formed by two interdigitated silver conductive patterns printed on opposing Mylar film as seen
in Figure 22. When the film is pressed against the conductive pattern more contact between
layers occurs and the resistance is decreased.

39

Figure 22: Interlink Electronics FSR-I sensor construction (Maalej et al.,, 1988)

Maalej et al., (1988) tested the 1.5 cm diameter single cell sensor at different
temperatures. They found that the single cell FSR-I exhibited 8% hysteresis (of 2 MPa full scale)
and nonrepeatability was less than 7% of full scale with four calibration cycles. The sensitivity of
the sensor ranged from 0.08 mV/kPa at low pressure to 36 mV/kPa at high pressure; thus, 450
times less sensitive at high pressure. Temperature related drift was -0.5%/ºC. Loading each
quadrant of the sensor compared to the whole sensor loading suggested that the sensor acted as
though the four quadrants’ resistance were in parallel as seen in Figure 23 . The authors noted the
benefits of the FSR-I in that it is thin, flexible, rugged, and low cost. Additionally, it operates
with simple and power efficient electronics. They found similar results for the FSR-I array used
in a shoe insole for gate analysis (Maalej et al.,, 1989).

Figure 23: Area affects on FSR-I resistance (Maalej et al., 1988)
40

In a similar fashion, Pax (1989) showed that the FSR-I’s response was not only a function
of applied load, but also area of contact as seen in Figure 24. These authors were also one of the
first to encase the FSR-I in another material to facilitate usage.

Figure 24: Variation in FSR-I output with contact area (Pax, 1989)

A paper by Yaniger (1991) [Vice President/Chief Scientist for Interlink Electronics,
manufacturers of the FSR-I] indicated that the FSR-I was neither a pure force nor a pure pressure
sensor- its output is related to the portion of sensor surface addressed, the type of actuator (hard
or soft) and the ratio of the area of the applied force to the electrode pitch (i.e., the number of
fingers covered by the actuator). As is true for most force sensors, the FSR can be used as a
pressure sensor when the area of applied force is large compared to the FSR dimensions. The
sensitivity to the area and distribution of the force means that either the FSR-I must be used as a
qualitative sensor, or the force footprint must be held constant in area, position, and distribution.
With increasing force applied to the sensor, the force drops following an approximate power law.
Repeatability of an individual FSR-I cycle-to-cycle is better than 2% and variation of between
different FSR-I’s of the same type are typically +/-15%. The FSR-I was stated to be a “slow
41

device (typical mechanical rise time 1-2 ms), and is relatively insensitive to vibration and
acoustic noise pickup.” There is a tendency for the sensor to drift under static load, but the drift
is reversible with a time constant “on the order of minutes” and totals 10-20% of the initial value.
Accuracy was limited to +/-15% and the author stated that “if accuracy better than 10% or so is
needed, strain gauges and other piezoresistives are the obvious choice.”
Jensen et al. (1991) sought to improve the accuracy of the FSR-I by placing an epoxy
dome over the sensing area so that any applied force effectively loaded the entire sensing area.
The 12 mm diameter sensing area FSR-I was used. The construct is shown in Figure 25 .
Sensors were calibrated by pressing them against a strain gauge dynamometer at 10 equally
spaced force levels between 0 and 30N in random order, with a 30 second rest period between
each calibration point. Calibration points were fitted to a second-order polynomial using linear
regression. Load was applied by flat circular styli with surface areas of 25, 98 and 1100 mm
2

attached to the end of the load cell. A convex curved surface (16 mm radius) was also used. The
output for sensors both with and without epoxy domes with varied contact areas is shown in
Figure 26. These data confirmed that the response of the sensors containing epoxy domes was
not affected by contact area, while output from the unmodified FSR-I varied considerably with
contact area.

Figure 25: Domed FSR-I (Jensen et al., 1991)
42

Figure 26: (A) Response of sensor with an epoxy dome for input forces applied using
different size surface areas, including a curved surface. (B) Response of a sensor without an
epoxy dome for forces applied using different size surface areas (Jensen et al., 1991).

Dynamic calibration was conducted by “slowly increasing” the applied force until a 30N
load was achieved and then “slowly” unloading over 10 seconds. Sensor step input response was
investigated with 18 and 29N loads. Time constants (to 63% of steady state level) were 0.57 ms
+/-0.06 ms for 18N and 0.52 ms+/-0.13 ms. “Slow dynamic loading” response was checked by
loading the sensor at “one pinch every 2s for 10s” (0.5 Hz). Force levels from 0-30N, 0-40N and
0-50N were tested dynamically. Average error increased with increasing static loads from
1.0N+/-0.8N at 0-30N loads to 2.1N+/-1.2N at 0-50N loads. Dynamic testing produced a similar
response with average error increasing with increasing dynamic load from 1.2N+/-0.7N at 0-30N
loads to 3.2N+/-1.5N at 0-50N loads. In discussion, they noted 20% variability in the offset
voltage.
Ferguson-Pell (1992, 1993), comparing an FSR-F (Force Sensitive Resistor from FSA
sensors, Verg Inc.) to an FSR-T system over a pressure range from 0 to 160 mm Hg, found the
43

FSR-F system had a hysteresis of 19% and a 3.3% change in readings under constant pressure at
2 minutes and 4.4% change at 10 minutes.
Hedman (1992) was one of the first reported usages of a force sensitive resistor for in
vitro biomechanical application. He used the FSR-I to quantify the loads in the L4-5 facet joints
via his ultrathin load cell (ULC). Hedman’s (1992) ULC’s used FSR-I’s that were 18 mm in
diameter with a single 9.5 mm sensitive diameter and they were 570-620 µm thick. As stated by
the manufacturer and shown by Hedman (1992, 1995, 1997) the electrical resistance of the FSR-I
is both a function of the magnitude of contact and the area of contact. As discussed previously,
Hedman utilized Fujifilm to quantify the area of contact. The area of contact was then compared
to packaged sensor force-time calibration curves obtained at each of three known contact areas.
Force values were linearly interpolated to the actual contact area. It was observed that the force
response was quite nonlinear with significant viscoelastic behavior, but with application of
standard linearization techniques, the response of the transducer was characterized and
coefficients of determination were found to be typically very close to one (Hedman, 1992).
Hedman accounted for the viscoelastic nature of the sensors by adjusting the loading ramp and
dwell times to minimize the effects. As seen in Figure 27, Hedman (1992) found that sensors
between 0.57 mm to 0.75 mm of thickness could induce between 22 N to 46 N of load in the facet
joint. This is similar to the results from Lorenz et al. (1983).

44

Figure 27: Induced facet force with sensor thickness (Hedman, 1992)
Hedman (1992) also found that 5-60% of the peak applied load (1000 N) was measured
in the facets at 4degrees of extension and 0-30% of the peak applied load at 8 degrees of flexion.
Hedman noted no drift using his calibration and testing methods.
Smith (1994) tested the 0.2 inch and 0.5 inch FSR-I’s and indicated they were “shown to
be of little use in precise measurements, as the manufacturer states. Also they poorly measure
dynamic, rapidly changing force. They are suitable for threshold switches, sensory feedback, and
any force measurement that does not require great accuracy.” Smith refers to “memory”
problems based on differences in resistance values (response) to load versus unload at the same
force. This would actually appear to be hysteresis and is shown in Figure 28. They also noted, as
shown in Figure 28, that the sensors’ response was dependent on how long they had been under
power. A likely related, but which was as a separate property, was that there was considerable
“settling time.” Under a constant load of 413 grams, at time 0 seconds, the 0.5 inch sensor read
15.2 kilohms, and at 300 seconds, it had dropped 24% to 11.5 kilohms. Although not attributed
by Smith, both these phenomenon are likely drift issues. Additionally, Smith did not attempt to
derive any mathematical equations for calibration curves, such as log-log curves.
45

Figure 28: FSR-I response variation with sensor warm up (note the up then down values on
the x-axis)(Smith, 1994)

Park and Yun (1996) provided very little methodological information on how they
calibrated their FSR-I’s, but they used exponential regression to determine a calibration curve of
the form Force (kg) = exp (-2.18+1.25mv). The coefficient of determination for the force
calibration regression was 0.99.
Tura et al., (1998) worked on integrating FSR-I’s to control the strength of grip on
objects in a prosthetic hand. Very little data was provided on their calibration methods, but they
did indicate they tested 2.54 cm diameter FSR-T by applying increasing loads with a 1.27 cm
diameter metal probe (presumably flat ended). They noted that the sensor only exhibited a
straight Log Resistance versus Log Force graph over a certain range of force values (not
specified): with very high forces tending towards a saturation value and with very low forces
producing more rapid changes in resistance. The sensitivity from the manufacturer was noted to
be 0.007 Kg/cm
2
.
46

Coronado et al., (1999) stated one of the aims of their work was to “modify a simple
force sensing resistor to obtain a reliable and repeatable fingertip force recording during pinch
grip.” As seen in Figure 29, they took a 12.7 mm diameter FSR-I mounted over a curved “rigid”
Cibatool surface with a rubber membrane on it (NFC 18-415-Type I, 18x31x8mm). Over the
sensor, another membrane was placed with the same circular dimensions as the active sensor area.
The top layer was another rubber membrane. The sensor was conditioned by the loading the
sensor 5 times for 10 seconds with increasing force until a force of 30 N was reached. Applied
force was recorded with a strain gauge dynamometer. Calibration was done both in a single force
application step load up to 15 N, held for “about” 9 seconds and in step-like force application of
5N, 15N, 20N, and 25N with a stop of “about” 3 seconds at each force level. Calibration results
are shown in Figure 30.

Figure 29: Modified FSR-I configuration (Coronado et al., 1999)

47

Figure 30: Modified FSR-I calibration results (Coronado et al., 1999)

Coronado et al. (1999) concluded that the modified FSR-I had minimum inter-subject and
inter-trial variability, and a good inter-trial repeatability; however, no numerical results to support
these statements were provided. They also noted that the sensor nonlinearities such as saturation
and hysteresis become significant when large transient forces occur.
Hollinger and Wanderley (2006) evaluated the FSR-I model 042, FSR-T A201, and the
FSR-L (Force Sensitive Resistor from LuSense) 12mm PSR Standard 151, with static, referred to
as “dead-weight tests” and ramp loads. Static tests were conducted at loads of 50g-1000g and for
either 240 seconds or 1200 seconds. Drift from load removal was assessed by placing two
weights of 500g and 200g on the sensor for 30 minutes, after which only the 500g mass was
removed and data was recorded. Ramp testing was conducted on a material testing machine with
five cycles from 0-20N and a final cycle from 0-25N, all at a constant compression rate of
48

0.1mm/minute to examine each sensor’s hysteresis. Unfortunately hysteresis was only
qualitatively described. Their results shown in Table 7 demonstrated that the FSR-I had
comparable drift as the FSR-L and FSR-T, a quicker response time, and was midway between the
two other technologies relative to hysteresis. Although not discussed in this research, another
positive attribute of the FSR-I is that it is substantially less expensive than the FSR-T.
Table 7: FSR performance results (Hollinger & Wanderley, 2006)

3.5.2 Tekscan FSR-T
Tekscan sensors, FSR-T’s, are far and away the most commonly used systems. They
come in both grid-based arrays and single load cell configurations. The grid based sensors are
commonly referred to by their product lines as F-Scan, F-Socket, I-Scan, or K-Scan systems. The
K-Scan system is noted to be their joint analysis system. A representative sensor grid is shown
on the left side of Figure 31 and Figure 32. The single load cell configuration is typically referred
to as a Flexiforce system and is shown on the right side of Figure 31. Per the company web site
(http://www.tekscan.com/technology.html), FSR-T sensors are available in a range of shapes,
sizes and spatial resolutions (sensor spacing). These sensors are capable of measuring pressures
ranging from 0-15 kPa to 0-175 MPa. Sensing locations within a matrix can be as small as
0.0009 square inches (.140 mm
2
); therefore, a one square centimeter area can contain an array of
170 of these locations.
FSR‐I
Interlink
Model 402
FSR‐L
LuSense
PS3 Std. 151
FSR‐T
Tekscan
A201
Static loads
Min drift 385s 3.05% 6.18% 10.34%
Max drift 385s 12.37% 11.74% 11.42%
Drift at 1875s 7.41% 6.14% 4.08%
Time to 90% Final Value 345s 427.5s 450s
Load removal drift 7.20% 17.30% 2.00%
Time to 90% Final Value 800s 350s 475s
Hysteresis (ramp loads) "Low" "Lowest" "High"
49

Figure 31: Grid based and single load cell sensors from Tekscan
(http://www.tekscan.com/flexiforce.html)

Figure 32: Cross sectional dimensions (mm) of one ink-covered grid conductor strip
(Hartmann et al., 2009)

The standard grid sensor consists of two thin, flexible polyester sheets which have
electrically conductive electrodes deposited in varying patterns. The inside surfaces of one sheet
forms a row pattern while the inner surface of the other employs a column pattern. The
intersection of these rows and columns creates a sensing cell, or what Tekscan calls a “sensel.”
50

The spacing between the rows and columns varies according to sensor application and can be as
small as ~0.5 mm.
Company quoted system performance is shown in Table 8. Resistance of the sensing
elements (sensels) reportedly varies inversely with applied load as seen in Figure 33. A standard
general purpose I-Scan system can sample 250,000 sensels per second. The system linearizes
sensor output into digital counts, or “raw” values on a scale from 0-255 (8-bit). Calibration
converts raw values into engineering units, such as psi or kPa.

Table 8: Tekscan typical system performance (http://www.tekscan.com/technology.html)

SENSOR PROPERTY STANDARD
Linearity <±3%
Repeatability <±3.5%
Hysteresis < 4.5% of full scale
Drift per log time 5%
Lag Time 5 µsec
Operating Temperature
15° to 140°F
(-9° to 60°C)
Thinness 0.004 in (0.1 mm)
Sensel Density
Up to 1,600 per sq. in.
(248 per sq. cm)
Pitch as fine as 0.025 in.
(0.6 mm)
Pressure Range
Up to 30,000 psi (207 MPa) (dependent on
sensor selection)

51

Figure 33: Tekscan variation in resistance with applied load
(http://www.tekscan.com/technology.html)

At a recent conference on the assessment of pressure measurement devices (Giacomozzi,
2010), a statement from the “PMD (pressure measurement device) management at TEKSCAN”
indicated Tekscan has developed new methods and guidelines for appropriate and optimal use of
the MatScan FSR-T. The application of these methods and guidelines ensures improved sensor
measurement accuracy, performance and ease of use. “The clinical community has a higher
tolerance for such variations in PDM sensors, and these are accepted in the majority of clinical
applications. However, for applications under scientific rigor, such as in research and
development, Tekscan recommends the use of an equilibrator (air bladder) device to assess,
optimize and maximize sensor uniformity and performance. Tests of individual sensing element
outputs and overall sensor output distribution should be performed with a fully-equilibrated
sensor.” Tekscan also recommends and provides equilibration procedures to significantly reduce
52

the sensel to sensel variation, if present. An air bladder equilibration of the MatScan sensor is
applicable when two main conditions are met: 1) sensor-bladder stabilization occurs after
“sufficient settling time” (not quantified); and 2) “trapped air within the senor is allowed to
escape (vent).” Although how it is vented or how to determine if this is needed or has been done
is not described. Of importance, “Tekscan does not recommend the use of an equilibrator device
to calibrate the MatScan sensor, or to conduct dynamic testing.” Since the study by Giacomozzi
(2009, 2010, 2010), Tekscan has introduced two new calibration procedures to the commercial
market. These new calibration procedures improve MatScan sensor measurement accuracy and
ease of use. These new calibration methods account for time-based compensations that occur
during dynamic events. They also significantly improve sensor output accuracy by reducing
hysteresis. “For optimal performance of the MatScan System, Tekscan recommends that the
calibration conditions mimic the events being recorded.” This last underlined quote is critically
important to recall while reading the rest of the literature review as the vast majority of published
research using the FSR-T does not “mimic the events being recorded” when calibrating the
sensors.
The first reported usage in the literature was for occlusal contact distribution
measurement (Maness et al., 1985, 1986, and 1987). This early version of the FSR-T was only
able to report positional and duration data. No load magnitude information was recorded. The
last paper in the series (Maness et al., 1987) noted that a second generation system was in
development which would record the magnitude of pressure at each contact. Harvey (1991) was
one of the first reported performance evaluation of the second generation-type FSR-T systems
and they examined the threshold force to activate a sensor and the reliability of the sensors over
multiple trials. They found contact force threshold of between 2.7 kg and 3.8 kg with a
53

statistically significant difference between the second and third load applications. They
concluded the sensors were valid (start of contact) and reliable when used twice.
Cranmer and Patterson (1992) was the first paper to quantify the load accuracy and
hysteresis of the FSR-T, using the F-Scan model. They indicated that Tekscan reported system
hysteresis of 10%, linearity of 5% for “up” (presumably increasing load), and creep of 10% (note
difference from the manufactures current values in Table 8). A “circular square piece of metal”
0.5 square inches flat actuator covered in Johnson and Johnson mole skin tape was used to apply
load to the sensor. No information was provided on their loading rate. Their sensor appears to
have been conditioned and calibrated at 25 psi (50lbs with a 0.5 in
2
applicator). Cyclic loads of
25, 50, 75, and 300 pounds were applied. Conspicuously absent from the results is accuracy
results. However, a review of the only figure shows an approximately 50% error at 50 pounds
and a greater than 60% error at 100 pounds. Hysteresis was approximately 10% below 50 pounds
and 20% at 100 pounds. “Above 100 pounds the output tends to flatten out.”
Ferguson-Pell & Cardi (1992, 1993) examined the load magnitude accuracy for
wheelchair-occupant pressure mapping. They found the Tekscan FSR-T system showed a
substantial hysteresis of +/-20% and average creep of 19%. The creep was found to be both
pressure magnitude and contact duration dependent. Creep was quantified for pressures of 50
mmHg and 100 mmHg after loading for 2 and 10 minutes. After 2 minutes, the FSR-T
demonstrated a creep of 7.5% at 100 mm Hg of applied pressure. There was 26% creep at 50
mmHg after 10 minutes of applied pressure. It was noted (Ferguson-Pell & Cardi, 1993) that
“Tekscan has developed both hysteresis and creep correction algorithms, but these were not
available for testing.”
Pavlovic et al. (1993) was the first to state they were testing the FSR-T for intra-articular
utilization (within the knee). They examined repeatability, hysteresis and accuracy. A layer of
54

3M #5480 (current specification from the 3M website indicate #5480 tape is
polytetrafluorethylene with a total thickness of 0.09 mm) tape was placed on each side of the
FSR-T and then it was “sandwiched” between two flat polished metal squares that matched the
sensor area. Load was applied with a material testing machine. The sensor was conditioned and
calibrated at 1,334 N (300 lb.). Three loading rates that replicated human walking, “knee
simulator speed,” and speed half ways between these two were used (numerical rates were not
defined). Repeatability testing found the sensor force increased with each additional trial
resulting in a 30N, or 2.3% of the total load range, increase over 30 trials. The sensors were less
accurate for unloading and the sensor hysteresis varied with applied load. Hysteresis was 5.6%,
8.0% and 8.5% at loads of 75N, 106N and 113, respectively. Lowest hysteresis was present at the
fastest loading rate. The sensors were noted to follow the patterns (emphasis added by original
author) of the applied force curves but the sensors read between 355 N and 400 N with a 111 N
applied load at the beginning and end of each test. The authors, hypothesizing that this was
related to the high calibration relative to measurement load, recalibrated the sensor at 111 N and
retested. At 111N calibration the sensor was accurate at the 111N load but measured 470-495 N
at 1,334 N. At an 890 N calibration, the sensor read 260 N at 111N applied load and 1,040-1,070
at the 1,334 load. They stated it was “important to note that the sensor must be calibrated at the
maximum load measured, and will only be accurate for a small range of loads. The authors
considered the hysteresis “unsatisfactory” with noted maximum differences, expressed as a
percentage of load range, of 24%, 22% and 21% for the walking, medium and knee simulator
loading rates. The authors concluded that sensor accuracy was poor, but comparable for all rated.
The sensor should be calibrated at the maximum load measured, but sensor readings low and high
relative to the calibration load will not be accurate.” Although not discussed in the paper, review
of their sensor response curves clearly shows a non-linear response.
55

Werner et al. (1995) examined day-to-day repeatability and the effects of loading rate,
creep, and temperature effects on FSR-T output. Calibration was noted to have been done 1
minute after a static load was applied. Day-to-day repeatability was assessed by applying 258 N
to six sensors once per day over 6 days. Creep was assessed by applying a 258 N load for 2hours.
Frequency response was performed by applying ramp loads of 3.1, 1.5, 0.78 and 0.08 MPa/sec.
Temperature effects were examined by taking readings at various temperatures under static load.
Day-to-day repeatability was 1.6% of applied force. The sensor exhibited a “substantial” creep.
The initial sensor output was lower than the applied force; it increased by 10.4% at 1 minute
(calibration time) and had increased by 30.8% after one hour. Creep followed a Force =
10.8*ln[time]+269.3N (r
2
=0.94). For ramp loading, the calculated force underestimated the
applied force by an average of 22% for all loading rates. Much of the deviation was attributed to
the fact that calibration values were taken after 1 minute of loading and data was sampled
quicker. Applying their creep equation resulted in a correct force value during dynamic testing of
a 7.7% underestimation. A 6º C decrease in temperature caused an average decrease in output of
19.4%, also reported as a 3% variation per ºC. The authors noted that their study suggested there
are limitations to the FSR-T technology in dynamic loading situations.
McPoil et al. (1995) compared the FSR-T (software version 3.622) to the EMED
capacitive based system from Novel (referred to as an FSC-N “force sensitive capacitor-Novel”,
discussed in capacitive sensor section). Bench testing involved applying pressures from 0-50
kPa, 0-100 kPa, 0-150 kPa 0-200 kPa, 0-300 kPa, 0-400 kPa, and 0-500 kPa where “each pressure
load was applied as quickly as possible, starting at zero pressure, maintained at the prescribed
level for a minimum of 2 seconds, and then rapidly released.” Unfortunately, no quantification of
the how long “as quickly as possible” was given for the pressure rise. Given this uncertainty, the
data is questionably applicable to dynamic events. Long term drift testing was conducted under a
56

steady load of 150 kPa out to 10 minutes. “Dynamic testing” was conducted, comparing the two
systems consistency and reported by the intraclass correlation coefficients. This is notably not
dynamic testing of the accuracy of either system, only consistency. The bench testing, as seen in
Figure 34, showed that FSR-T had an average error of 4% at 50 kPa and a continually increasing
error with pressure up to a 24% error at 500 kPa. The total creep for the FSR-T was 11.6% at 10
minutes, but the change over time was non-linear.

Figure 34: Absolute error with increasing pressure (McPoil et al., 1995)

Repeatability of the FSR-T was examined between trials (peak 1 and peak 2 among three
steps), between sessions on different days, between time in session 1, and between time for
session 2 (2
nd
day). The results are shown in Table 9.

57

Table 9: Intraclass correlation coefficients for dynamic testing (McPoil et al., 1995)

The authors noted that “the reliability and validity problems reported in this study with
the F-Scan insole system raise serious questions regarding the ability of the F-Scan insole system
to accurately and consistently measure normal forces and, as a result, plantar pressures.”
Woodburn and Helliwell (1996) noted nearly identical results as Ferguson-Pell and Cardi
(1992, 1993) and Pavlovic et al.(1993) with the F-Scan (a dedicated in shoe system from
Tekscan) FSR-T system having hysteresis of 21% and creep of 19%. Sensor variability was
“acceptably low” between approximately 1-5%, with several instance of more than 10%
variability at very low force levels. Repeatability tests simulating a normal gate cycle but at 2.48
kN of applied load found that the sensors developed a 10% error within 9 cycles, and had 18%
and 25% errors at 100 and 500 cycles, respectively. They noted that output was more stable over
time when higher loads were applied. They did note that the F-Scan system had several good
features including a thin unobtrusive sensor, good spatial resolution, adequate sample rates and
user friendly software. However, after testing the F-Scan they concluded that “despite some
encouraging first impressions, the F-scan is not entirely suitable for accurate and repeatable in-
shoe pressure measurements.”
58

Pitei et al. (1996) examined the longer term time-dependent behavior. FSR-T’s were
tested with an Instron material testing machine running a ramp displacement at either 20 mm/min
(0.1 Hz) or 50 mm/min or (0.25 Hz) to a pre-set load. Under load control to a fixed load value,
they observed an upward drift in apparent peak pressure in each cycle. There was a mean
increase of 14.7+/-2.1% from the initial peak to the final peak at 20 mm/min and 12.8%+/-2.9%
at 50 mm/min. As seen in Figure 35 the sensor obtained an asymptotic, but positively sloped
value after 74+/-6 seconds at the low frequency load rate. A similar asymptotic value was
obtained after 58-59 seconds at the higher frequency load rate.

Figure 35: Maximum pressure variation under cyclic load to constant peak force (Pitei et
al., 1996)

Even after a rest of 5 minutes between trials, Pitei et al. (1996) found an upward drift in
the sensitivity of the same sensor with repetitive trials as shown in Figure 36. The authors noted
the response was only recoverable after a rest period “much longer” than their allocated 5 minute
rest between trials.

59

Figure 36: Change in readings during cycle and between cycles (Pitei et al., 1996)

Martinelli et al. (1996) examined the force detection error (accuracy), contact area error,
repeatability, homogeneity, creep, one axis bending and two axis bending of the FSR-T and the
FSC-N. Only the results of the FSR-T I-Scan 5051 sensor will be discussed here. After
conditioning the sensor, a silicone sheet was applied at the interface” to “distribute the load
uniformly on the entire area.” No additional information was provided regarding the “silicone
sheet.” The two step power law calibration provided with the Tekscan software was used. The
sensor was calibrated only on a flat surface to “investigate the potential error in force detection
when the sensor was loaded in a bending situation. Calibration data was “recorded after 5
seconds from the beginning of the holding time, which lasted for 30 seconds, to equilibrate the
sensels.” Calibration forces of 500N and 1000N were used. Three geometric configurations were
used flat (circular), spherical, and cylindrical. “Rigid steel indenters were used because the aim
was to apply forces and not pressures.” No further data was presented on the flat indenters.
Spherical surfaces had radii of 30 mm and 50 mm into a “fully conforming” mating spherical
60

surface. Cylindrical loading utilized a (upper) male surface with a 40 mm radius and female
surface (lower) with a 50 mm radius. “Loads were increased during a 5 second period, held for
10 seconds, and released over 5 seconds. Data were acquired after 5 seconds of the static
condition” (10 second after the start of load application). Each dynamic trial consisted of 10
cycles. The authors noted that creep is an important parameter negatively influencing the results
of a test because it limits the cycle frequency. Sinusoidal loads applied with a period less than the
time needed for sensor recovery will produce a creep artifact. When loaded by the flat ends of
two different sized cylindrical indenters, the FSR-T had force measurement errors of -12% to
+20%. Area error was 2%. Repeatability of the FSR-T was not even reported because “the
ISCAN did not perform accurately in the 10-cylces loading test.” Errors during loading on a
cylindrical surface were 9% with no difference between directions of curvature. The FSR-T
showed crinkle artifact during pretesting with spherical surfaces. Therefore, the spherical surfaces
“were not used because the radii of the spherical contact surfaces were too small to allow loading
without large crimps and potential sensor damage.” Creep artifact (Cr) was 18% with memory
artifact recorded after the unloading time of 15 seconds. No artifact remained after 30 second of
relaxation.
Buis and Convery (1997) dynamically tested the F-Scan FSR-T in an Instron material
testing machine and found similar results as Pitei et al. (1996). They noted inter-cell variations
on the order of +/- 50% from the mean value, but the equilibrium software from the manufacturer
“eliminates this variation.” However, they noted that “the output of a typical cell demonstrated
both a lack of repeatability and a difference between loading and unloading.” Cyclic tests did
show a good dynamic response between applied load and the FSR-T output, but it took
approximately 10 cycles of load to obtain stable results. They were one of the first researchers to
examine the effects of interface curvature on FSR-T response. “Flat, cylindrical and spherical
61

matching contact surfaces were studied” under constant compressive load. Unfortunately, no
additional description, such as materials used or radius of curvature, of the contacting surfaces
was provided. However, as seen in Figure 37, for nearly all loads, there was an approximately
300% decrease in the sensor output when a spherical indentor was used compared to the flat
indentor under the same applied force. They concluded that “it must be recognized that the actual
pressure levels recorded are not absolute. Sensitivity of loading rates and hysteresis are two
problems which still exist.”

Figure 37: Effects of indentor shape on F-Scan FSR-T response (Buis & Convery, 1997)

Woodburn and Helliwell (1997), in an extended abstract with the exact same title as his
prior paper (Woodburn and Helliwell, 1996), used a small jig-mounted force meter to test groups
of individual sensing units for within and between sensor accuracy in full size and adjusted (cut)
sensors. “An Instron servohydraulic materials testing unit was used to evaluate creep, hysteresis
62

and the repeatability of output forces over repeated loading cycles for full size and cut sensors.”
This was the extent of their method description. They found there was average calibration error
of 4%, creep of19%, hysteresis of 21%, within sensor error of between 1 to 23%, and between
sensor error of between 1 and 26% (highest within and between sensor errors were noted for the
low range of tested forces). Measured errors in the force reached a 10% error rate at between 9
and 397 cycles, errors were between 8 and 18% at 100 cycles and between 13 and 33% at 1000
cycles. The authors concluded that “The F-Scan calibration protocol is inaccurate. Significant
within and between sensor variability in output forces can be expected. The creep and hysteresis
properties of the sensor are poor. The F-Scan sensor does not yield repeatable measurements.
When the sensor is adjusted to a smaller size, output is adversely affected.”
Luo et al., (1998) examined the accuracy of the F-Scan FSR-T by bypassing the
manufacture’s software and directly measuring the individual sensel output with custom DAQ
software. Testing with an MTS material test system showed the sensors were “highly dependent
on the contact surface hardness” with 3 times greater output and 6.4 times greater standard
deviation on hard surfaces versus soft surface at the same test load and contact area. Sumiya et
al. (1998) found similar effects of contact surface hardness. For dynamic load rates of 0.0241
MPa/s and 0.241 MPa/s there were statistically significant difference in the calibration
coefficients between the two load rates and compared to static loading. Luo et al. as shown in
Figure 38, found a rather pronounced temperature effect, with exponential type variations in
sensors output around standard body temperature of 37ºC.

63

Figure 38: Temperature effects on F-Scan output (Luo et al., 1998)

Luo et al. concluded that “the sensor is sensitive to surface conditions, loading speeds,
and temperature. Variations also exist from sensor to sensor. In order to have accurate
measurements, calibration was recommended in actual clinical or experimental conditions prior to
use, including surface contact conditions, loading speeds, and temperature environment.”
Sumiya et al., (1998) in a separate, but sequential article in the same journal issue as Luo
et al., (1998), also examined the accuracy the F-Scan FSR-T. They examined the response using
F-Scan software version 3.42 and a new version 3.622 that utilized a special calibration they
referred to as the Baumann compensation. Via dynamic testing that consisted of dropping a
weight 3 mm onto the sensor, as seen in Figure 39, they showed that the sensor had a mean
dynamic response time (time to 90% of stable state) of 0.32 s with a range of between 0.24 and
0.41 seconds. This was noted to be substantially slower than the manufacturer’s quoted value of
5 µs. It should be noted that the authors erroneously refer to this time as the dynamic response
time when it is actually the settling time under constant load preceded by an impact. As indicated
64

by the characteristic dynamic impact spike early in Figure 39, the impact is over after
approximately 0.050 seconds. The load response after that is the settling time under the constant
static load of the previously dropped weight. 100% output is achieved at approximately 1.3
seconds which would imply a frequency response of approximately 0.77 Hz (1/1.3 seconds). As
the mean vertical ground reaction frequency in gait is at approximately 12.21 Hz (Stergiou,
2002), the FSR-T appears to react nearly 16.6 times too slow to capture all the power of a foot
impact during walking.

Figure 39: FSR-T dynamic response time (Sumiya et al., 1998)

Sumiya et al. (1998) also compared the F-Scan response to the ground reaction forces
(GRF) measured in the more traditional and mature technology of a foot to force plate loading
during gait. As seen in Figure 40, they found that relative to the load cell, the F-Scan required
from 11% to 17% longer time to reach the first peak force, but gave almost equivalent force
values of between 93% and 101% of the load cell. At the second peak the time was similar;
between 96% and 103%, but the force were 14% to 17% lower.
65

Figure 40: Phase and magnitude lags of F-Scan (FSR-T) (Sumiya et al., 1998)
Sumiya et al. (1998) concluded that “rather than using F-Scan measurement to accurately
obtain actual values, it should be used for relative comparisons…” Only slight differences were
noted with the Baumann compensation and it was noticed to be “inconsistent and theoretically
unacceptable.”
Ahroni et al. (1998) set out to examine the “reliability of the F-Scan” system. What they
essentially did was examine the repeatability and precision, but not the accuracy. Although they
had low coefficients of variation (0.15 and 0.155) and relatively high ICC values (0.751 and
0.755), they only showed there was low variance and good repeatability. Reliability when
assessed with ICC generally is looking at how well two different instruments agree, typically with
one instrument considered the reference or gold standard. They used the same instrument and
only determined how different the measurements were using that one instrument. This provides
no information on the accuracy, therefore only partial data on reliability as there is no attempt to
quantify the accuracy. Their statement that the system “demonstrated excellent to good
reliability” cannot be proven by their method.
66

Otto et al. (1998, 1999) stated they were going to quantify the output drift and
characterize the dynamic response of the K-Scan FSR-T. Testing was conducted within a
pressure vessel with hydraulic fluid pressing on a Nitrile membrane over the sensor on an
aluminum base. The use of the hydrostatic pressure vessel permitted “unambiguous applications
of homogeneous contact stress.” Calibration was conducted per the manufactures software at
20% and 80% of anticipated load. As seen in Figure 41, during continuous static loading the
sensor initially underestimated the true value but quickly, typically in less than 30 seconds,
crossed over to overestimation. For the long-term continuous loading cases the sensor reached a
steady state overestimate after about 10-20 min. Drift created progressively accumulating
relative error up to +58 percent (overestimation) at the end of the 3-h period for the 15 MPa case.
Like drift, intersensel variability increased with time. For repeated intermittent loading (load on
for 2 seconds, off for 5 minutes, and repeated for 3 hours), the sensor also initially underestimated
the pressure, but the sensor reading decreased after the initial recording, asymptoting to a steady
state underestimate after about 10 min with an approximately -22% error (underestimation).
During the short-term loading static experiments (this appears to be identical to the early part of
the long turn static loading, but with different results) the sensor initially underestimation the
pressure, became equal at 25-35 s, and after that point relative error decreased monotonically with
an increase in load. The rate of drift increase was greatest immediately after load application and
the absolute rate of drift was higher at higher loads. Although not discussed by the authors, as
demonstrated in Figure 41, some other source of error had been introduced as evidenced by the
fact that in Figure 41(a) the highest error is at the highest pressure, but in Figure 41(b), which is
essentially the same testing but stopped at 15 minutes, the pattern has reversed with the highest
error at the lowest pressure and lowest error at the highest pressure.

67

Figure 41: Sensor drift response (a) long term continuous and intermittent loading, (b)
short term continuous loading (Otto et al., 1999).

Frequency response to a sinusoidal input was estimated by assuming the system acted as
a first-order mechanical system. The time constant ( τ) was calculated utilizing the time between
corresponding events ( β
1
) and frequency ( ω). Magnitude ratios were calculated from the peak-to-
peak contact stress “excursion.” The authors noted that “for an ideal first-order system, an
increase in frequency will result in an increase in the time constant and a decrease in the
magnitude ratio.” However, as shown in Figure 42 and Figure 43, they got nearly the opposite, as
the time constant decreased with increasing frequency and the magnitude ratios remained flat or
increased slightly at the highest frequency. Also not discussed by the authors, the time constants
for the low (valley) values of the sensor response were different than the high (peak) values.
Given that the output of a first order system with a sinusoidal input is a sinusoid of the same
frequency but different magnitude and possibly phase shifted, the peaks and valleys cannot have
different time constants for this to be a first order system. Additionally, these graphs imply an
infinite frequency response as the sensors are reported to respond faster, lower time constant, as
frequencies increase. The authors' conclusions that “the sensors appeared able to accurately
measure dynamic contact stresses up to at least 20 Hz” is not only contrary to prior research
(a) (b)
68

(Ferguson-Pell & Cardi, 1992, 1993, Pavlovic, 1993, Werner et al., 1995, Woodburn & Helliwell
, 1996, Pitei et al., 1996, Luo et al., 1998, Sumiya et al., 1998) but is also contradicted by their
own data which shows with static loading (a step response), the sensor takes 25 to 35 seconds to
get to the input value. Assuming τ is to the 63.2% level (T= τ, 1-1/e = 63.2%), the quickest τ
would be is approximately 15.8 seconds (63.2% of 25 seconds-assuming a linear rise). The
authors attributed this apparent contradictory time constant response relative to their first order
assumption to drift; however, their description is not supported by their results. As seen in Figure
43, if this had been a drift issue, as the frequency increased the magnitude ratio should have
progressively increased but it does not. Additionally, if this had been a drift issue, then a first-
order model should not have been used. Oddly, no graphs of pressure versus time were provided
showing the dynamic testing.

Figure 42: Erroneous time constant results (Otto et al., 1999)

69

Figure 43: Magnitude ratios (Otto et al., 1999)

Urry (1999), in a very extensive and exhaustive review of the literature on plantar
pressure measurement sensors, noted a particular scarcity in the literature of dynamic
performance data. He concluded that “the dynamic behavior of these sensors, have been less
prominently portrayed in the literature.”
Valdevit et al. (1999) tested the single sensor Uniforce FSR-T. Their sensor was
preconditioned by subjecting it to fifty load cycles between 0 and 400 N at 0.1 Hz. Static
calibration was done with a rubber backed mounting platform and was then brought into contact
through alignment pins with a platen. A static load of 100 N was applied and load was recorded
at 10, 60, 300 and 1800 s and the sensor was dry, wet, kinked or folded. A folded sensor
involved folding the sensor area in half in the perpendicular direction to the long axis. Kinked
was where the folded sensor was unfolded back flat and a residual kink was present at the bend
70

line. Figure 44 shows their results with sensor exhibiting a clear continuously increasing output
under constant load.

Figure 44: FSR-T sensitivity change (Valdevit et al., 1999)

Dynamic testing was conducted in a MTS material testing machine running at 50 N/s to a
peak load of 400 N. Valdevit et al. (1999) was also one of the first to investigate the response of
the FSR-T to curved surfaces. They placed the sensor on a 32 mm radius surface and loaded it
with a surface of 26 mm radius. They do not describe the material characteristics of the curved
surfaces. Their results for various surfaces under dynamic testing are shown in Figure 45.
Notably there was over a 400% increase from the “hard” to “dry” (rubber backed) condition and
an approximately 16% decrease in sensitivity for a curved versus dry surface. They do not
quantify nor describe how “hard” and “soft” surfaces were established. They did note that the
introduction of the sensors in the facet joints could elucidate effects of fusion on the load transfer
to the facets. They also stressed the need for appropriate calibration under conditions which
closely resemble those it will be used in.
71

Figure 45: Effects of surface conditions on FSR-T sensitivity (Valdevit et al., 1999)

Harris et al. (1999) set out to compare the pressure reading from Fujifilm and a K-Scan
sensor (a version of FSR-T from Tekscan). However, this study has a failing similar to the other
studies in that the calibration was conducted on flat surfaces and the joint tested (Wright Medical
Technology, Size 3, Advance Semi-congruent total knee arthroplasty) was curved with an
approximate 31 mm or smaller radius (Wright Medical Technology product literature,
www.wmt.com. 2010). They did appear to match interface material properties from calibration to
testing. The FSR-T was calibrated according to the manufacturer's guidelines, one side of the
sensor was placed on a “polished steel plate” and the other side was covered by a 25 mm thick
polyethylene “block” designed to cover as much of the array as possible. The difference in
calibration conditions to utilization conditions would call into question their ultimate numerical
results. During their flat plate calibrations, they did find nearly a 10% progressive non-linear
increase in measured contact area with increasing load, despite a constant indentor area as seen in
Figure 46. An interesting omission in this study is that the authors constructed a curved alumina
ceramic surface of 11.19 mm diameter to check area measurements, tested the FSR-T in that
curved surface in a material testing machine, and reported the area differences but they did not
report any of the FSR-T measured forces at that radius.
72

Figure 46: Measured area change with constant area loading (Harris et al., 1999)

Chen and Gates (2000) conducted a similar GRF comparison of the F-Scan FSR-T to
forceplate data as Sumiya et al. (1998), but with a larger sample of 30 volunteers each with 10
steps of force plate and F-scan data (5 right foot and 5 left foot). Notably they used F-Scan
software version 3.611 that has the Baumann compensation (Sumiya et al., 1998). They found no
statistically significant differences in the force values, but the F-Scan system showed a
statistically significant delay (P<0.05) in the standard bi-modal (Figure 40) time to the first peak,
time to minimum after first peak, and time to second peak.
Ferguson-Pell et al. (2000) was one of the first to explicitly indicate a sensor should be
“free from error of measurement on curved surfaces..” although this had been shown to have a
significant effect on FSR-T sensitivity by Buis & Convery(1997) and Valdevit et al. (1999).

73

They also listed the following specifications for a sensor to measure interface pressure between
support surfaces, pressure garments and the skin:
(a) small and thin, e.g. 1 mm in thickness, 10 mm in diameter, and highly flexible;
(b) have a continuous output;
(c) be able to measure shear as well as normal forces;
(d) be free from error of measurement on curved surfaces and from the effects of
temperature and moisture;
(e) low cost.
Ferguson-Pell (2000) tested the single sensel FlexiForce sensor. A major limiting factor
associated with their testing that makes it nearly unusable for many real world applications was
that “readings were taken after achieving a stable level mostly 1 minute after loading.” In light of
the enormous amount of previous research (Woodburn & Helliwell , 1996, Pitei et al., 1996, Buis
& Convery, 1997, Luo et al., 1998, Sumiya et al. , 1998, Valdevit et al., 1999) including some of
the author’s own (Ferguson-Pell & Cardi, 1992, 1993) showing significant creep and a highly
non-linear and variable response during the initial 0-60 seconds of loading, this type of delay in
measurement eliminate one of the main benefits of the FSR-T system, which is recording of a
continuous time response. This method is essentially taking substantially time delayed
“snapshots” at a single instance of time. For testing of drift, this type of delay would be
acceptable. But for repeatability testing, linearity testing, hysteresis testing, and curvature testing,
this data would only apply to testing where a load was applied and held constant for “mostly 1
minute after loading.” This would not be relevant or applicable to most types of dynamic
biomechanical testing, including gate and most dynamic orthopedic tests such as within the knee
joint during walking or spine flexibility tests. The study was also limited, as stated by the
74

authors, by the fact that they only tested the ultrasensitive sensor that has an operating range of 0-
0.45 kg.
Despite the substantial limitations of the study, the authors did quantify in a more
systematic and detailed way than prior studies, the effect of curvature on the sensor output.
Curvatures of radii of 8.0, 10.4, 13.6, 16.1, 32.5 and 51.7 mm were used with “readings were
taken 1 min after loading” at pressures from 0 to 50 mmHg (0-6.67 kPa). As seen in
Figure 47, using the 20 mmHg applied pressure point, there is an approximately 700% increase in
the sensor output at a radius of 8.0 mm compared to a flat reading at the same pressure.

Figure 47: Curvature effects on FSR-T output (Ferguson-Pell et al., 2000)

As seen on the left side of Figure 48, curving the FSR-T also causes an artifactual, non-
zero, offset in the sensor reading while under 0 mmHg of pressure. Not only is there a shift in the
zero value with curvature, but as shown in the right side of Figure 48, the sensitivity can change
approximately 83% from a radius of 10.4 mm to a radius of 13.6 mm and approximately 100%
from a radius of 10.4 mm to a radius of 32.5 mm.
75

Figure 48: Offset and sensitivity changes in FSR-T with curvature (Ferguson-Pell et al.,
2000)

Ferguson-Pell’s data in Figure 47and Figure 48 would raise valid questions on the
accuracy of data reported in studies in which the sensors are calibrated on a flat surface, say in a
material testing machine, but data is reported from within a curved joint such as the elbow, knee,
ankle or spinal facets.
Wilson et al.(2000) stated in the very first line of his paper that “The I-scan system can be
used to measure continuously changing force and pressure distributions at biomechanical
interfaces” (I-scan is the “Industrial” version of the FSR-T). However, similar to Ferguson-Pell
et al., (2000), “in all tests, a 5 s long ramp function increased the load to the desired value, where
it was held for 5 seconds, at which point the force measurement was recorded.” They are also
taking a “snapshot” discrete reading of the data after 10 s of load applications. This is not a
measurement of the “continuously changing force and pressure distribution.” The authors do note
“it is important to note that the tests performed have been designed to minimize the effects of
drift and hysteresis that the I-scan system displays, and that the accuracy with which the system
measures a changing dynamic load may be substantially poorer than the static accuracy
76

measured.” Given these limitations, compared to the material testing machine, the I-scan had a
mean error in force readings of 6.5% +/-4.4% (mean+/- S.D.), but this force accuracy was
assessed through a flat-ended cylindrical indentor with the sensor backed by 1/8 inch thick
“rubber.” Force distribution error was 0.86%+/- 0.58% using two “flat-ended cylindrical
indentors.”
Polliack et al. (2000) examined the F-Socket Tekscan FSR-T and the Rincoe SFS (FSR-
R) which are both FSR based sensor systems designed for prosthetic fit assessment, but only the
FSR-T will be discussed. Accuracy, hysteresis, drift, and curvature effects were evaluated.
Accuracy was assessed on “flatbed” and within a “mould” at pressures from 69 kPa to 531 kPa
after having been calibrated per manufacturer specifications at 172 kPa. Flatbed testing utilized
the Tekscan flat bladder calibration system and was described as “ideal conditions.” Mould
testing was conducted by placing the sensors at various locations on the positive mould of a
transtibial stump that was inserted in a hydrostatic pressure vessel. Hysteresis testing was
conducted from 0 to 345 kPa at the same load up and down rates. Drift test measurements were
taken at 5 minute intervals for 20 minutes at constant pressures of between 69 kPa and 414 kPa;
however, the FSR-T was re-calibrated at each hold pressure before each test. Curvature effects
were examined over radius from 17.1 mm to 99.9 mm (note: the units on the radius were not
reported, but they were determined by a review of similar type test setup by the same authors
(Polliack et al., 2002). The FSR-T was reported to have an 8% error in flatbed and 11% error in
the molded configuration. Hysteresis errors were 42% in flatbed and 24% in mould. Drift errors
were 12% flatbed and 33% in mold. No correlation (no consistent trend with increasing or
decreasing radius) between degree of curvature and accuracy errors was observed, but there was
between 3.8%+/- 5.2% and 17.7+/17% (at the smallest radius) error across all the curvatures.
77

Morin (2000) investigated the effects of contact compliance and changes in system
hardware. The Tekscan 9811 FSR-T (1-500 kPa) sensor was used with two calibrators in a total
of four calibration conditions. The first method used the standard Tekscan bladder configuration
composed of a lower metal plate covered in latex rubber, the Tekscan array, an inflatable rubber
bladder, and a metal plate. The last three calibration methods used a hand-held calibrator that
loaded four sensels at a time with the array on a flat non compliant surface, and approximately 3
cm lateral of the spine on the back of a LC Simulator mannequin and of the spine of a human
subject. Linear calibration coefficients were obtained with the four methods. The slope of the
calibration curve (sensitivity) compared to the standard Tekscan bladder calibration varied on
average by 36%, 19% and 7.8% for flat hand-held, manikin hand-held and human-hand-held,
respectively. Variability in the slope also increased with increasing surface compliance. Changes
in the hardware cuff (plug from DAQ to the sensor array) produced average differences from 5%
to 32%. The authors concluded that “the Tekscan sensor system must be carefully calibrated to
obtain the best results in terms of accuracy and repeatability. Calibration should be performed
under conditions that are as close as possible to the actual measurement condition.” They also
noted that calibration of the Tekscan system remains a challenge and a good understanding of the
factors which affect Tekscan sensor performance and contribute to variability in the output are
essential.
Stuhmiller and Sih (2001) started by acknowledging the problems of the F-Scan but then
attempted to improve the accuracy by using different calibration techniques. Notably they were
using software version 4.46. The authors followed the manufacturer's calibration requirements,
but then also performed calibration after 1 second, 10 second and 60 seconds of standing. The
values were then compared with load cell data of four trials of slow walking. A third method
utilized the 10 second calibration but considered four trials of slow walking, fast walking, and
78

running to examine the effects of different speeds on the calibration. In all trials the F-Scan
underestimated the total load compared to the force platform, sometimes by as much as 44%. As
would be expected given the dynamic aspects of foot strikes during gait, the 1 second load
calibration always performed best with mean errors of between -11 and -27% compared to mean
errors of -36 to -43% using the 60 second load calibration. The authors modeled the response of
the F-Scan as a linear spring, similar to the manufacturer's linear model, but optimized their
coefficients to minimize the root-mean-square error. A second model referred to as a standard
linear solid (SLS) model was a spring and dashpot in parallel connected to a second spring in
series. The linear spring model resulted in revised F-Scan errors between -5% to +8%. The SLS
model resulted in error of between -4% and +14%. Across a wide range of walking speeds, the
SLS model performed best with a mean error of less than 11% compared to the manufacturer's
calibration error of 31%.
Hsiao et al. (2002) using a systematic stochastic method examined the effects of applied
pressure, duration of pressure application, calibration pressure used and measurement week
(measurements taken in different weeks) on the accuracy and precision of the F-Scan FSR-T and
the FSC-N using a rubber bladder and a “smooth” wood-wood interface. The authors noted “in
each calibration, the insole was loaded for 5 minutes before the pressure was recorded. The 5-
min calibration was intended to reduce the amount of time-dependent change of the insoles.”
However, during testing, data was collected after 2 s, 5 min, 10 min and 15 min at each pressure.
A 4 (applied pressure) x 4 (duration) x 2 (calibration pressure) x 3 (week) design was used for
ANOVA statistical analysis. Three main effects (calibration pressure, applied pressure, and
duration) seven two-way interactions and three three-way interactions were statistically
significant. Lower error where reported -3.8% to 4.5% when applied pressure was commensurate
with the calibration pressure, when calibration pressure was incommensurate with applied
79

pressure, the errors ranges from -26.3% to 33.9%. The authors noted this may explain why
McPoil et al. (1995), who calibrated at approximately 66 kPa, but tested from 50-500 kPa, and
Luo et al. (1998) who tested from 48 to 241 kPa and calibrated at 241 kPa, obtained different
results.
Wilson et al. (2003) noted that the results of many earlier studies could not be extended
to the natural joint because “the system’s performance depends on the compliance of the
materials between which the sensor is compressed, the geometry of the interface, and the sensor
type and shape.” Despite the title of the paper “Accuracy and repeatability of a pressure
measurement system in the patellofemoral joint”, as noted in the abstract, they only tested the
“repeatability of the system’s measurements of patellofemoral contact forces…” not the accuracy
in the joint. Even after acknowledging the geometry of the interface was critical to system
performance, the calibration and accuracy were established by compressing the sensor with a
“machined flattened cylindrical indentor” that was covered with 3.2 mm thick rubber resting on
an aluminum plate covered in lubricated 3.2 mm thick rubber. The authors did note in discussion
that “Measurements of force and force distribution were much more repeatable in our flat
simulation of the patellofemoral joint than in a cadaver specimen. This may be because the sensor
does not respond as consistently when it is loaded in the curved, nonconforming joint between
two materials with variable compliance than in a flat, conforming interface with a uniform
compliance.” A stated limitation of their study was they assessed the accuracy at a flat interface
and “no human or animal joint is perfectly flat…”
Wilson et al. (2003) utilized the FSR-T to quantify the forces in the patellofemoral joint.
As had been done in his prior studies (Wilson et al., 2000), extolled the virtues of the FSR-T
being able to take “continuous measurements of force,” yet “in all trials, we increased the load
steadily to the desired level over 5 s, held it for 5 s, then recorded the force measurement.” This
80

discrete snapshot 10 seconds after the load has been applied is certainly not “continuous.” They
did, however; remain consistent and repeat this same method while testing within the
patellofemoral joint. For calibration, each sensor was placed between machined aluminum plates
coated with 3.2 mm thick lubricated rubber. As seen in Figure 49, the patellofemoral joint is a
highly curved joint. Kwak et al. (1997) has shown the patella is concave in the lateral regions,
convex in the central region and concave in the medial region. Curvatures to as small as 250 m
-1

(radius of 4 mm) on the patella and as small as 175 m
-1
(radius of 5.7 mm) on the mating femoral
surface were reported. Given the highly curved interface, the accuracy of data from a sensor
calibrated on a flat surface and then used in this curved joint are questionable.

Figure 49: MRI of patellofemoral joints (A) 0º, (B) 45º, (C) 90º, and (D) 135º knee flexion
(Moro-oka et al., 2002)

Brimacombe et al. (2005) examined the effects of different calibration methods on the
accuracy of the FSR-T. “To simulate the loading in a prosthetic knee joint, each sensor was
compressed between a flat disc of UHMWPE (38.2 mm in diameter) and a large aluminum
plate.” Similar to prior studies, snapshots of data were taken after the force had been “ramped up
over 10 s, held for 5 s, and decreased for 10 s.” Sensors remained unloaded for 120 seconds
between load applications. As seen in Figure 50 the femoral condyles of knee joint have
A B
C D
81

relatively highly curved surface with major curvatures in the A-P (anterior-posterior or sagittal
view) of radii between 19.4 and 19.9 mm (Howell, 2010) and lateral curvatures of radii between
17.8 and 22.1 mm (Nuno, 2003) . The mating surface of the tibial plateau has also been shown to
be curved with radii between 18 and 25 mm (Dam, 2007). Given the generally curved interface
of the knee joint, the accuracy of data from a sensor calibrated on a flat surface are questionable.
The authors did find that a 3-point quadratic calibration lowered the error to 1.5%. They also
“recommended that investigators design their own calibration curves. Since output is dependent
on experimental protocol (sensor type, interface materials, sensor range in use, etc.), sensor
behavior must be investigated for each application.”

Figure 50: Curvatures of the knee joint. Lateral x-ray(left) and anterior view (right)
(Howell et al., 2010)

Hoffmann and Decker (2005) examined inaccuracies in measurement of the FSR-T due
to the grid spacing in the array. A main disadvantage of the system was noted to be high rigidity
in the vicinity of the sensor crossing. They noted that because the sensing array is a combination
of “sensing area” and “inactive area” the sensor must be calibrated with a material having
stiffness similar to that of the material to be tested.
82

Rousseau et al. (2006a) was one of the first to use the single sensel FlexiForce (A101-
500) FSR-T’s in the spine facet joints. However, the only calibration and conditioning
information provided was “our sensor calibration demonstrated that the output voltage varied
linearly with the force regardless of the pressure area.” The highly curved nature of this joint
(Tulsi et al, 1993, Panjabi, 1993), may make the author’s statement about the sensors true for a
flat surface, but potentially highly inaccurate for curved surfaces. The facet forces reported in
this study are at best not supported by the method and are potentially substantially in error due to
creep and hysteresis (Ferguson-Pell & Cardi 1992, 1993, Woodburn & Helliwell, 1996, Pitei et
al., 1996, Buis & Convery, 1997, Luo et al., 1998, Hsiao et al., 2002), potential curvature effects
(Buis & Convery 1997, Valdevit et al., 1999, Ferguson-Pell et al., 2000), calibration versus
utilization interfaces and pressures (McPoil et al., 1995, Luo et al., 1998, Sumiya et al. , 1998,
Valdevit et al., 1999), and temperature effects (Luo et al., 1998). Facet loads at L5-S1 were
reported, as seen in Figure 51, when the spine was subjected to 650 N of compressive and 550 N
of shear load at various flexion-extension angles. Tests were conducted on intact spines and
spines with a Prodisc II prosthetic.

Figure 51: Facet loads intact and modified L5-S1 (Rousseau et al., 2006a)
83

Similar research by Rousseau et al. (2006b) noted the sensor output was recorded at 5 Hz
and averaged using data acquisition software (Labview 6.1, National Instruments, Austin, TX,
USA). They calibrated their sensors by applying “pre-determined forces via contact surfaces of
different areas and demonstrated that the output voltage varied linearly with the force regardless
of the pressure area. The calibration ratio was 500 N/V (±5%).” Facet loads at L5-S1 were
reported as seen in Table 10 when the spine was subjected to 650 N of compressive load and 550
N of shear load and various flexion-extension angles and lateral bending angles.
Table 10: Facet loads for flexion and extension and lateral bending (Rousseau et al., 2006b)

Wilson et al. (2006) was one of the first to use the grid-based array FSR-T (I-Scan with
Tekscan software version 5.1) to measure lumbar facet loads. They noted “it was not clear how
accurately and repeatably these sensors will assess load in the facet joints because sensor
performance is dictated by such factors as compliance of interface materials and sensor type
used.” It is interesting that the above quote is nearly identical to what the same authors noted in
their 2003 paper, except the current statement is lacking the statement from the earlier work that
Flexion/Extensi
on
Lateral Bending
84

accuracy and repeatability were also affected by “the geometry of the interface.” This current
omission is significant given the small and highly curved geometry of the facet joints.
Wilson et al. (2006) noted conditioning and calibration followed the manufacturer's
recommendation and previously established methods (Harris et al., 1999; Wilson et al., 2003 –
both reviewed above). They stated each 11x11 grid of 1.6 mm
2
sensels was conditioned and
calibrated between machined aluminum plates covered in 3 mm thick foam (nominal compressive
strength: 50 kPa at 25% deflection, 90 kPa at 50%) and lubricated with surgical lubricant.
Accuracy of the FSR-T was assessed by placing the sensor grid in the L3-4 articular facet and
loading the isolated (all other spinal materials –disc, ligaments, etc. removed) to pure
compressive loads of 25, 50 and 100 N. Within the curved lumbar facet, the I-Scan FSR-T
overestimated the applied compressive force by 50+/-9%, 35+/-7% and 18+/-9% at 25, 50, and
100 N respectively. In contrast to their prior studies using snapshots of data, FSR-T data was
collected at 5 Hz in the current study. The authors noted in discussion that “accuracy depends
critically on sensor preparation, calibration method and on the load levels assessed.” Despite
their prior comments about geometry effects (Wilson, 2003), no mention of that as a possible
source of inaccuracy is explicitly stated. They did note that “the limited accuracy of the sensors
in our tests may be explained, in part, by the fact that the applied loads were small relative to the
sensor’s measurement range.” They noted that their findings of lower accuracy at low force
levels is consistent with results from other measurement devices, but notably none of their
references used FSR-T’s. They do show reasonably good repeatability of the system. Results
using a 2-point calibration and linear calibration are present in Figure 52 for mock injured L3-L4
levels (nucleotomy, facet joint capsule sectioning, and posterior ligament sectioning).

85

Figure 52: “Injured” L3-4 fact loads measured by FSR-T under 7.5 Nm (1.3º/sec) extension
loading (Wilson et al., 2006)

Bachus et al. (2006) compared to the I-Scan FSR-T to Fujifilm. As shown in Figure 53
the FSR-T and Fujifilm were loaded between a “finely ground steel plate” (P) and the “finely
ground flat face of a cylindrical steel peg” (C). There was no description of the data acquisition
from the FSR-T. Using their test configuration, they found the FSR-T was more accurate than
Fujifilm for estimating area and pressure between flat steel plates. However, their Fujifilm was
scanned “within 2 hours after each stain was made” despite prior research (Liggins & Finlay,
1992) recommending a 50 hour development time for Fujifilm with a documented +/-5% change
in the film over that time. Given the prior research and known effects of contact interface
material properties and shape, it is unclear what relevance, if any, these results would have to
testing in a biological specimen that are typically not flat and have substantially different material
properties than steel.

86

Figure 53: FSR-T loading configuration (Bachus et al., 2006)

Martinelli et al. (2006) compared the FSR-T ISCAN 5051with the FSC-N Ankle Joint
Pressensor (AJP) (discussed in FSC-N section) and sought to answer the following questions: (1)
what force errors occur in the two devices when measuring applied forces through flat and curved
surfaces; (2) what errors in contact area measurements occur in the two sensors in various
geometric configurations; (3) are the force measurements repeatable in the same test; (4) do the
sensors show the same response to an applied load regardless of its location on the surface; (5)
how are sensors affected by creep artifact; and (6) do sensors respond with the same accuracy
when bent. A material testing machine was used to apply known with machined surfaces and
counter surfaces made of polyvinyl chloride (PVC). As seen in Figure 55 three geometric
configurations were used: flat (circular), spherical, and cylindrical. Rigid steel indenters, not
deformable at the pressure range used during testing, were used. Small (area, 0.9 cm2) and
medium (area, 3.3 cm
2
) flat indenters (A) were used with different kinds of setups to cover a
contact area ranging from 0.9 cm
2
to 6.6 cm
2
. Two pairs of spherical counter surfaces (B) were
used with radii of 30 mm and 50 mm, respectively. The cylindrical surfaces (C) had radii of 40
87

mm (upper plate) and 50 mm (bottom plate). These authors noted in their introduction that
bending (curvature) may affect the accuracy of the sensors. However their selected radii f
curvature are less than what is present in the ankle joint (the region the FSC-N Ankle Joint
Pressensor was deigned to be used in). As seen in
Figure 54, the talus portion of the ankle joint has radii of curvature between 16 and 28 mm, with
most values between 24 and 26 mm.

Figure 54: Radius of curvature on superior Talus surface (Demirci et al., 2007)

Data from the force sensor of the materials testing machine and from the two sensors
were compared. Loads were increased during a 5-second period, held for 10 seconds, and
released over 5 seconds. Data were acquired after 5 seconds of the static condition. All tests were
based on the same loading history. The setup of the test was comparable to the one described by
Wilson et al.(2003) with respect to loading and unloading phases, but they did not use the silicone
interface. Before testing, calibration and preconditioning of the FSR-T sensor were performed.
To distribute the load uniformly on the entire area, a silicone sheet was applied at the interface.
88

Subsequently, the sensor was calibrated according to the manufacturer’s protocol, during which
calibration was performed at two force steps, which allowed generation of calibration curves
based on a power law equation. The FSR-T sensor was calibrated only on flat surfaces with a
silicone sheet to uniformly distribute the load. The flat surface was used to investigate the
potential error in force detection when the sensor was loaded in a bending situation. The data
were recorded after 5 seconds from the beginning of the holding time, which lasted for 30
seconds, to equilibrate the sensels. Calibration of the FSC-N AJP sensor was conducted with the
use of a pressure bladder (Figure 64).

Figure 55: The flat (A), spherical (B) and cylindrical (C) load application fixtures
(Martinelli et al., 2006)

The maximum force error (E
force
)

was lower for the FSC-N (range, −3% to +5%) as
compared with the FSR-T system (range, −12% to +20%) over all the force levels investigated.
The FSR-T system had an area error (E
area
) of 2%, whereas the error of the FSC-N depended on
the total area applied and on the position on the sensor. The error was within 6% for areas
between 3.3 cm
2
and 6.6 cm
2
. Lower sensor accuracy was observed in area detection of smaller
areas, where E
area
of as much as 20% were observed. The repeatability was better for the FSC-N
89

system for flat and cylindrical surfaces. The FSR-T did not perform accurately in the 10-cycles
loading test. For the FSR-T sensor, different force detection errors depended on the location of
load application with the E
force
related to the position of the applied load on the sensor surface,
suggesting sensor homogeneity was low. The E
force
in the cylindrical load setup was within 8%
for the FSC-N and 9% for the FSR-T. The FSC-N had an E
force
of 8% with the 50-mm radius
sphere and 12% with the 30-mm sphere. The FSR-T showed crinkle artifact during pretesting
with spherical surfaces. Therefore, the spherical surfaces were not used with the FSR-T system
because the radii of the spherical contact surfaces were too small to allow loading without large
crimps and potential sensor damage. Creep artifact (Cr) was 16% for the FSC-N and 18% for the
FSR-T sensor. Memory artifact was recorded after the unloading time of 15 seconds for the FSR-
T system. After 30 seconds of relaxation, no artifact remained to affect subsequent
measurements.
Tochigi et al. (2006) utilized the FSR-T to examine contact forces in the tibiotalar
articulation of the ankle. They were one of the first groups to calibrate their sensors in situ. They
noted “to account for differing joint curvatures and sensor variability, sensors were calibrated in
situ for each specimen.” The calibration curve was then used on subsequent tests of that
specimen. Very little data acquisition information is provided, but their in situ calibration would
appear to overcome many of the potentially critical failings of prior studies.
Drewniak et al. (2006, 2007) attempted to examine the accuracy of circular contact area
measurements with the FSR-T. The paper abstract indicated they were attempting to determine
the accuracy of measuring the contact area of “flat-ended circular indenters of varying known
sizes.” However, what they actually did was use the “flat-ended circular indenters of varying
know sizes” and inserted a “1.85 mm thick foam rubber pad (cut to the same diameter as the
indenter) with a modulus of elasticity of approximately 1 MPa” between indenter and the sensor.
90

It appears the authors have assumed the “foam rubber pad” would not distort and get wider under
their applied loads of 1000 to 7000N. They noted highest errors for the smallest indenter at the
highest loads of 27% and the lowest error of 5% for the largest indentor at the lowest load.
Although not stated, it is presumed that since the percentage errors are all positive, the FSR-T
was over predicting the area. Also an “unexpected finding” was that the percentage error got
larger with larger applied loads. The authors attributed these findings to the FSR-T. However,
all of these findings could be accounted for by the fact that they inserted a rather compressible
(hence increasing area) material between the indenter and the sensor. With the small indentor
(smallest area) and large force, the contact pressure is the largest and necessarily the contact
strain, hoop stress and displacement of the “foam rubber” would be greatest. Conversely, at the
largest indentor (largest area) and lowest force, the contact pressure would be the lowest. The
actual area of the “foam rubber pad” under load was never reported. This obvious problem raises
doubts on the relevance of their results. They did report increased accuracy when filtering out
sensels that were reporting values greater than two standard deviations from the mean value.
They did note that the FSR-T sensors are sensitive to the temperature, the compliance of the
indentor and supporting surfaces, and the length of time that the indentor is in contact with the
sensor. Variability of these factors could lead to further inaccuracies.
Papaioannou et al. (2008) also sought to improve the accuracy of the FSR-T calibrated
with the traditional “Bladder Method” by using a “Pin Method.” The Pin Method applied a
sensel-by-sensel load by means of a square ended 9 mm by 9 mm pin attached to a material
testing machine. Two identical Tekscan 5315 mats were used. The I-Scan software was
indicated to allow linear, one point calibration (zero is the second point) or a power law or two-
point calibration. The two point power law calibration was used for the study. Loads were held
for 100s. The average raw sensel values +/- 1 standard deviation are shown in Figure 56 with the
91

Pin Method calibration pressures. Notice the distinct difference in rise times for the sensels
relative to the pin pressure from the material testing machine. They found up to a 12% difference
in the ability of two identical sensors to assess the maximum pressure that the sensor can record.
In the sensel-by-sensel loading experiments, it was observed that the sensel’s raw readings
“overshot” when compared to the applied load by up to 50% for most of the duration of the one
hour experiment. The standard deviation of the accumulated sensel recordings increased during
the second hour of dynamic cyclic loading. An “unacceptable” drift was present always in the
continuous dynamic loading experiments. At times the sensels take about 200 to 400s to return to
their base zero reading. When the rest interval was not long enough, the sensors stayed preloaded
when the pin applied the next load sequence (see response in Figure 56 at time the interval
between 600 to 800s). The sensor was more accurate during loading than unloading similar to the
findings of Pavlovic et al. (1993). They concluded the “Pin Method”, although far more
laborious, is a better method for equilibration and calibration of sensors, particularly when
saturation is prevalent.

Figure 56: Pin Method calibration relative to average raw sensel response (+/- 1 SD)
(Papaioannou et al., 2008)
92

Niosi et al. (2008), from Wilson’s (2003, 2006) lab, utilized the Tekscan 6900 FSR-T as
shown in Figure 57 to measure facet loads in the lumbar L3-4 facets under 7.5 Nm pure moment
flexibility testing. Conditioning and calibration was conducted per the previously reviewed
papers of Harris et al. (1999) and Wilson et al. (2003, 2006). New sensors were used for each
specimen to minimize the effects of sensor deterioration.

Figure 57: Tekscan 6900 FSR-T sensor [1 of 4 fingers] (Niosi et al., 2008)

As seen in Table 11, the average total peak facet loads (average of right and left for
lateral bending and axial rotation, sum of right and left for flexion and extension) was greatest in
axial rotation (56 N), followed by extension (27 N), lateral bending (13 N), and lowest in flexion
(7 N). Right axial rotation loaded the left facet joint whereas left axial rotation loaded the right
facet joint. The contact force increased with increasing extension and was minimal in flexion as
shown in Figure 58. In lateral bending the contact force pattern was less consistent among
specimens, but typically alternated between sides for left and right applied moments.
93

Table 11: Facet contact forces [Average+/-S.D.] (Niosi et al., 2008)

Figure 58: Facet load at L3-4 with 7.5 Nm flexion and extension moment (Niosi et al., 2008)

The authors noted that the accuracy of the Tekscan was considered a limitation in the
present study. However, despite the low accuracy for small forces, the repeatability of the force
measurements suggests that relative differences in loading can still be assessed. Their results
were compared with other research on facet loads as seen in Table 12.

94

Table 12: Comparison of facet loads (Niosi et al., 2008)

As previously discussed under strain gauges, Zhu et al., (2008) revisited utilizing strain
gauges to quantify spinal facet loads. Zhu, who is also from the same lab as Wilson et al.(2003,
2006), may have reverted back to this technology due to previously reported high errors
associated with the FSR-T in the spinal facets (Wilson et al., 2006). Within this current study,
they did use the FSR-T as a quantitative check of their strain gauge data. They cited Wilson et al.
(2006) as foundation and presumably methodology for calibration of the FSR-T’s.

Table 13: Contact force in the facet joint measured by the FSR-T under 7.5 Nm moments
(negative is compressive force) (Zhu et al., 2008)

95

Additionally, Zhu et al., (2008) provided a summary of the intact facet forces reported in
the literature as seen in Table 14.
Table 14: Intact facet loads reported in the literature

Park et al., (2009) used FSR-T’s to examine the effects of various shoulder rotator cuff
supraspinatus tendon repair techniques. However, the description of the “pressure sensor
preparation” is so poor that it makes it impossible to properly assess the accuracy of their results.
They note that the FSR-T sensor was “sealed between 2 layers of clear waterproof tape using a
seam roller” and that “during pilot studies, it was shown that the sealing method did not affect the
sensitivity or repeatability” of the FSR-T. This is never further quantified. Calibration was
discussed in one sentence only and involved “using a 2-point 15- to 30-N calibration before
experimentation using the Instron 4411 load cell.” Notably there is no discussion of calibration
times, interface materials, geometry, temperature, curvature, etc.
Ramruttun et al. (2009) noted the high error rates from Wilson et al.(2006) when using
FSR-T’s for lumbar facet force measurement and sought to develop a methodology to reduce
those errors by better matching the interface materials during calibration. The sensor was
calibrated in an Instron MTS while positioned between two aluminum plates covered in 1.5 mm
thick Shinetsu KE1300 T Silicone rubber (presumably flat). They noted that the facet joint
articular cartilage is reported in the literature to have a Young Modulus of 11.0 MPa and Poisson
96

Ratio of 0.4 and their closest match for interface material, the Shinetsu KE1300 T Silicone
rubber, had a Young Modulus of between 1.17 MPa at 10% compression and 5.86 MPa at 50%
compression and a Poisson Ratio of between 0.4 and 0.5. It should be noted that the Young’s
modulus they cited might not be correct as it ultimately came from either cow or pig specimens.
These authors site Williams et al. (2007) as the source; however that paper cites Sharma et al.
(1995) who ultimately cites McCutcheon (1962) as the source. McCutcheon (1962) did testing
on cows and pigs. To check the accuracy in the facet joint, the authors harvested 5 pairs of
cadaver facets and each superior and inferior facet was separately potted. The calibrated FSR-T
was placed between the disarticulate facets and loaded in the Instron at 25, 50, 100 and 150 N
with the load perpendicular to the facet surfaces. Their absolute accuracy results are presented in
Figure 59. The FSR-T overestimated the applied force for the 2pt calibration (burgundy bar) by
19.36+/- and 14.00+/-9% for the 25 N and 50 N, respectively, while underestimating the applied
force by 3.7+/-10% and 11.27+/-13% for 100N and 150N, respectively. In light of the large
amount of prior research that show FSR-T’s are very sensitive to geometry and particularly
curvature, it is somewhat surprising that these authors did not try calibrating their FSR-T’s in the
same configuration as they used to check the accuracy. Not only would this have exactly
matched the material properties, but it would have at least accounted for geometric/curvature
effects within the limits of biological variability. The lack of accounting for geometric
characteristics could be a source of their relatively high errors in accuracy.

97

Figure 59: Force measurement errors of FSR-T in lumbar face joint (Ramruttun et al.,
2009)

Ramruttun et al. (2009) tested their calibrated FSR-T’s within the facet joints at L3-4 and
L4-5 of an L2-S1 human test model under +/- 10 Nm flexion/extension, right/left lateral bending,
and clockwise/anticlockwise axial rotation with a follower preload of 300N. The results from the
L4-5 level with the linear and 2-point calibration method are shown in Figure 60.

98

Figure 60: FSR-T measured L4-5 facet forces (E-Extension, F-Flexion, RLB and LLB –
Right and left lateral bending, CAR and AAR – Clockwise and Anticlockwise axial
rotation) (Ramruttun et al., 2009)

Hartmann et al. (2009) examined the effects of interface material property characteristics
on FSR-T response using a FEM model of the senor. The model was based on detailed
assessment of the sensor geometry with a measuring microscope (Figure 32) and component level
tensile testing of the sensor’s material properties. This is one of the first papers to report on
sensor material properties. It is interesting to note that the FE model of the sensor used
component tensile modulus but modeled loadings plates using compressive modulus in a cited
range for cartilage. A Mylar material properties sheet from the manufacturer
(http://usa.dupontteijinfilms.com) indicated a compressive modulus of approximately 2.7-2.8 GPa
and a tensile modulus of approximately 4.9 to 5.1 GPa. The current study found a tensile
modulus for the FSR-T Mylar of 2.9 GPa. They noted that a sensor's results will be compromised
when there is a mismatch between the moduli used during calibration and that used during
99

testing. As seen in Figure 61, the authors reported an error due to sensor loading of
approximately 1% per 10 MPa of modulus discrepancy in the cited range of cartilage at 12 MPa.
However, at that cited range of cartilage modulus, only approximately 10% of the load is actually
on the sensing elements.

Figure 61: Percentage load on conductor intersection (Hartmann et al., 2009)

Brimacombe et al. (2009) in a paper essentially identical to Brimacombe et al. (2005),
also from Wilson’s lab, sought to determine the best calibration method for the FSR-T. Four
calibration methods were investigated, two provided with the Tekscan software and two custom
calibrations. The two Tekscan calibrations are a single parameter linear model of the form y=mx
and a two-parameter power calibration of the form y=Ax
b
. User-defined calibrations methods
were a three-point quadratic polynomial of the form y=a
2
x
2
+b
2
x+c
2
and a ten-point polynomial
calibration of the form y=a
1
x
3
+b
1
x
2
+c
1
x+d
1
with a nonlinear least-squares fit of the cubic
polynomial. Raw sensor data was saved so that each calibration method could be applied to the
same data. For all tests an interface was selected to mimic the material interface of a standard
100

prosthetic joint with an articulation of between plastic and a metallic surface. As shown in Figure
62, each sensor was compressed between a flat polyethylene disc (diameter of 38.2 mm) and a
flat metal plate. Similar to prior studies, snapshots of data were taken after the force had been
increased “linearly over 10s, held it constant for 5s, and decreased it to 0 linearly over 10s.”
Sensors remained unloaded for 120 seconds between load applications. It is extremely puzzling
why this lab which states that their methods are designed to mimic the loading conditions within
the articulation of a joint, conduct their intra-articular testing taking continuous data (Wilson,
2006, Nioshi, 2008, Zhu et al., 2008), but take calibration points only after the load has been on
for 15 seconds. This snapshotting of data negates the main benefit of this system, being able to
record continuous data.

Figure 62: FSR-T loading interface (Brimacombe et al. 2009)

The raw output was nonlinearly related to the applied loads. Of the Tekscan methods, the
power calibration was the most accurate with a 2.7% RMS error of the tested range (0.39 +/-0.14
MPa). Both user defined calibration methods were more accurate than the two Tekscan methods.
The quadratic and cubit calibrations had RMS errors of 1.2% and 0.6% of the full tested range.
The three-point calibration, which only required one more calibration point than the Tekscan
power calibration, was twice as accurate. Repeatability of the ten-point calibration was excellent
with a standard deviation of only 0.2%. A stated limitation of the study was that they did not
101

quantify other factors known to affect Tekscan accuracy. They did not test what effect curved
contact surfaces might have on the sensor’s output because these effects were not directly
relevant to their stated goal of assessing differences between calibration algorithms. The
accuracy of the system may be lower than reported when used in dynamic loading situations due
to the effects of shear loads, loading between curved surfaces, or small contact areas. Readings
are also subject to drift and sensor output and output variability depend on the stiffness of the
contact materials and only one combination of materials was tested. Additionally, Tekscan
sensors can experience a loss of sensitivity over time. It can be concluded that the calibration
method used is one of many variables that must be planned, monitored, and controlled
3.5.3 Other FSR Technology
A few other companies such as FSA/Verg, Sensitronics, and LuSense produce FSR based
thin sensors used in force or pressure sensing and mapping. However, very little literature was
found describing these sensors and their performance. The following brief summary of other
FSR suppliers is provided.
3.5.3.1 Vista Medical
Vista Medical has a piezoresistive system called FSA (Force Sensing Arrays). This
equipment has also been referenced in the literature as coming from Verg, an affiliated company.
The company web site (www.pressuremapping.com) indicates its sensors, the FSR-F are “paper
thin, flexible , adaptable, modular - Singles, sets or matrixed arrays - up to several thousand
sensors, multiple sizes, shapes and pressure ranges…. The technology is piezoresistive, pressure
range 0 - 100 PSI ( 0- 35 kg/cm²), array size up to 32 x 32 sensing points, 1 – 1024 scanning
frequency, 10,000 sensors per second, sensor, size10 mm and up, and thickness 0.6 mm (.023" ).”
No information is provided on sensor performance such as accuracy, hysteresis, non-linearity,
frequency response, etc.
102

3.5.3.2 Sensitronics
Sensitronics (http://sensitronics.com) is indicated to be run by the “inventor of the Force
Sensing Resistor” Franklin Eventoff, who is identified as the founder and former president and
CEO of Interlink Electronics (FSR-I). The company’s website indicates its force –sensing
resistor characteristics are as follows: size –any including compound curves, range of 1 oz to 20
lbs or pressure from 1 psi to 125 psi, single part repeatability of +/- 5% of established nominal
resistance, force resolution of 1% full scale, and device “rise time” of 1 msec. No publications
were noted on the website.
3.5.3.3 LuSense
LuSense, part of IEE (International Electronics and Engineering)
(http://www.iee.lu/products), has a few references in the literature (Hollinger & Wanderley, 2006,
Emborg et al., 2007, 2009) to a model CPS 155 FSR-L. However, the IEE website has no
references to the products identified in these articles. The website does indicate the company
sells FSR’s. No technical information was found on the website, but Table 7 above shows some
performance data for the FSR-L (Hollinger & Wanderley, 2006). Relative to the FSR-T and the
FSR-I, the FSR-L had “lowest” hysteresis, fastest load recovery time, but substantial load
removal drift.

3.6 Capacitive based sensors and sensor arrays
3.6.1 Novel FSC-N (Force Sensitive Capacitor-Novel)
The most widely used capacitive based force sensors are those produced by Novel,
GmbH of Germany. The company's website does not provide a clear description of the
technology, but it does state “Pliance systems offer the state of the art technology for pressure
distribution measurement between soft and curved surfaces. The systems consist of a flexible and
103

elastic measuring mat, a multi-channel analyzer, a calibration device and a software package for
PC's….Pliance works with capacitive transducers in a matrix configuration. The elasticity of the
sensor mats permits perfect conformability to 3-dimensional deformations. The pressure
transducing elements contain high-tech elastomers manufactured by Novel. Restoring force,
range of force, threshold, hysteresis, temperature effect, frequency response and other
characteristics are determined during the manufacturing process. This makes it possible to adapt
the sensor characteristic to different measuring needs. The elastic measuring mats are available
in various sizes, sensor configurations and pressure ranges.”
(http://novel.de/novelcontent/pliance).
Novel primarily offers three different FSC-N based systems; the Emed, Pedar, and
Pliance systems. The Emed system is described as “high resolution sensor platforms for the fast
diagnosis of foot deformities or malfunction and as a pre and postoperative control in surgery.”
The system is very similar to a force plate, but it records pressure distribution. The PEDAR
system is an “in-shoe measuring system for gait analysis and control of footwear function in
therapy.” And the Pliance system is a “highly compliant elastic sensor mat system for the control
of seating and spine balance and the positioning of handicaps in wheelchairs to prevent decubitus
ulceration” (www.novel.de).
A recent update to the Novel website indicates Novel “offers three product lines, Emed,
Pedar and Pliance systems. Emed systems are pedography platform systems for functional foot
diagnosis. Pedar systems are in-shoe pressure measurement systems for the analysis of the foot-
insert-shoe function. Pliance systems are a product family for the analysis of pressure distribution
on all kinds of contact surfaces.” As the EMED and Pedar systems are specifically designed for
podiatric and gait analysis, the discussion will be limited to the Pliance system where possible.
All three systems appear to be based on the same capacitive technology. Given the extremely
104

limited information on the FSC-N accuracy, any discussions of accuracy for any of the three
systems will be included. Table 15 shows technical data for the Emed platform and Pedar system
in Table 16.

Table 15: Technical Data for the EMED platforms (emeATM_end.pdf, www.novel.de)

Table 16: Technical Data for the Pedar system (InsoleCatalogue072007.pdf, www.novel.de)
Technical data
sensor thickness (mm) 1.9
thickness of leadings 1.5
number of sensors 84 - 99
pressure range (kPa) 15 - 600
hysteresis < 7%
resolution (kPa) 2.5
temperature drift of Offset < 0.5 kPa/°C
frequency response (0-100 Hz) < 2dB
min. bending radius (mm) 20
pressure change due bending (kPa) < 20

105

According to the Novel website, the Pliance sensor family has been developed by Novel
for the special needs of researchers and clinicians. “All sensors are individually calibrated and
provide accurate and reliable pressure data. Standard sensors are available in various shapes and
sizes. They can be configured as single sensors or arranged in a matrix to fit different measuring
surfaces. Various pressure ranges can be achieved as well as various sensor thicknesses.
Flexibility and elasticity are two of the relevant characteristics of novel sensors. Proper material
selection and design give novel sensors the ability to conform around highly contoured sites
without wrinkling. An assortment of coatings can be applied to novel sensors. “Some novel
sensors can be sterilized and utilized in physiological environments in vitro and in vivo during
surgical procedure" (sensors_eng.pdf, at www.novel.de). Technical data from the company’s
website for the Pliance system are shown in Table 17.

Table 17: Technical Data for the Pliance system (Novel pliance_ftm.pdf)

Despite touting the accuracy of its systems, Novel only provides accuracy data for the
Emed system as shown in Table 15. Accuracy data was not found for the Pedar or Pliance
systems on the company’s website. The company does provide an Excel spreadsheet listing 906
references. However, many of the references are to internal Novel documents, unpublished
proceedings or are in German. In fact, sorting the references by date reveals that of the first 33
technology capacitive matrix sensor
number of sensors 256 ~ 1344
sensor resolution 2.0, 2.5, 6.0 cm² / sensors
pressure range 1 ~ 60 kPa
hysteresis < 3 % FSO
temperature coefficient 0.05 kPa x °C
elasticity approx. 4 %
min.bending radius 40 mm
106

references from 1983-1989, 26 are in German and only one of the other seven references could be
located through Ovid-Medline or via Google.
A document on the Novel North American web site titled “Why choose Novel” lists
seven papers that demonstrate “the accuracy and reliability of the Novel systems has been
discussed in much of the literature.” All of these documents are included in the current
chronological review.
The earliest document on Novel’s list of paper is by Hughes et al. (1991) who evaluated
only the reliability, and not the accuracy of, the EMED F-scan ground level pedography platform
system. Data recordings were taken of 10 volunteers walking at slow, medium and fast speeds, as
set by a metronome. Twenty-five recordings were taken at each speed for a total of 75 recordings
for each subject. As depicted in Figure 63, exponential increases in the coefficient of reliability
with increased number of walks was shown. Although the paper title is the “The Reliability of
the pressure measurements: the EMED F system” the result probably actually represents the
reliability of the combined methodology including the learning of the participants and increased
habituation. For this to be a pure assessment of the reliability of the FSC-N system, it must be
assumed there is no variability in volunteer response during repeated trials. The authors
concluded that “few measurements gave a sufficiently high level or reliability for a single
recording to be taken as representative.”

107

Figure 63: Coefficient of reliability as a function of the number of2 walks onto platform at
medium speed (Hughes et al., 1991)

At a recent conference on the assessment of pressure measurement devices (Giacomozzi,
2010), a statement from the “PMD (pressure measurement device) management at Novel”
indicated “accuracy better than ±5% for pressure values up to 1.25 MPa. The hysteresis of the
Novel Emed system is less than 3%.’’ The readers were directed to “detailed results see
references” (no references given). It further stated that sensor property “may change slightly over
time, calibration must be checked periodically.” “However, the only way to determine exact
accuracy of each individual sensor it is indispensible to use the Trublu calibration system.”
In a one page extended abstract, Kalpen and Seitz (1994) compared the FSC-N Pedar
system to the data from a Kistler platform (load cell). No data or graphs were presented, but they
reported the difference between the measured forces with the FSC-N and the body weight were
108

always less than 5%. The hysteresis of the tested measuring insoles was maximally 5% over the
entire measuring range related to the maximum load of 60 N/m
2
. They did note that the scanning
mode of the in-shoe device caused a delay time between the force values recorded with the Pedar
system and the Kistler platform. “When the heel area of the measuring insoles was loaded a
delay time of up to 40 ms had to be considered, decreasing to 20 ms if the forefoot area is
loaded.” To investigate the performance of the in-shoe device the measuring insoles were placed
directly under the foot and in leather shoes. In the first case the forces measured with the in-shoe
device and the normal forces from the Kistler platform were “similar with the exception of the
initial peak.” This was not quantified and potentially is a significant problem as focal accurate
peak contact pressures are particularly clinically relevant. In leather shoes, the forces recorded
with the FSC-N were much higher than the normal forces from the Kistler platform. The authors
provided the equally unsupported explanation that “this was caused by additional force
components resulting from stress in the bent leather sole and transferred horizontal force
components through the mechanics of the shoe.”
McPoil et al. (1995) tested both FSR-T’s and FSC-N’s. His methodology was previously
described in the FSR-T section with performance and comparison graphs of the two systems
shown previously in Figure 34 and Table 9. The methodology has questionable relevance to
dynamic testing, but as previously noted the data presented was not dynamic testing of the
accuracy of either system, but only consistency as represented by the Intraclass Correlation
Coefficients (ICC). The bench testing, as seen in Figure 34, showed that the FSC-N had an
average error of 16% at 50 kPa decreasing to 0.8% error at 500 kPa. The total creep for the FSC-
N was 3.4% at 10 minutes, and the error was linear with respect to time. In comparing the FSR-T
to the FSC-N, it was indicated the EMED insole has a linear response to applied loads with
minimal error noted at the various levels of pressure tested. This was especially true at high
109

pressures. The FSC-N (Emed) insole in comparison to the FSR-T (F-Scan) insole demonstrated
less creep during the application of a continuous and constant pressure. The “dynamic testing”
conducted on the FSC-N insole indicated a high level of reliability for between trials, between-
sessions, between-time, and both peak 1 and peak 2.
Quaney et al. (1995) compared the Emed FSC-N to the Pedobarograph using the same
subjects walking barefoot over both systems. Intraclass correlation coefficients were calculated
for each system and are presented in Table 18. There were no definitive statements of either
system’s accuracy. They did conclude that the systems did not produce identical measurements,
but “despite these differences, the values for each system are reasonably well correlated.”
Table 18: Intraclass correlation coefficients (Quaney et al., 1995)

Quesada et al. (1996, 1997, and 2000) conducted a series of pressure vessel and gait tests
to directly compare the FSC-N (Pedar) to the FSR-T (F-scan). In the first study (Quesada et al.,
1996) they utilized the Novel Air Bladder calibration set up as shown in Figure 64 with the
attached analog reference pressure gauge.
110

Figure 64: Trublu Novel Air Bladder calibration set up (hand%20sensor_eng.pdf,
www.novel.de)

Each set of insoles was placed in the calibration setup stacked one on top of the other and
loaded in 100 kPa steps up and down from 0 to 500 kPa. Average and standard deviations were
computed across a 3 cm x 3 cm area in the forefoot region. They found that across all data sets
and all conditions the mean absolute differences between the area average pressure and the
bladder pressure gauge value were 10 and 86 kPa for the PEDAR and F-Scan systems,
respectively. They concluded that the PEDAR system is likely the system of choice when the
greatest accuracy and repeatability are desired. Subsequent testing (Quesada et al., 1997, 2000)
involved relative comparisons of the two systems by stacking each sensor array onto of the other
into a show and conducting volunteer walking tests. Multiple tests with each volunteer were
conducted and the order of sensor stacking was reversed between tests. The authors noted that
which sensor array was closest to the foot (on top) had an effect on the pressure reading.
111

Intuitively, it would be expected that the lower sensor would likely show lower pressure as there
is an additional interface (the other sensor array) between it at the incongruent sole of the foot to
better distribute the force hence lowering the peak pressure as shown in Figure 65. However the
authors did not attribute these differences to this effect and instead asserted “differences in peak
values and variability of dynamic plantar pressure recordings can be more directly attributed to
differences in system performances, although the effects of insole orientation cannot be
definitively isolated.” The authors noted that their methodology and data cannot provide a direct
indication as to which system's measured peak pressure values are more correct. However; the
capacitive system demonstrated significantly lower overall variability of recorded dynamic
pressure than did the resistive system.

Figure 65: Effects of FSC-N sensor position (Quesada et al., 2000)

Boyd et al. (1997) assessed the reliability and validity of the Pedar FSC-N by comparing
in shoe measurements to Kistler force plate data. The area of the foot was divided into 8 regions
and the reliability within each region was assessed by the interclass correlation coefficient (ICC).
112

For peak pressure, ICC varied from a poor value (0.11) to an excellent value of 0.89. Validity of
the peak force values F1 and F2 (see Figure 40) showed ICC values of 0.84 and 0.81,
respectively. However, the authors noted that the forces from the FSC-N were consistently lower
with an average difference of 16% for F1 and 5.8% for F2. It was noted that placement of the
insole profoundly affects pressure measurements, especially in the great toe area.
In a similar fashion as Boyd et al. (1997), Barnett et al. (1999, 2001) also compared Pedar
FSC-N data to Kistler force plate data during gait tests. Data was sampled at either 99 or 100 Hz
and volunteers wore three different shoes (Trainers, Oxfords, and Slip-on deck type) during the
study. The five clinically relevant variables shown in Figure 66 were considered for comparison.
Their specific results are show in Table 19. Overall, for barefoot and shod data combined, there
was a mean difference between the two systems of 1.8% for temporal parameters, 6.3% for
impulse data, and 13.4% for force data. The authors noted that the Pedar FSC-N system has an
inbuilt threshold of 20kPa, below which force and pressure data is not recorded. This was noted
to be a 2 N threshold in the earlier paper (Barnett et al., 1999). It was noted that during the gait
cycle, part of the plantar surface loads were around or below this threshold. These characteristics
contribute to data obtained with Pedar FSC-N being generally lower than the Kistler load cell.

Figure 66: Gait parameters for comparison (Barnett et al., 2001)
113

Table 19: Comparison of Means (Barnett et al., 2001)

As described above in section 3.5.2, Hsiao et al. (2002) using a systematic stochastic
method examined the effects of applied pressure, duration of pressure application, calibration
pressure used and measurement week (measurements taken in different weeks) on the accuracy
and precision of the F-Scan FSR-T and the FSC-N using a rubber bladder and a “smooth” wood-
wood interface. Two insoles were tested, a new and a second that was one year old. Both sensors
were calibrated on the Novel rubber bladder (Figure 64) according to the manufacture’s
specifications. During testing, data was collected after 2 s, 5 min, and 10 min at each pressure
between 30 kPa and 500 kPa. Each experiment was repeated three times for 3days. A 4 (applied
pressure) x 3 (duration) x 2 (insole) x 3 (day) design was used for ANOVA statistical analysis.
All four main effects were statistically significant (p<0.01). In addition, the applied pressure x
day and insole age x applied pressure were statistically significant (p<0.01). Long duration
testing was conducted using the manufacturer's recommended extended calibration procedure and
114

data collection at 2s, 5, 10, 15 and 20 minutes. Each experiment was repeated three times for
3days. A 4 (applied pressure) x 5 (duration) x 3 (day) design was used for ANOVA statistical
analysis. All three main effects were statistically significant (p<0.01). In addition, the duration x
applied pressure and applied pressure x day were statistically significant (p<0.01). Lower
pressure threshold were examined by placing static weights on the sensors equivalent to applied
pressures of 12, 24, 35, 47 and 59 kPa. Accuracy was found to decrease significantly with the
decrement in applied pressure with the highest accuracy at 47 and 50 kPa with average errors of
+/- 1.3%. Measuring pressures below 35 kPa resulted in errors from -5.9% to -57.2%. Likewise,
precision increased with applied pressure. They noted the Pedar FSC-N had unacceptable levels
of accuracy and precision when measuring pressures below 35 kPa. The FSC-N system produced
greater accuracy and precision when the insole was new and measurements were taken (1) after a
manufacturer-specified calibration procedure, (2) in the 50 ± 500 kPa pressure range and (3)
within a few seconds after pressure was applied. Under this condition, the measurement error
was in the range 70.6 to 2.7%, and the magnitude (upper bound minus lower bound) of the 95%
tolerance intervals was in the range 13.5 ± 18.7%. Measuring 535 kPa with the Pedar system is
not recommended.
Polliack et al. (2002) used a prototype Pliance FSC-N sensor custom built by Novel to
examine pressure distributions within a lower limb prosthetic. The sensor was noted to be a 4x4
array of 16 capacitive sensors mounted in a 2.5 cm x 2.5 cm silicone substrate. Sensor thickness
was 0.64 mm with each sensor having an area of 0.25 cm
2
. Up to 16 sensor arrays could be
placed at various stump locations for a total of 256 sensors. Testing utilized the maximum
system sampling rate of 50 Hz. The sensors had a maximum pressure range of 600 kPa (88 psi).
The system can record a maximum of 4000 continuous frames or a maximum of 80 seconds at the
maximum 50 Hz sampling rate. Initial testing was conducted on using the Novel flatbed chamber
115

(Figure 64). Subsequently, a custom pressure vessel was constructed with a range of 0-350 kPa
(0-50 psi) for contoured surface testing. The custom pressure vessel allowed a positive mold of a
trans-tibial amputee stump with the electrical leads and attached sensors to be placed inside. The
sensors were attached at 9 sites as described in Table 20 with Micropore (3M Corp.) tape. Site 8
was later eliminated from analysis based on set up restrictions. Notably the smallest radius of
curvature tested was 17.1 mm (1.71 cm) [Table 20 indicates this minimum radius is at site 3;
however the text indicates this curvature is actually at site 4 - Patellar tendon]. Before each test,
the sensors were calibrated with the flatbed chamber at 5 pressure levels in 100 kPa increments
from 0-550 kPa (0-80 psi) starting at 50 kPa. Contoured mould tests were conducted at 50 kPa
increments from 0-350 kPa (50 psi).

Table 20: Sensor attachments sites on trans-tibial stump mould (Polliack et al., 2002) [note
correction to site 3 and 4]

Testing examined the sensor accuracy, effect of curvature on accuracy, hysteresis, and
drift. Loading was incremented at 50 kPa every 10 seconds to the 5 pressure levels. Drift was
measured for the flatbed and mould tests at 5 minute intervals for 20 minutes at a constant
116

pressure of 150 kPa (20 psi) and 400 kPa (60 psi) for the flatbed and at 100 kPa (14 psi) and 300
kPa (43 psi) for the mold. Results are shown in Table 21.

Table 21: Summary of results [mean+/- SD, n=sample size] (Polliack et al., 2002)

Comparing how the text describes Table 20 in conjunction with the errors shown in
Figure 67, the largest error of 31% +/-21% occurred at the smallest radius of 1.71 cm (17.1 mm).
The authors noted “The highly contoured patellar tendon site of the positive mould (radius of
curvature = 1.71 cm) used for all tests, may to be too great for the sensing elements to provide
consistent and repeatable results. This is surprising in view of the small geometry of each sensing
element as well as the general flexibility of the sensor matrix.” In all measures, the mold
(curved) sensors had both higher mean error and nearly two to three times higher standard
deviations. The durability of the sensor was described as “good” with only two sensing elements
out of a total of 160 failing. Small tears did appear on the edges of a few matrixes, but they did
not extend to the sensing elements. `

117

Figure 67: Pressure mean error +/- 1 SD at each stump location (Polliack et al., 2002)

Arndt (2003) examined sensor creep of the Pedar FSC-N (in-shoe foot pressure sensor)
under repeated cyclic gait loading for up to three hours. Sensors were calibrated using the
“standard Novel equipment and software.” Only two subjects were tested, with two different
boots, in normal gait and with a backpack weighing 49% of their bodyweight. Samples were
taken for 10 seconds after 0, 1, 2, and 3 hours of walking on a treadmill at 3.6 km/h. The sensors
were found to exhibit a total sensor difference of between -3 and +17% after three hours. The
average total sensor creep was found to be between 0.38 N/min to 0.96 N/min. Creep was noted
to be great for the heavier of the two subjects.
Martinelli et al. (2004) examined the performance of a prototype FSC-N Ankle Joint
Pressensor (AJP) designed for the ankle joint. The sensor was 1 mm thick, composed of 194 cells
(2.5 mm x 2.5 mm), a sensing area of 2.8 by 4.3 cm, and pressure range from 4.5 to 250 N/cm
2
.
Testing was conducted in a material testing machine at loads of 50, 100 and 150 N. Using flat
cylindrical indentors, contact area estimation was in error of 20% and 16% with contact areas of
3.3 cm
2
and 0.9 cm
2
, respectively. Cyclical load tests were conducted for 25 cycles at frequencies
of 0.1, 0.25 and 1 Hz with both flat and spherical indentors. The force error was found to be
below 2% for each load in all force ranges, did not depend on surface used, and the sensors did
118

not show a “memory effect” during cyclic loading. Utilizing spherical indentors of 3 cm and 5
cm radius “the sensors did not appear affected by crinkling artifacts, even if in the 3 cm radius the
force error increased to 15%. Creep and hysteresis were investigated respectively by applying a
known force for five minutes and running a cycle through three steps of 50 N, 100 N and 150 N
at a cycle frequency of 0.25 Hz. Creep manifested in an overestimation of the applied load of
about 15% after 300 sec and hysteresis after the 8th cycle.
Muller et al. (2004) used a custom made FSC-N 16x16 element array that was folded
spherically into a hemispherical cavity to simulate an artificial hip joint. No information was
given on calibration and the only description of load data was a note that at least some tests had
external loads of 250 or 400 N. No accuracy data was noted and the only sensor performance
information was that “the sensors were found to respond linearly to applied load."
Wasielewski et al. (2004, 2005) modified a standard FSC-N array to monitor
intraoperative pressures in total knee arthroplasty surgeries. The sensors were noted to have been
calibrated intraoperatively prior to insertion. Numerical results were presented, but no relative
sensor performance testing was conducted. The authors did note that the sensor array was
“highly conformable” to curved surfaces.
Hurkmans et al. (2004, 2006a, 2006b) examined the validity of vertical ground reaction
forces measured over a period of 7 hours utilizing a custom “humidity-proof” version of the
Pedar FSC-N. FSC-N data was collected from five subjects every hour for seven hours while
walking or standing on a Kistler load cell that also recorded the ground reaction force. The FSC-
N was calibrated using the Novel Trublu calibration device (Figure 64) and before measurement,
the system was turned on 1 h in advance (acclimatization period) and zero settings were done at t
=0 and t =1h. Their preliminary tests found a negative drift in the system data, which stabilized
after 1 h; based on this, the Novel company recommended an acclimatization period of 1 h after
119

which a second zero setting should be performed. After the second zero setting at t = 1h,
dynamic and static measurements were performed every hour for 7 h. For dynamic
measurements each subject walked at his own walking speed and positioned himself in front of
the force plate so that the third right footstep was placed on the platform. This was repeated 10
times for each subject every hour. For the static measurements, the subjects stood still on the left
leg only, followed by standing still on the platform for 10 s on the right leg only. As seen in
Figure 68 the drift was found to be relatively minor for the first three hours and then increase
after four hours. The mean drift after 7 hours was 14% (132N) for walking experiments and 16%
(141N) for standing experiments. Their testing indicated the drift was predominantly an offset
(all output values are increased at a certain time by the same value) drift as opposed to gain drift
(output values are increased by a multiplication factor). A Sensor Drift Correction algorithm
(SeDriC) was used to correct for the offset drift. After the correction, dynamic gait data first peak
(Figure 66) error was 5 to 12% and the second peak error was 1.1 to 3.4%. These differences
were only statistically significant for hours 1, 2, and 3. It was noted that all FSC-N data was
lower than the Kistler force plate data.

Figure 68: FSC-N drift measured during swing phase of five subjects (Hurkman, 2006b)

120

Hurkmans et al. (2006c) conducted a similar study where the same FSC-N Pedar insole
was used. Prior to testing, each array was calibrated in the in the Novel Trublu bladder device
(Figure 64) with calibration curves for 4, 7, and between 10 and 60 N/cm
2
at 5 N/ cm
2
intervals.
All measurements were taken after the system had been on for 1 hour and allowed to reach a
“thermodynamic equilibrium.” Static loading was conducted with the Novel bladder by applying
a load for 7 h with the Trublu calibration device at a pressure of 5 N/cm
2
for 7 h. Data were
collected for 2 s at 0, 10, 20 and 60 min, and subsequently every hour. To investigate whether the
type of drift is an offset drift the continuous load of 5 N/ cm2 was temporarily increased by 5
N/cm for 2 s every hour, and data were also collected during these 2s. This experiment was
repeated 3 days later, to determine day-to-day repeatability. A similar protocol was also used
with a load of 10 N/cm
2
. Long-term dynamic loading was investigated by placing the Pedar
Insole, between two steel plates mounted on the material testing machine. The array was then
loaded for 1200 cycles (session 1) with 300 N (load cycle: 2 s load-on, and 1 s load-off), using a
loading speed of 1 mm/s. Then the insole was not loaded for 3 h and then again loaded for 1200
cycles with 300 N (session 2). The system was active during the entire period of about 8 h. Data
were recorded during the first 10 cycles and then after every 100 cycles (12 data collection
periods in total). During the period between the two loading sessions, the Pedar force was
measured for 10 s every hour at 50 Hz. This procedure was repeated 3 days later. The protocol
was also performed with a cyclic load of 500 and 1000 N. The percent error of measurement for
the insoles during the static experiments ranged from -2.2% and 0.3% at hour 0. As seen in
Figure 69, from hour 1–7, the percent error of measurement varied from 10% to 43% depending
on the duration of loading and applied load. The relative drift after 10 and 20 min was 9–9.5%
when loading with a static load of 5 N/cm
2
, after which a more steady increase in drift was found
of 12.3% and 12.6% at hour 1 to a drift of 26.3% and 33.8% at hour 7. The difference in
121

measured force between the temporarily increased load and the continuous load (10 minus 5 N/
cm
2
, and 20 minus 10 N/ cm
2
, respectively) for all tested insoles was constant over time after hour
1, which indicated that the drift was an offset drift. During the first hour, the continuous load
curves showed a larger increase in force than the temporarily increased load curves. This
resulted, therefore, in a larger difference compared to hours 1–7.

Figure 69: Static FSC-N sensor drift (Hurkmans et al., 2006c)

As seen in Table 22, the percent errors for the dynamic loading experiments were much
larger for the 300 N loads (between -19.3 and -22.3% at the beginning) than for the 500 N and the
1000 N cyclic load, during both sessions. For the 300 N loads, the difference between the Pedar
data and the loading device data remained more or less the same with errors of between -16.4%
and -19% at 1200 cycles. The measured force showed an increase over time with the 500 N and
1000 N loads, leading to smaller errors with the applied load. During the load-off periods of the
dynamic experiments with 300 N loads, the mean total force was zero (Fig. 6). An increase in
force of 10–12 N was measured during the load-off periods at the end of sessions 1 and 2,
122

respectively, for the 500 N cyclic loading experiments. The mean increase during the 1000 N
was 22–55 N at the end of sessions 1 and 2, respectively. The mean percentage day-to-day
differences for the 300, 500 and the 1000 N load at session 1 were 3% (range 1.9– 4.4), 2.1%
(range 0.5–4.4) and -2.9% (range -4.2 to -0.8), respectively. The mean percentage day-to-day
differences for session 2 were 1.9% (range 0.1–3.9), 0.6% (-1.4 to 2.1) and -1.9 (range -4.0 to -
0.7) for the 300, 500 and the 1000 N load, respectively.

Table 22: Dynamic measurement error of the FSC-N

As described in the FSR-T section, Martinelli et al. (2006) conducted a detailed
comparison of the FSR-T and the FSC-N (refer to prior section for methodology) including using
flat, curved and spherical loading surfaces (Figure 55) although of questionable relevance to the
FSC-N Ankle Joint Pressensor (AJP) they evaluated. The maximum force error (E
force
) for the
FSC-N (range, −3% to +5%), area error (E
area
) was within 6% for areas between 3.3 cm
2
and 6.6
cm
2
. Repeatability was better for the FSC-N system for flat and cylindrical surfaces. The E
force
in
the cylindrical load setup was within 8% for the FSC-N, 8% with the 50-mm radius sphere and
12% with the 30-mm sphere. Creep artifact (Cr) was 16% for the FSC-N.
123

Rikli et al. (2007) evaluated a custom designed FSC-N designed to evaluate the pressure
distribution in the radioulnocarpal joint. The only mention of calibration was “the sensor was
calibrated and then introduced through the incision.” Hysteresis was evaluated by “loading in a
test machine” [no discussion of what the test machine is or what the interfaces are] with
increments of 5 N/cm
2
from 5 to 60 N/ cm
2
, and then unloaded. In the first test, each step load
was held for 20 seconds and in the second test the step was held for 60 seconds. Hysteresis did
not exceed 7% and was less with the step load was held at the long 60 second hold. Temperature
effects were examined by loading the sensor with a constant load of 7.5 N and then raising the
temperature from 17 ºC to 45 ºC and back down to 23 ºC. The sensor values were found to
decrease with increasing temperature at a rate of -0.06 N/ºC. Humidity was examined by
submerging unloaded and loaded sensors in a mepivacain 0.5% solution. Unloaded sensors were
found show a decrease of the measured forces of 4 N/ cm
2
and the loaded sensors did not change.
The response of the sensor to gas sterilization was also examined and was found to change the
output values less than 1%. Reproducibility was checked by repeatedly loading the sensor with a
cadaver wrist and the mean coefficient of variation was found to be 2.1% for the baseline force
and 1.2% for the recorded force. In addition to providing no information on the methods of
calibration the authors did not quantify or even consider the effects of the joint curvature. Joint
curvatures in the scaphoid radius are between 4.1 and 18.7 mm (Werner et al., 2007).
Putti et al. (2007) evaluated the consistency (repeatability) of the Pedar in-shoe FSC-N
and then, presumably after establishing the repeatability they wanted to “establish a range of
normal in shoe pressure values.” As has been done by prior authors, they appear to be assuming
without any foundation, that high repeatability implies accurate data. The 99 sensor FSC-N was
calibrated in the Trublu bladder calibration system (Figure 64). The mean number of days
between repeated data collection was 11.8 days (range of 1-32 days). Fifty-three volunteers were
124

used. Eight steps of in-shoe data were collect on each run. For analysis the foot was divided into
10 regions and 18 parameters monitored by the Pedar software were considered. In total, 122
parameters were assessed. Their results are presented in Table 23.

Table 23: In-shoe repeatability results (Putti et al., 2007)

The coefficient of repeatability (CR) was defined using the formula “(coefficient of
repeatability/mean)*100”. For the 122 parameters assessed, the highest CR was 15.3% with 114
out of 122 parameters (93.4%) having a CR less than 10%. The highest CR of 15.3% was stated
to be “clinically acceptable” given the normal variations in steps and the Pedar system was
considered repeatable. Putti et al. (2008) published a very similar study evaluating the EMED
FSC-N platform and found the CR was less than 10% in 111 of 122 parameters with the highest
CR of 16.9%. Based on this testing, the platform based FSC-N was also deemed to be repeatable.
Accuracy data was not present in either study by Putti et al.. Although repeatable, these studies
are insufficiently designed to conclude these results are in fact accurate and therefore
representative of the “normal foot.” Similar testing and repeatability results were observed by
Gurney (2008).
125

Ellis et al. (2008, 2009), utilized a custom Pliance FSC-N 32 sensor array with a
resolution of 1.85 sensors/cm
2
. In the one page extended abstract, the authors noted the loading
apparatus consisted of an aluminum base plate that was advanced with a turn screw mechanism
against the plantar foot. “Test pressures” ranged from 0-6 Bar in increments of 0.5 Bar. Root
mean square error (RMSE) between the input and measured pressure was used to assess
accuracy.” (Note: in the 2008 paper the authors do not describe how the input pressure was
measured). The “objectivity” of the plantar pressure system was assessed by applying standard
loads in the “Trublu calibration device” (Figure 64). In reality, all they showed was that the
sensor was still calibrated, not that it was accurate for measuring plantar pressure. Ten subjects
were used with loading of forces of 25%, 37.5% and 50% of body weight. Subjects were also
tested in vertical posture and gait across an EMED-X sensor platform (a platform mounted FSC-
N). Given the lack of a clear description of how the input pressure was measured, the results
should be considered carefully. The authors noted the sensors were linear (R2=0.9996) and
accurate (RMSE = 0.15 Bar). The ICC (2, 1) values demonstrate that the sensors were reliable.
The authors concluded their measurement system provides an accurate, linear, and reliable
method to measure plantar pressure parameters in the supine subject. This study should not be
considered as substantially proving the accuracy of this system for plantar pressure measurements
as all they only showed the sensor was still calibrated, not that it was accurate.
Hillstrom et al. (2008) examined the performance of the plantar pressure measurement
systems: the EMED-X (FSC-N) and the MatScan (FSR-T). The authors noted that construct
validity requires that a system be both accurate and reliable. “To be reliable is necessary but not
sufficient because it is possibly to be repeatedly wrong.” Both systems were tested in a bladder
based calibration test apparatus, shown in Figure 70, at regular discrete values between 0 and 850
KPa in 50 KPa steps. Measured values from each system were compared to the applied pressure
126

from a reference transducer on the calibration device. Due to sensor saturation, the FSR-T
MatScan required two sets of calibration curves. “Short term” and “long term” loading profiles
were noted, but not defined. Figure 2 in the paper is noted to show the “short term” loading
regime and appears to show a rise to approximately 850 KPa in steps with a 5 second rise and
then a 5 second hold, and a similar step decrease back to zero. “Short term” accuracy was a
quantified as a root mean square error (RMSE) and it was for the 13.2 KPa (1.56% FS) from the
FSC-N and 49.93 KPa (5.83% FS) for the FSR-T. Utilizing the standard FSR-T calibration
technique resulted in an RMSE for the FSR-T of 21%.

Figure 70: Bladder calibrator and test fixture (Hillstrom et al., 2008)

Potthast et al. (2008) utilized the Pliance Ankle Joint Sensor FSC-N for quantifying the
contact pressure across the talocrural (ankle joint) [composed of the distal tibia, distal fibula and
dome of the talus]. In their literature review, they explicitly mention prior studies that had
problems “such as the ‘crinkle artifact’ caused by the use of film between curved joint surfaces.”
Ironically, their FSC-N was “calibrated by the producing company with 15 steps for a maximum
range of 2 MPa.” No further calibration information was provided. As shown in
Figure 54, the talus has radii of curvature from 16-28 mm. In light of their measurements on
highly curved surfaces with a sensor that was not likely calibrated correctly, the results of this
study should be viewed with caution.
127

Johanson et al. (2009) utilized a Pliance S2039 Patella Sensor FSC-N. The authors stated
their purpose was to “use existing validated technology in the measurement and distribution of
patellofemoral forces and pressures…” They cite the previously discussed works of Martinelli et
al. (2006) and Putti et al. (2007) in support. However, the smallest radius Martinelli et al. (2006)
tested was 30 mm (also this had the highest error in force reading of 12%). Considering
curvatures as small as 250 m
-1
(radius of 4 mm) on the patella and as small as 175 m
-1
(radius of
5.7 mm) on the mating femoral surface have been reported (Kwak et al., 1997), citing Martinelli
et al. (2006) does not support a claim of validity for the proposed application in the
patellofemoral joint. Putti et al. (2007) is a study of plantar pressure systems and is even less
supportive of the validity in this application. The current author did note their sensors were
calibrated in the (flat) Trublu calibration device (Figure 64) prior to each test.
Lai and Li-Tsang (2009) examined the Pliance X FSC-N system that was stated to have
been designed for low interface pressure applications between skin and pressure garment or
bandages. A 10mm by10mm 4mm thick plastic disc was placed under the pressure sensor such
that all forces of the loading weight were transferred only to the sensor. The test was to examine
the accuracy of the pressure device by measuring the proportionality of the sensor response to a
range of standardized loads. The device was calibrated before the experiment. “Readings per
weight were recorded 1 min after loading with 10 repetitions.” Pearson’s Moment Correlation
was used to determine the linear relation between the applied forces and sensor values.
Evaluation of the test–retest reliability of Pliance X System was also conducted through analysis
of intra-class correlation coefficient, ICC (3, 10). The same procedure with five trials was
repeated by three independent assessors for its inter-rater reliability which was analyzed by ICC
(2, 5). In vivo pressure measurement were conducted to examine the effect of human skin
elasticity on the sensor accuracy, the sensor was re-assessed with the measurement of the
128

interface pressures between the human skin surface and a sphygmomanometer cuff [25,40]. The
sensor was taped onto the skin underneath the cuff of the mercury manometer which was used to
measure the pressure. Eight anatomical locations, namely lateral aspect of upper arm (10 cm
above elbow), forearm (10 cm below elbow), thigh (10 cm above knee) and calf (10 cm below
knee) were selected on five volunteers. Measurements were taken starting from 10 to 50 mmHg
with increments of 10 mmHg and the procedure was repeated five times. Linearity and
repeatability on rigid surface had coefficient of variation ranges from 1.097% to 8.450%, with the
higher variations in the lower range forces. The difference between the applied pressure and
sensor output were less than 1 mmHg (0.175 ± 0.264 mmHg). Correlation between the applied
pressure and sensor output was revealed with Pearson’s Moment Correlation r = 0.998 and
adjusted r
2
= 0.997. Test–retest reliability was revealed with an ICC (3, 10) of 0.998. The inter-
rater reliability ICC (2, 5) among the three independent assessors was 95% with CI from 0.995 to
1.00. In vivo pressure measurement showed no significant discrepancies between the readings of
sphygmomanometer and the Pliance X system. The maximum mean percentage difference was
6.42%.
Giacomozzi (2009, 2010) conducted a “scientific project aimed to design, validate and
implement dedicated testing methods for both in-factory and on-the field assessment of plantar
pressure measurement devices (PMD). The five commercial PMD’s described in Table 24 were
used.

129

Table 24: Characteristics of the five tested PMD’s (Giacomozzi, 2010)

The following tests were conducted over 5 randomly selected areas: (1) 100 kPa steps of
static pressure from 0 to 600 kPa and back to 0, each step lasting 5 s, the area being completely
offloaded after each step; (2) sinusoidal pressure cycles (0–500 kPa; 0.75 Hz; at least 10 cycles)
applied through the proportional valve; (3) constant pressure (350 kPa; 60 s) to investigate creep.
Center of pressure (COP) measurement was also extensively tested. Interestingly, Tekscan (FSR-
T) refused to participate in the study (Giacomozzi, 2008), but its system was obtained on the open
market. The Tekscan MatScan FSR-T was calibrated with a “special on-site calibration
procedure” previously described (Hillstrom et al., 2008). Early results including the RSSCAN
data are show in Figure 71. However, in the full paper (Giacomozzi, 2010), the authors noted
that “RSSCAN data had not been reported in this Technical Note, since a deeper investigation has
been agreed with the Company.”

130

Figure 71: Static pressure measurement results measured over the whole surface
(Giacomozzi, 2008)

The results of their testing are presented in Table 25. Relative to the FSC-N and FSR-T,
the authors concluded the FSC-T showed high linearity, low creep, low hysteresis and high
correlation under slow sinusoidal loading, high accuracy and precision in center of pressure
(COP) estimation, low variability of all performances over the whole sensor matrix. After
“dedicated calibration” the FSR-T showed high linearity and moderate spatial variability, low
even if variable creep, low hysteresis and high correlation under sinusoidal loading, high
accuracy and precision in COP estimation except for one tested area.

131

Table 25: Results of PMD testing

3.6.2 XSENSOR FSC-X (Force Sensitive Capacitor-XSENSOR)
X-Sensor is a capacitive based system primarily designed for seating comfort, laying
body pressure distribution, and some industrial applications. The smallest standard sensor size is
4 inches by 4 inches (http://www.xsensor.com/media/upload/X3-Sensor-Chart.pdf). No research
papers were noted on the company website (http://www.xsensor.com/pressure-imaging/research-
articles) that are pertinent to the smaller sized sensor that would be used in an orthopedic
application.
3.7 Mathematical and Finite Element Models
Early mathematical analysis (Morris et al., 1961, Fiorini et al., 1976) using static
equilibrium models showed that a normal male could develop approximately 42 lbs of load, also
reported as 271 lb/in^2, in the facet joints by simply bending forward to 40 degrees. A prolific
132

publisher of lumbar spine F-E models (Shirazi-Adl & Drouin, 1987) noted that a number of
previous models up to that point had either not included the facet joints or improperly modeled
them. The results of his non-linear F-E model showed that in pure compression, the external
axial force is transmitted primarily by the intervertebral disc, with the facet joints only carrying a
small percentage of the load. The facets were noted to carry large loads in extension. Results
were compared to prior empirical data.
As computing power increased and the costs of cadaveric specimens increased, there has
been a progressive increase in the number of mathematical models, primarily finite element (F-
E), of the spine. An Ovid search using the key words “finite element” and “lumbar” returned 344
results from 1996 to 2010. Further limiting the results with the addition of the keyword “facet”
still left 72 references, with 13 of them in 2009 alone. Mathematical models have the inherent
advantage of being able to easily change various parameters, such as removing elements,
changing size(s), or changing material properties and quickly and at relatively low cost examining
the effects of the changes. However, the validity of the model is often judged by how accurately
it represents data taken from actual anatomical specimens. This is typically true for both
component level material properties as well as whole specimen response. The validity of this
form of analysis is therefore dependent on the previous methods that provide quantitative data for
validation.
3.8 Preliminary Evaluation of Existing Technology
In evaluating existing technology, the most prevailing limiting factors were found to be
thickness, ability to accurately measure on curved surfaces, and cost. The most non-intuitive
finding of this technology review was that the published literature on the systems' capabilities
was never found to be accurate on the systems that were tested when it came to thickness or
measurement accuracy on curved surfaces.
133

Initial screening of the possible new technologies was done via a literature search through
Ovid Medline from 1950 to the present and via web searches.
As noted by the volume of literature reviewed above, two general types of sensor
technologies were found to be the most prominent. These involved sensing elements that were
either resistive based systems (strain gauges-resistors, conductive ink, or piezoresistive) where
the load change was quantified based on a voltage drop due to a change in resistance such as the
FSR-T, or capacitive based systems where the load was quantified based on a change in
capacitance such as the FSC-N.
Systems were evaluated from Sensors Products Inc. (Xsenor and Tactilus), Vista
Medical/FSA (Force Sensitive Applications), Novel (FSC-N), Pressure Profiles (Tact Array),
Tekscan (FSR-T), force sensitive resistors (FSR) from Interlink Electronics, and ultimately
pressure sensors from Precision Measurement Inc. The vast majority of these systems clearly
focused their designs in the comfort, podiatric and more therapeutic arena. Most companies had
systems for evaluation seating (comfort, pressure ulcers, ergonomics, wheel chair design) or
podiatric sensors (shoe fit, gait abnormalities, pressure ulcers).
Given the unique design requirements of the facet joints and the many ultimately
unsubstantiated claims made by sales representatives, it was found more productive to bypass the
sales departments and deal directly with the engineering or design departments of these various
companies. Numerous e-mails and phone calls were exchanged with the various companies and
technical personnel. Representatives from Vista Medical flew out for a meeting at the lab to
discuss design requirements. This author organized and attended a meeting at Pressure Profiles to
discuss designs with its engineering staff. A prototype array was sent by Sensors Products Inc.
Extensive testing was carried out on the Tekscan system.

134

3.8.1 Sensor Products Inc.
These sensors appeared the most promising based upon the initial reviews. Therefore, a
prototype sensing array was acquired from the company. The sensor was ultimately too wide and
too thick. The engineering staff indicated they would not recommend the system with curvatures
smaller than 0.5 inches (12.7 mm). The system was quite expensive. Initial checks of the system
accuracy relative to the in lab MTS machine showed unacceptably high discrepancies.
3.8.2 Vista Medical/FSA/Verg Inc.
Oddly, this company had one of the largest lists of references, but did not provide any
information on frequency response, hysteresis, or accuracy in its publications. The provided
information on the system seemed to stress the system as a relative comparison tool or as a
method to determine the location of high relative contact forces, but the absolute numerical force
readings are not noted or emphasized. Following a meeting with the company’s representatives
in which the company representatives stated the system would not work for the proposed
applications, this technology was deemed insufficient and no further inquiry was made.
3.8.3 Novel
Novel was the second to last company that was consulted. It was informed of the
problems with the other companies’ sensors and it indicated it could meet all the design
requirements. Based on the company's representations, the company's system met the physical
requirements, but the system with only 3 sensors was priced at $49,980 in 2003. Follow-up
contact was made with the company in July of 2007 in the hope that advancement in technology
had resulted in a reduction in the price. This was not the case. The company's new system now
costs $74,000. This was considered too expensive in 2003 for unproven technology and even
more expensive in 2007.

135

3.8.4 Pressure Profile Systems
According to the company’s website, this system had been used by Dr. David Mason at UC Davis
to measure the normal pressure and forces in the canine elbow and their changes under various
surgical techniques. This is a close analog for the current study. Dr. Mason was contacted via e-
mail and phone and he indicated he had spent a year looking at other technology before using this
system. Dr. Mason did indicate he had problems with sensor durability and the sensors lacked a
large range of sensitivity. Overall, he indicated “it worked pretty well for me.” This author met
this company’s engineering staff after numerous exchanges of e-mails and a preliminary sensor
design. A cost quote was generated. Ultimately it was determined that to give the sensor some
protection from the saline sprain, it would be greater than 1.2 mm thick. According to the
company's engineering staff, a 0.5 inch (12.5 mm) curvature would induce “as much as 50%
error.” Although the total system cost was favorable, the geometric problems superseded.
3.8.5 Tekscan
Based on an initial review of the literature, postings by users on Biomch-L, and personal
communications with another USC graduate student who had used these sensors for his Ph.D.
research, the Tekscan sensors were determined to be too non-linear, calibrations were loading rate
dependent, the sensor required pre-condition before every usage, and they were not durable. This
technology was therefore initially not considered. However, three recent publications (Rousseau
et al., 2006a; Rousseau et al. 2006b, and Wilson, 2006) were reviewed that utilized this
technology to quantify loads within the lumbar facet joints. In the light of the recent publication,
it was deemed necessary to re-investigate this technology.
A total of 24 tests were conducted related to the Tekscan. The initial test series looked at
the Tekscan’s performance on flat and curved surfaces (r=9.0 mm) relative to the selected sensors
from Precision Measurement. The second series of tests looked at the performance of the
136

Tekscan as a function of loading area. In short, the Tekscan did not perform well. The response
was non-linear, strain rate dependent and contact area dependent. An example plot of the data
from testing on a flat surface is shown in Figure 12. The left vertical axis and red graph is the
percentage error compared to the Enduratec material testing machine’s calibrated load cell and
the right vertical axis and green line are the normalized force output from the Tekscan. As seen
in Figure 12, the Tekscan shows maximum error of nearly 25%. This data is consistent with
testing of the Tekscan reported by Wilson et al. (2006) where errors of between 18+/-9% and
35+/-7% were cited over the load range considered in Figure 72. This level of error was
determined to be unacceptable. Figure 73 shows an overlay of the Enduratec material testing
machine calibrated load cell, normalized Tekscan force and normalized Precision Measurement
pressure sensor data.

Figure 72: Tekscan Error Relative to Tekscan Measured Load
137

Figure 73: Normalized performance of the Tekscan, Enduratec and new pressure sensors.

3.8.6 Interlink/Force Sensitive Resistors (FSR)
After going through four different companies with no success, the previous technology
utilized by Hedman (1992, 1995) was reevaluated. New FSR sensors were ordered from the
same company (Interlink Electronics, Camarillo) as from Dr. Hedman’s original research. The
pre-test, test and post-test calibration and sensor preparations outlined by Hedman (1992, 1995)
were followed. The output signal from the FSR signal is dependent on both the magnitude of the
load and the contact patch or contact area size. The original methodology developed by Dr.
Hedman with the FSR’s involved recording the sensor output voltage while the sensor was loaded
with three different area plungers (63 mm
2
, 95 mm
2
and 126 mm
2
) over a range of applied forces.
Then a graph of the Log of the applied force (Log F) was plotted versus the Log of the sensor’s
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Time (sec)
Normalized Magnitude

LOAD CELL
Pressure Sensor #1
Pressure Sensor #2
Force Transducer
138

resistance (Log R) for each diameter plunger. In Dr. Hedman’s prior research, this type of Log-
Log data produced a series of linear plots with a logical offset of the curves related to the plunger
area. In the spine, the FSR was backed by Fuji pressure sensitive film that left a visual record of
the total combined contact area for the entire test (no temporal resolution). The contact area for
the test was then calculated and a new Log F versus Log R curve was created by interpolating
between the curves base on the measured contact area. However, while calibrating the new FSR
sensors it was found that they were no longer giving a logical progression of the curves. As
shown in Figure 74, not only do the curves cross or intersect, but the smallest diameter plunger is
in between the medium and large diameter plungers. Additionally, during the calibration and
testing process, many of the sensors delaminated. Both of the these observations were discussed
with Dr. Hedman and the only logical conclusion was the materials or manufacturing process for
the FSR had changed and they were no longer suitable for this application.

Figure 74: Calibration Plots for FSR’s.
Log F Vs. Log R
y = -39.8481x + 126.1748
R
2
= 0.9786, 126 mm
y = -48.4984x + 152.9594
R
2
= 0.9905, 95 mm
y = -61.8902x + 194.8029
R
2
= 0.9876, 63 mm
0
0.5
1
1.5
2
2.5
3.11 3.115 3.12 3.125 3.13 3.135 3.14
Log R
Log F
126 mm
95 mm
63 mm
Linear (126 mm)
Linear (95 mm)
Linear (63 mm)
139

A summation of the data is shown in Table 26. Following the preliminary evaluation and
extensive literature review, the sensors from Precision Measurement were ultimately selected for
further testing and validation as they preliminarily met all the design requirements. However,
given this researcher’s experience with erroneous manufacturer published claims and apparently
contradictory claims in the literature with other systems, extensive independent validation was
necessary and was subsequently performed.

Table 26: Existing Technology Evaluation (* refers to unverified claims in product
literature, n/a – no information was located, # combined thickness of the sensor, Fujifilm
and tape)

Criteria
Company
Width and
Height Thickness Curvability
Water
Spray Durability Hysterisis Accuracy
Steady State
Response Cost
Sensor Products Inc.
(Xsensor/Tactilus) N/Y N (1.2 mm) N Y
300
impressions* +/-5%* +/-10%*
Not likely-
capacitive
elements $27,686.50
Vista/FSA N/N N N N n/a n/a n/a
Likely-
Piezioresistive
$4000 (just
sensors)
Novel Y*/Y* Y* Y* Y* Y* <3%* +/-10%*
Not likely-
capacitive
elements
$49,980 (3/03)
$74,000 (7/07)
Pressure Profiles
(Tact Array) Y/Y N (>1mm)
N (50% error
at r=12.5 mm) Y* n/a +/-5%* n/a
Not likely-
capacitive
elements $13,715.00
Tekscan (Flexiforce) Y/Y Y (0.21 mm) N Y Y +/-4.5%*
+/-5%*
(+25%) Y-Resistive Ink
$21/sensor
w/circuit
Force Sensitive
Resistors (FSR) Y/Y
Y (0.46 mm)
(0.570-.620mm)# Y N N n/a N
Likely-
Piezioresistive $4.00/sensor
Precision
Measurement Y/Y Y Y Y N +/-1.0%+/-7%Y-Strain Gauge$220/sensor
140

The preliminary selection criteria shown in Table 26 are described as follows:
 Width and Height - width (FCW) of approximately 12.0 mm and a height (FCH) of 15.1
mm
 Thickness - total sensor and any coating less than or equal to 0.8 mm. This included
stack up in sensor edge and cables.
 Curvability - be able to curve the sensor in the concave portion of a circular curved
surface with a 9 mm radius and induce less than 5% full scale signal and maintain linear
load response.
 Water Spray – Be water resistant to saline spray.
 Durability – be able to survive at least 5 test cycles
 Hysteresis – less than or equal to 5%
 Steady State Response – able to record steady state load down to 0 Hz. With less than
10% drift.
 Cost – Where available, the total cost of all hardware, software, sensor design, and at
least two sensors.
141

CHAPTER 4: SENSOR PERFORMANCE
CHARACTERISTICS

4.1 Introduction
There is a high frequency of back complaints in modern society with 60–85% of all
people experiencing back pain at some time in life (Anderson, 1997). Although the etiology of
low back pain remains controversial; one study of chronic spinal pain (Manchikanti, et al., 2009)
found a 31% prevalence of facet joint pain in patients with chronic lumbar spine pain. Despite
this finding, relatively little research has been performed accurately examining intra-articular
facet joint loads. More and more motion sparing spinal prosthetics are being introduced with
little understanding of their effects on one of the three main load bearing elements of the spine,
the facet joints.
Numerous methods have been used to measure the loads transmitted intra-articularly
through various joints within human and animal bodies and between various body parts and the
external environment. Early attempts (Ewing et al., 1972, Prasad et al., 1974, Prasad & King,
1974, King et al., 1975, Butterman et al., 1991) included externally mounted strain gauges that
inferentially measured the load within the joint by external measurements. These externally
mounted strain gauges are also prone to strain coupling (Zhu et al., 2008). Fuji Prescale Pressure
Sensitive Film (Fujifilm Corp., Tokyo, Japan) allows direct intra-articular measurements and has
been used in this capacity since 1980 (Fukubayashi & Kurosawa, 1980, Wiseman, et al., 2005).
However, the film has no temporal and limited spatial sensitivity as well as being capable of only
capturing the total aggregate load. The film is also sensitive to load rate (Fujifilm website 2010,
Rudert et al., 1987), temperature (Fujifilm website 2010, Liggins and Finlay, 1992), humidity
(Fujifilm website 2010, Liggins and Finlay, 1992), requires interpretation of color gradations of
variable optical density often with a 5
th
order polynomial fit (Fujifilm website 2010, Liggins et
142

al., 1995, Liggins & Finlay, 1997, Liau et al., 2001, Hale & Brown, 1992, Caldwell et al., 1993),
is prone to fluid contamination with protective measures possibly altering the factory calibration
curves(Liggins et al., 1995, Hedman, 1992), has a 50 hour development time (Liggins and Finlay,
1992), and may be subject to “crinkle” artifacts on bicurvillinear surfaces (Caldwell et al., 1993).
Newer technology such as force sensitive resistive arrays from Tekscan (FSR-T) (Boston, MA,
USA) show promise and continually improve, but have been shown to have high hysteresis
(Ferguson-Pell & Cardi, 1992, 1993, Pavlovic et al., 1993, Woodbum & Helliwell, 1996), high
creep (Ferguson-Pell & Cardi, 1992, 1993, Werner et al., 1995, Pitei et al., 1996, Otto et al.,
1998, Otto et al., 1999), calibration requirements around a priori load and material interface
properties (Pavlovic et al, 1993, Luo et al., 1998), temperature sensitivity (Werner et al., 1995,
Luo et al., 1998) poor dynamic response (Werner et al., 1995, McPoil et al. 1995, Pitei et al.,
1996, Sumiya et al., 1998, Otto et al., 1998, Otto et al., 1999), repeatability problems (McPoil et
al. 1995, Woodbum & Helliwell, 1996, Pitei et al., 1996), memory artifacts (Pavlovic et al, 1993,
Pitei et al., 1996), and up to 300%-700% errors when used in physiologically relevant curved
surfaces (Ferguson-Pell et al., 2000, Buis & Convery, 1997). A similar but capacitive based array
type sensor from Novel, a force sensitive capacitive Novel (FSC-N) (Munich, Germany), has not
been as thoroughly tested as the FSR-T but has been found to have lower hysteresis (Werner et
al., 1995, Giacomozzi, 2009, Giacomozzi, 2010), better linearity (Werner et al., 1995,
Giacomozzi, 2009. Giacomozzi, 2010), and less creep (Werner et al., 1995, Arndt, 2003,
Martinelli et al., 2006, Giacomozzi, 2009, Giacomozzi, 2010). However similar problems with
measurements on physiologically relevant curved surfaces (Polliack et al., 2002, Martinelli et al.,
2006) with errors of 31%+/- 21% at a 17 mm radius (Polliack et al., 2002) have been identified.
The majority of testing on the FSC-N has focused on plantar pressure (gait) accuracy and
reliability with very little intra-articular research.
143

Prior to implementing a sensor in in-vitro or in-vivo environments, it must be thoroughly
tested and evaluated under controlled laboratory conditions (Liggins and Finlay, 1992, Liggins, et
al., 1995, Liggins & Finlay, 1997, Liau et al., 2001, Hale & Brown, 1992, Caldwell et al., 1993,
Hedman, 1992, Ferguson-Pell et al., 2000, 24, Werner et al., 1995 McPoil et al. 1995, Woodbum
& Helliwell, 1996, Pitei et al., 1996, Buis & Convery, 1997, Luo et al., 1998, Sumiya et al., 1998,
Otto et al., 1998, Otto et al., 1999, Polliack et al., 2002, Martinelli et al., 2006, Giacomozzi, 2009,
Giacomozzi, 2010, Wilson et al., 2003, 2006). Our objective was to develop a new sensor array
capable of more accurately quantifying the spatial and temporal pressures, with particular
attention to eliminating sensor array curvature artifacts. This chapter presents the evaluation of a
new sensor array and method for directly measuring the time-dependent and spatial distribution
of pressure. The advantages and limitations of this new sensor array were thoroughly explored by
subjecting it to a variety of commonly performed bench top tests. Specification for interface
pressure measurements sensors have been proposed (Ferguson-Pell et al., 2000, Giacomozzi,
2009, Giacomozzi, 2010, Wilson et al., 2006, Wilson et al., 2003) and the following criteria and
parameters for the current system will be discussed and evaluated:
1. Thickness – Total sensor array less than or equal to 0.8 mm thick.
2. Scalable geometry – Down to less than 12.0 mm in width and 15.1 mm in height.
3. Flexibility in sensor spacing - Spatial resolution including individual element
location, number of sensor and effects of sensor spacing on adjacent sensors (cross-
talk).
4. Curvature effects – Less than 1% zero load offset and less than 1% change in
sensitivity on curved surfaces down to a 9 mm radius sensor array curvature.
5. Frequency response – Less than 2% loss of signal power (less than 0.09 dB) at
frequencies from steady state (0 Hz) up to 5 Hz.
144

6. Non-linearity – Less than 1% of full scale (FS).
7. Drift – Less than 1% of full scale out to at least 700 seconds.
8. Hysteresis – Less than 1%.
9. Repeatability – Coefficient of variation less than 2%.
10. Total cost of utilization including hardware, software, and sensors.

Although non-linearity, drift, hysteresis, and frequency response are all part of the overall
sensor array accuracy, and in some instances are difficult to differentiate, the test methods
employed here endeavored to isolate and test each specific property as thoroughly as possible.
The total sensor array uncertainty will be a combination of the above described system
parameters in addition to any random parameters (Fraden, 2003).
4.2 Methods
4.2.1 Description of new sensor array construction.
The individual sensor elements are composed of two beryllium-copper surfaces with a
strain gauge between them. The individual sensors used for this study have a manufacturer stated
linear operating range from 0 – 3.4 MPa (Model 060, Precision Measurement, Ann Arbor, MI,
USA). The sensor’s three lead wires are Teflon coated #36 (0.127 mm diameter) stranded copper
wire. As shown in Figure 75, the individual sensors are approximately 1.5 mm in diameter and
0.3 mm thick. The new system consisted of an array of these miniature pressure sensors arranged
in various spatial distributions and embedded in electrically resistant durable Kapton (poly 4, 4'-
oxydiphenylene-pyromellitimide) tape. The terminology “sensor array” will be used henceforth
to describe the tape-multi-sensor-tape construct versus the non-taped individual “sensor”
elements. All results are reported for the sensor in the sensor array (encapsulated) configuration.

145

Figure 75: Pressure sensor element geometry (left) and sample sensor array configuration
(right)

4.2.2 Test apparatus
Calibration and evaluation of the sensor and sensor array were performed within a
custom designed pressure vessel shown in Figure 76. A pressure vessel is a commonly utilized
method of sensor evaluation (Rudert et al., 1987, Buis & Convery, 1997, Otto et al., 1998, Otto et
al., 1999, Polliack et al., 2002, Giacomozzi, 2009, Giacomozzi, 2010, Hoffman & Decker, 2005)
because it guarantees a uniform pressure distribution and eliminates potential confounding effects
of variation in sensor contact area, which may be present when attempting to calibrate the sensors
between two surfaces with an externally applied load. The pressure vessel was constructed with a
section of 75 mm diameter, 350 mm long, A312 stainless steel pipe (96 MPa burst pressure,
Biomechanical Research & Testing, Long Beach, CA, USA) with screw-on end caps. Sensor
array wiring pass through was facilitated by a hermetically sealed 30-wire pass through plug
146

(3.45 MPa rated) placed on the left end of the vessel as shown in Figure 76. On the other end was
a high accuracy, PX35 High Accuracy (6.9 MPa, accuracy of 0.25% combined linearity,
hysteresis, and repeatability), NIST traceable, pressure transducer (Omega Engineering,
Stamford, CT, USA). Figure 76 shows the calibration system prior to sealing the left end of the
cylindrical pressure vessel. For each sensor configuration, the array was wired to the internal
portion of the wire pass through (right side of the left end cap in Figure 76), the array and wires
were placed in the vessels, and the left end cap was screwed on to the pressure vessel. A hermetic
seal was established by sealing all threads with PTFE (Polytetrafluoroethylene) thread sealing
tape.
All data were collected on a 16-bit TDAS Pro (Diversified Technical Systems, Seal
Beach, CA, USA) data acquisition system certified to the NHTSA, FAA, ISO 6487 and SAE
J211 data acquisition practices. As the underlying sensing element was a strain gauge, each
channel was configured in a half bridge configuration with 5 volts excitation. Data were
collected at 250 Hz per channel.

147

Figure 76: Calibration pressure vessel set-up (prior to sealing).

After embedding the sensors in the Kapton tape, the new sensor array was tested in the
pressure vessel shown in Figure 76. The new sensor array’s signals were compared to those of
the high accuracy pressure transducer. Initial 1 Hz testing of the array found that embedding the
sensors in the Kapton tape did change the factory calibration values by approximately 6.48%+/-
2.62% (mean +/- SD) (N=58). Therefore, all factory sensor calibration values were adjusted
based on the 1 Hz calibration data so that the embedded sensor’s output matched the high
accuracy pressure transducer’s output. All reported data from subsequent testing utilized the
corrected 1 Hz calibration values.

Hermetically Sealed Wire
Pass and Left End Cap
Data Acquisition System
Air Pressure
Inlet Valve
Reference Pressure
Transducer
New Sensor Array
Air Pressure
Release Valve
Pressure
vessel
Thread Sealing
Tape
148

4.2.3 Specifications
4.2.3.1 Thickness – Total sensor array less than or equal to 0.8 mm thick
Since many articular joint spaces are in fact only virtual spaces with no true empty space
between adjacent mating surfaces, it is desirable to have the thinnest possible sensor to minimize
the effect of the sensor itself. It has been shown (Hedman, 1992) that in the lumbar facet joint, as
the thickness of the sensing element increases beyond 0.8 mm, there is an exponential increase in
the sensor array induced pre-load in the facet joint. Other articular joints such as the human tibia-
femoral (knee) or talocrural joint (ankle) exhibit much greater space between the articular
surfaces (Howell et al., 2010, Dam et al., 2007, Nuno & Ahmed, 2003, Demirci et al., 2008). The
more rigorous requirement of the lumbar facet joint will be utilized in order to maximize the
utility of the new sensor array because meeting the thickness requirements for the lumbar facet
joints would allow the sensor array to be placed in more spacious joints.
4.2.3.2 Scalable geometry - Down to less than 12.0 mm in width and 15.1 mm in
height.
A sensor designed to capture human lumbar facet loading must fit within the facet width
and facet height (Panjabi et al., 1993). A sensor designed to fit within the L2-3 through L5-S1
facet joints must maintain active sensing elements within a width (FCW) of approximately 12.0
mm and a height (FCH) of 15.1 mm. The width was calculated by taking the smallest of the
average of the left and right widths at the L2-3 articulation (the L2 inferior facet) minus one
standard deviation. The height was calculated with the same method applied to the respective
heights (L3 superior facet) (Panjabi et al., 1993). Height and width of the new sensor array were
measured with a micrometer.
149

4.2.3.3 Flexibility in sensor spacing - Spatial resolution including individual element
location, number of sensor and effects of sensor spacing on adjacent sensors
(cross-talk).
Sensor spacing was evaluated by considering the physical size of the sensor element and
examining the densely packed arrangement of the sensors touching in a side-by-side
configuration as show in Figure 77. As these individual sensor elements were being placed
between two pieces of Kapton tape, there was the possibility that applying force to one sensor
would induce artifactual readings in an adjacent sensor. To examine whether this was occurring,
two sensors were encapsulated in the tape immediately adjacent to one another as shown in
Figure 77. Force was applied to each sensor via a rubber dowel, making sure no portion of the
dowel touched the adjacent sensor while simultaneously recording readings from both sensors.

Figure 77: Position of two sensor elements for proximity affects testing (small hash marks
are millimeters)

Similar testing was also conducted with the five element sensor array used in the
curvature tests which had the sensors separated by 0.5 mm as shown in Figure 78.
150

4.2.3.4 Curvature affects - Less than 1% zero load offset and less than 1% change
in sensitivity on curved surfaces down to a 9 mm radius sensor array
curvature.
Curving a sensor can induce sensor offset and sensor error, as well as change the sensor’s
sensitivity (Ferguson-Pell et al., 2000). A sensor array was constructed as shown on the left side
of Figure 78 with five sensor elements in a 10 mm wide row with the sensor elements spaced
approximately 0.5 mm apart. This sensor array was then adhered with double stick tape to the
interior of cylindrical metal collars at various radii as shown on the right side of Figure 78. The
sensors were oriented with the sensitive side (flat) portion facing out as they were subsequently
used in the cadaver spines. Effective collar radii were 5.5, 7, 8, and 9 mm. The sensor array was
tested with each radius collar in the no load state (0 MPa) and in the pressure vessel at 0.96 MPa.
Finally, the limits of bending effects were examined by bending the unloaded sensor array to 90
degrees.

Figure 78: Sensor array for curvature tests (left) placed in cylindrical collars (right) of
various radii

151

The smallest radius of curvature for the lumbar facets was found to be a mean radius of
curvature of 8.7 mm at the L2 vertebral level and largest at L5 at 16.6 mm (Panjabi et al., 1993,
Tulsi & Hermanis, 1993). These values represented the mean calculated values minus one
standard deviation. Testing was conducted at the smaller radius as this was the more rigorous
requirement and most likely to induce error.
4.2.3.5 Frequency response – Less than 2% loss of signal power (less than 0.09 dB)
at frequencies from steady state (0 Hz) up to 5 Hz.
It was desired to have a sensor array that was at least capable of capturing data up to 5
Hz. Air was cycled in and out of the pressure vessel shown in Figure 76 at a rate of
approximately 5.5 Hz. The power spectral density (PSD) was utilized to quantify frequency
response of the sensor array relative to the reference sensor. All sensor data and reference
transducer data were extracted over the same time window. The mean value (DC response) was
subtracted from all signals and the PSD was calculated. Results are reported as both a percentage
difference in power of the sensor array relative to the reference sensor and as the relative decibel
(dB) loss.
4.2.3.6 Non-linearity – Less than 1% of full scale (FS).
Linearity of the sensor array is an expression of the extent to which the actual measured
response departs from the idealized best-fit straight line response at constant temperature
(Ferguson-Pell et al., 2000, Fraden, 2003). The linearity was defined as the maximum deviation
from a 3-point straight line drawn from the 0 psi, half maximum pressure (0.48 MPa) and the
maximum measured pressure of 0.96 MPa. Results are stated as a percentage of the of the normal
full scale (FS) output.
4.2.3.7 Drift – Less than 1% of full scale out to at least 700 seconds.
Drift of the sensor array is a measurement of the long term change in sensor measurement
typically relative to a held non-zero reference sensor value (Ferguson-Pell et al., 2000, Fraden,
152

2003, ISO/IEC, 2007). Various arrangements of sensor elements were placed within the pressure
vessel shown in Figure 76. The pressure was then raised to a nominal value of either 0.96 MPa or
1.38 MPa and held for between 300 to 700 seconds. Drift was determined by taking the
difference from the references pressure transducer and subtracting the sensor value at readings
taken at 100, 250 and 700 seconds, respectively. Drift is reported as the actual pressure drift and
as a percent of the full scale sensor range.
4.2.3.8 Hysteresis – Less than 1%.
A hysteresis error is a deviation of the sensor’s output at a specified point of the input
signal when it is approached from the opposite direction (Ferguson-Pell et al., 2000, Fraden,
2003). Hysteresis is the lagging or a retardation of the effect when the measured quantity (force
in this case) acting upon a body is changed. For this sensor array, hysteresis is the difference in
sensor output when comparing readings taken during increasing load to those during decreasing
load at the same loading rate. Readings relative to the reference transducer were compared for
increasing and decreasing pressures at a load and unload rate of approximately 0.689 MPa/sec
(100 psi/sec) with data sampled at 250 Hz.
4.2.3.9 Repeatability – Coefficient of variation less than 2%.
Repeatability (reproducibility) is the closeness of agreement between measured quantity
values obtained by replicate measurements under similar specified conditions (Ferguson-Pell et
al., 2000, Fraden, 2003, ISO/IEC, 2007). This is also considered a measurement of sensor
precision (ISO/IEC, 2007), but notably not a measurement of accuracy. It was tested by
recording data at pressures of 0.250, 0.500, 0.750 and 1.000 MPa. Pressure was applied at
approximately 0.100 MPa per second with no more than 10 seconds between tests. This was
repeated 36 times at each pressure level for a total of 144 measurements. Repeatability was
153

quantified by the coefficient of variation, which expresses the variation as a percentage of the
mean and is calculated as shown in the following equation.
%

100 %
Coefficient of variation was considered “excellent” when below 2.0%, “very good” between
2.0% and 3.0% , “good” between 3.0% and 4% and “fair” between 4.0% and 5.0% (ACI 214R-
02, 2002).
4.2.3.10 Total cost of utilization including hardware, software, and sensors.
Total cost of utilization was examined by considering the cost of sensor elements,
software, hardware and the cost of sensor replacements.
4.3 Results
4.3.1 Thickness – Total sensor array less than or equal to 0.8 mm thick.
As shown in Figure 75, an unencapsulated single sensor element has a thickness of
approximately 0.3 mm. The thickness of the combined array of Kapton tape-sensor-Kapton tape
was measured with a micrometer. The sensor array was approximately 0.6 mm thick or
approximately 25% less than the design criteria for the lumbar facets. Using a 0.6 mm sensor
array thickness induced facet loading of 20 N or less (Hedman, 1992).
4.3.2 Scalable geometry - Down to less than 12.0 mm in width and 15.1 mm in height.
Individual sensor elements are approximately 1.5 mm in diameter. The sensor elements
can be arranged immediately adjacent to one another with the only limitation being spacing
required for the wiring. Figure 75 demonstrates an array of 7 sensors within the specified 12 mm
by 15 mm spatial limitations. Arrays for smaller joints could easily be accommodated by
decreasing the number of sensors or more closely positioning them.
154

4.3.3 Flexibility in sensor spacing - Spatial resolution including individual element
location, number of sensors and effects of sensor spacing on adjacent sensors (cross-
talk).
There are no inherent limitations on the number of sensor elements that can be used. The
individual sensors can be arranged touching or side by side with the only limitation being
sufficient space for the 0.127 mm diameter wires. Since each sensor in any given array is its own
uniquely wired sensor, arrays can be easily rearranged or optimized and individual sensor
elements can be replaced within the array as needed.
With the two sensor elements placed within the tape immediately adjacent to one another
as shown in Figure 77 and focal contact load applied, Table 27 shows that at most there was a
1.64% contribution from the adjacent sensor. When the sensor elements were placed 0.5 mm
apart as shown in Figure 78 (left), there was effectively no contribution (average change 0.03%)
from loading on adjacent elements.
Table 27: Proximity results

Loaded Sensor
Pressure (MPa)
Adjacent Sensor 0
mm spacing
Pressure (MPa)
Adjacent Sensor
0.5 mm spacing
Pressure (MPa)
% Change at
0 mm spacing
(%)
% Change at
0.5 mm spacing
(%)
0.5 0.0057 -0.0002 1.139 -0.036
1 0.0107 0.0003 1.065 0.026
1.5 0.0110 0.0006 0.732 0.041
2 0.0193 0.0000 0.964 0.000
2.5 0.0257 0.0000 1.028 0.000
2.7 0.0341 0.0004 1.263 0.017
3 0.0404 0.0015 1.346 0.048
3.45 0.0567 0.0048 1.643 0.140

155

4.3.4 Curvature affects - Less than 1% zero load offset and less than 1% change in
sensitivity on curved surfaces down to a 9 mm radius sensor array curvature.
Sensor array curvatures with radii of 5.5, 7, 8, and 9 mm were examined in the no load
condition and under 0.96 MPa of pressure using the sensor array in Figure 78. Performance was
compared to baseline values obtained with the sensor array laying flat.
As shown in Table 28, under the no load, or 0 MPa, condition, the pressure across all
sensors remained at approximately 0 MPa (within the 0 load noise band) for all four radii tested.
Even at the smallest radius of 5.5 mm, no appreciable difference was observed in any sensor. To
examine the limitations of sensor array bending at the no load condition, one additional test was
conducted with the sensor array bent between two sensor elements at a 90 degree angle to the
adjacent sensor thus producing a theoretically infinitely small radius. At this extreme bending,
there was an approximately 0.048 MPa reading at each of the two sensors immediately adjacent
to the bend. Under 0.965 MPa, there were no difference in the pressure readings and no changes
in the sensor sensitivity for the 7, 8 and 9 mm radii. When curved around the inside of a 5.5
radius cylinder, the sensors exhibit a 0.66%+/-0.97% (N=25) change in sensitivity.

Table 28: Affects of sensor array curvature

Radius of
curvature (mm)
No load output
(MPa)
0.965 MPa Output
(MPa)
0.965 MPa
sensitivity change
(%+/- SD)
5.5 0 0.958 0.66 (0.97)
7 0 0.963 0
8 0 0.961 0
9 0 0.964 0
~0 0.048 n/a n/a

156

4.3.5 Frequency response – Less than 2% loss of signal power (less than 0.09 dB) at
frequencies from steady state (0 Hz) up to 5 Hz.
Higher frequency pressure readings were obtained by rapidly cycling the inlet and outlet
pressure valves on the pressure vessel in Figure 76. PSD results of the cyclic data are shown in
Figure 79. The new sensor array is capable of capturing the majority of the signal power with
less than 1.5% difference (less than a 0.07 dB difference) in power up to approximately 5.5 Hz.

Figure 79: A typical PSD results for the reference transducer and the new sensor array

4.3.6 Non-linearity – Less than 1% of full scale (FS).
As reported by the sensor manufacturer, individual sensor element non-linearity using a
three point method at 0, 1.72 and 3.45 MPa was 0.456+/-0.369 FS (N=58). The new sensor array
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0123456789 10
Power (kPa^2/Hz)
Frequency (Hz)
Reference Pressure Sensor
New Sensor 5770
New Sensor 5773
157

non-linearity was tested and found to be 0.58%+/-0.13% FS (N=58). The linearity is observable
in Figure 80.
4.3.7 Drift – Less than 1% of full scale out to at least 700 seconds.
Table 29 shows the results of drift tests with the sensor array held at nominal pressure
values of 0.96 MPa and 1.38 MPa.
Table 29: Sensor array drift response

Time (sec)
Ave Diff (S.D.)
@ 0.96 MPa
(MPa)
% Full
Scale
Ave Diff (S.D.)
@ 1.38 MPa
(MPa)
% Full
Scale
100
1
-0.0019 (0.0048) -0.054 0.00412 (0.0034) 0.121
250
1
-0.0056 (0.0062) -0.162 0.0045 (0.0042) 0.130
700
2
-0.0008 (0.0122) -0.024 -0.0180 (0.0115) -0.522
1: N=24, 2: N=12

Drift at the highest nominal hold pressure of 1.38 MPa was less than 0.2% FS out to 250
seconds and less than 0.6 % F.S. at 700 seconds.
4.3.8 Hysteresis – Less than 1%.
Individual sensor array hysteresis was reported by the manufacturer as 0.631% +/-
0.358% FS (N=58). The new sensor array hysteresis was examined by taking the sensor values at
a specific reference pressure during decreasing load and subtracting the sensor value at the same
reference pressure during increasing load (Morris, 2001, Fraden, 2003). This difference was
reported as a percent of the FS pressure. Whole sensor array hysteresis was 0.78% +/-0.18% FS
(N=42). A typical sensor array channel pressure versus reference pressure graph is show in
Figure 80.

158

Figure 80: Typical sensor hysteresis curve

4.3.9 Repeatability – Coefficient of variation less than 2%.
Table 30 shows the results of the repeatability test for each of the 36 measurements.
Across all four reference pressure values the coefficient of variation remained less than 2% and
less than 1% for pressures of 0.500 MPa or greater.
Table 30: Repeatability Testing

Reference Value
(MPa)
Average Measured Value
(MPa)*
Standard Deviation
(MPa)
Coefficient of Variation
(%)
0.250 0.250 0.005 1.934
0.500 0.504 0.005 0.932
0.750 0.763 0.005 0.604
1.000 1.025 0.004 0.386
*N=36
159

4.3.10 Total cost of utilization including hardware, software, and sensors
Total system cost was relatively low as standard commercially available data acquisition
systems could be utilized (the underlying sensing element of the individual sensor is a strain
gauge), with no specialized viewing software required. Additionally, individual sensor elements
within an array could be replaced as needed without necessitating replacement of the entire array.
In small quantities, individual sensor elements cost approximately $220 (US$) per sensor.
4.4 Discussion
The objective was to develop and thoroughly evaluate with laboratory bench top testing a
new sensor array. It is anticipated that compliance with the specified evaluation will allow the
new sensor array to accurately quantify the intra-articular pressures within curved physiologic
joints, such as the lumbar facet joints. Existing technology exhibited higher errors when the
arrays were curved at physiologically relevant radii (Caldwell et al., 1993, Ferguson-Pell et al.,
2000, Buis & Convery, 1997, Polliack et al., 2002, Martinelli et al., 2006, Wilson et al., 2006).
With the new sensor array, in addition to maintaining good overall sensor array performance,
particular attention was given to eliminating sensor array curvature artifacts at physiologically
relevant radii.
The most commonly utilized systems similar to the new sensor array are FSR-T’s (force
sensitive resistor-Tekscan). The FSR-T has been shown to exhibit negligible errors at radii down
to 32.5 mm with a substantial jump in errors to approximately 225% at 16.1 mm, 325% at 13.6
mm, 700% at 10.4 mm and 750% at an 8 mm radius (Ferguson-Pell et al., 2000). Additionally,
curving the FSR-T added a substantial offset (non zero reading at zero pressure) and lowered the
FSR-T's sensitivity by nearly 200% at the two smallest radii (Ferguson-Pell et al., 2000). Lumbar
facet joints exhibit curvatures of between 8.7 mm and 16.6 mm (Panjabi et al., 1993, Tulsi &
Hermanis, 1993). This curvature error may explain why prior testing of the FSR-T in a lumbar
160

facet joint exhibited errors as high as 50% (Wilson et al., 2006). The less used, but likely more
accurate FSC-N, had very little performance data for curvature in the literature, but it was shown
to have an error of 31%+/-21% at a 17.1 mm radius and a mean accuracy error of 9.96%+/-9.10%
for radii between 29.1 and 99.9 mm (Polliack et al., 2002). Both of these commercially available
systems will likely exhibit substantial error even in much larger joints such as the knee joint
(femoral-tibia) and ankle joint (talocrural) which have been shown to have curvatures between
17.8 mm and 25 mm (Howell et al., 2010, Dam et al., 2007, Nuno & Ahmed, 2003) and 16 mm
and 28 mm (Demirci et al., 2008), respectively. The new sensor array’s performance was
unchanged down to a radius of 7 mm, which was approximately 2 mm below the design
requirements. This superior performance may be due to the composite nature of the sensor arrays,
with the space between the sensors providing the bulk of the curvature and the sensors themselves
remaining relatively flat. This polygonal curvature behavior could be investigated further using a
shadow profilometer in a future study.
Enclosing the individual sensors in Kapton tape did degrade the sensor performance by
increasing average non-linearity from 0.456% FS to 0.58% FS and hysteresis from 0.631% FS to
0.78% FS. Considering the Kapton tape is a relatively plastic-like polyimide film, sticking two
layers of this on the sensor (front and back) would be expected to degrade the linearity and
hysteresis. However the degradation was minimal with resulting values remaining well below the
stated goals.
The new sensor array system also satisfied all other the design requirements with sensor
array thickness well below 0.8 mm, scalability of sensor array size, density and orientation,
frequency response beyond 5 Hz, good linearity, minimal drift, acceptable hysteresis, excellent
repeatability, and relatively low cost. Additionally, the componentized nature of the current
sensor array system allows for a very versatile and near limitless selection of sensor array sizes
161

and distribution. This flexibility also facilitates the utilization of the sensor elements as possible
sources of tactile-like feed back in prosthetic designs.
This new sensor array offers the potential to measure dynamically, with good spatial
resolution, pressure on highly curved surfaces and should be extremely useful in understanding
the overall intra-articular mechanics. While the new sensors and sensor arrays were rigorously
tested in a laboratory bench top setting, the sensor arrays were designed for use in human lumbar
facet joints which are small and have a tight radius of curvature. It is reasonable to conclude that
they will perform equally well in geometrically less demanding joints and surface interfaces such
as the knee and ankle (Howell et al., 2010, Dam et al., 2007, Nuno & Ahmed, 2003, Demirci et
al., 2008), however future testing is needed to determine the utilization of this sensor array in
such conditions.
Preliminary testing has confirmed the ability of the described sensors to measure in vitro
human lumbar spine facet contact pressures. Additional future studies will comprehensively
evaluate the in vitro performance in human lumbar facet joints.

162

CHAPTER 5: PRELIMINARY IN-VITRO RESULTS

5.1 Introduction
The objective was to develop a new sensor array capable of more accurately quantifying
the spatial and temporal distribution of intra-articular pressures within the lumbar spine facet
joints. This chapter presents the evaluation of a new sensor array for directly measuring the
dynamic time-dependent and spatial distribution of intra-articular pressure within cadaveric
lumbar spine L4-5 facet joints. Specification for interface pressure measurements sensors have
been previously proposed (Ferguson-Pell et al., 2000, Wilson et al., 2006, Giacomozzi, 2009,
Giacomozzi, 2010, Martinelli, 2006, Welcher et al., 2011). Prior bench top testing examined
criteria including thickness, height and width, sensor density, effects of sensor spacing on
adjacent sensors (cross-talk), curvature effects, frequency response, linearity, drift, hysteresis, and
repeatability (Welcher et al., 2011). Finally, total cost of utilization including hardware,
software, sensors and durability were qualitatively evaluated.
5.2 Methods
5.2.1 Sensor construction
The individual sensor elements are composed of two beryllium-copper surfaces with a
strain gauge between them. The individual sensor elements used for this study have a
manufacturer stated linear operating range from 0-3.4 MPa (Model 060, Precision Measurement,
Ann Arbor, MI, USA). The sensor element’s three lead wires are Teflon coated #36 (0.127 mm
diameter) stranded copper wire. As shown in Figure 81, the individual sensors are approximately
1.5 mm in diameter and 0.3 mm thick. The new sensor array consisted of an array of seven of
these miniature pressure sensors arranged as shown on the left side of Figure 81. The sensors are
163

embedded between two pieces of electrically resistant and durable Kapton (poly 4, 4'-
oxydiphenylene-pyromellitimide) tape.

Figure 81: Pressure sensor element geometry (left) and sample seven sensor array
configuration (right)

5.2.2 Facet geometry and sensor geometry constraints
Thickness
Since many articular joint spaces are in fact only virtual spaces with no true empty space
between adjacent mating surfaces, it is desirable to have the thinnest possible sensor to minimize
the effect of the sensor itself. It has been shown (Hedman, 1992) that in the lumbar facet joint, as
the thickness of the sensing element increases beyond 0.8 mm, there a significant exponential
increase in the sensor array induced pre-load in the facet joint. The thickness of the combined
array of Kapton tape-sensor-Kapton tape was measured with a micrometer (±0.001).

164

Height and width
A sensor designed to capture facet loading must fit within the facet width (FCW) and
facet height (FCH). It has been demonstrated (Panjabi et al., 1993) that both facet width and
height generally get larger proceeding caudally down the spine. Throughout the whole spine, on
average, the superior facets of the inferior vertebrae are both wider (average of 0.425 mm larger
from L1 to L5) and taller (average of 0.275 mm taller from L1 to L5) than the mating inferior
facets of the superior vertebrae (Panjabi et al., 1993). To ensure coverage of the maximum
amount of articular surface, the geometry of the larger superior facets was used. As shown in
Table 31, a sensor designed to fit within the L1-L2 facet joints must maintain the active sensing
elements within a width (FCW) of approximately 10.5 mm and a height (FCH) of 14.0 mm. The
widths and heights were calculated using the smallest average (superior L2 facet for the L1-L2
articulation) of the left and right superior facet values minus one pooled standard error of the
mean and rounded to the nearest ½ mm.
An array of seven sensors was used as shown in Figure 82. A coordinate system was
established with the 0 mm, 0 mm sensor position at the anatomical lateral and superior-inferior
(SI) center of the facet surface. Negative X-values are medial of the facet surface vertical midline
and negative Y-values are inferior to the facet surface horizontal midline. Horizontally, five
sensors elements were placed with approximately 0.5 mm between the edges of each adjacent
sensor. Sensors were placed at the anatomical center 0 mm, 0 mm and laterally along the
Superior-Inferior (S-I) midline at +2, +4, -2, and -4 mm from the center, respectively.
Approximately 1.25 mm of tape was left extending beyond the most medial sensor (-4 mm, 0
mm) resulting in the total sensor array extending approximately 6 mm medially from the center 0
mm, 0 mm sensor. Two additional sensors were placed along the lateral mid line with their
centers at 0 mm, +5 mm and 0 mm, -5 mm. Approximately 1.75 mm of tape was left vertically
165

extending beyond each sensor giving the array a total height of 15 mm. The sensor array
geometry was then approximated as an ellipse to match the general facet shape (Panjabi et al.,
1993, Tischer et al., 2006), centered at the origin (0 mm, 0 mm), a minor horizontal axis of 12
mm, and major vertical axis of 15 mm. The array was trimmed to the ellipse dimension on the
medial portion. Most of the lateral portion did not need to be trimmed down as it protruded out of
the cut facet capsule. A limitation of this study is that the height and width of the sensor array is
based upon calculated means of many facets and if the facet under study is wider, there will be a
medial shift of the sensor array center relative to the anatomical center. The converse would be
true for a narrower facet.

Figure 82: Sensor positions in the array (to scale)

166

Curvature
As noted in the introduction, contemporary technologies for intra- articular facet force
measurement such as FSR-T and FSC-N have been shown to have errors from -11.27% to 50%
when used within lumbar facet joints (Martinelli et al., 2006, Giacomozzi, 2009). It was therefore
deemed important to try and determine some descriptive measure of facet curvature and assure
the sensor array could accurately measure pressure at those radii.
An approximation of facet curvature was obtained by combining two series of published
data on facet geometry (Panjabi et al., 1993, Tulsi & Hermanis, 1993). Tulsi and Hermanis
(1993) used a metal rod to determine the zygapophyseal facet depth “K” relative to a horizontal
line across the top of the facet as shown in Figure 83.

Figure 83: Typical L5 Vertebrae with Metal Measurement Rod “K” in Solid Black (Tulsi &
Hermanis, 1993)

167

Approximating the facet as a semi circle, Tulsi’s “K” measurement is very similar to the
chord height “h” shown in Figure 84.

Figure 84: Basic Circle Descriptors (r=radius, c=chord length, h=chord height, s=arc
length)

The effective radius of the shaded region in Figure 84 can be calculated with Equation 1.
Equation 1

Combining Equation 1with the published data’s nomenclature (Panjabi et al., 1993, Tulsi
& Hermanis, 1993), Equation 1can be rewritten as:
Equation 2

Inputting the data from the two respective studies into Equation 2 yields the tightest
idealized radius of approximately 8.66 mm as shown in Table 31.

 
2 8 8
4
2 2 2
h
h
c
h
h c
r  


2 8
) (
2
K
K
FCW
r  
168

Table 31: Lumbar facet height, width and curvature data

5.2.3 Preliminary in vitro lumbar spine facet testing
Two frozen human lumbosacral (L1-Sacrum) specimens were utilized as described in
Table 32. Each specimen was allowed to thaw at room temperature and was meticulously
dissected and cleaned of any surrounding paraspinal musculature, spinal nerves, and spinal cord.
Special attention was directed to preserving all osteoligamentous portions of the lumbosacral
spine. Radiographs of each specimen were examined along with donor history to insure that no
tumors, previous fractures, or metabolic diseases affecting the tissues were present. Each
specimen was anchored in a two-part polyurethane potting solution (BJB Enterprises, Inc., Tustin,
CA). The most superior and inferior vertebral bodies were submerged in the potting solution to
the approximate level of the horizontal vertebrae midline, with the spine oriented so that the L4
vertebral body superior endplate was horizontal similar to a normal erect posture. Hydration of
the specimen was maintained prior to testing by wrapping it with gauze soaked in isotonic saline
and during testing by routinely spraying it with isotonic saline solution.

Tulsi (n=112, M=64, F=48)
1
Panjabi (n= 24 facets/level)
2
Vertebrae Level
Mean Depth
Superior Facet
Male (K)
Mean Depth
Superior Facet
Female (K)
Superior Facet Width
(FCW ave L and R)
Superior Facet
Height (FCH ave L
and R)
Mean Radius
Male
Mean
Radius
Female
mm (SE) mm (SE) mm (SE) mm (SE) mm mm
L1 1.360 (0.007) 1.167 (0.008) 10.35 (0.42) 12.45 (0.38) 10.53 12.06
L2 2.075 (0.005) 1.881 (0.006) 11.25 (0.61) 14.6 (0.46) 8.66 9.35
L3 2.091 (0.005) 1.897 (0.005) 13.85 (0.57) 15.95 (0.56) 12.51 13.59
L4 2.245 (0.005) 2.052 (0.005) 14.7 (0.58) 16.7 (0.43) 13.15 14.19
L5 1.945 (0.005) 1.752 (0.006) 15.6 (0.46) 17.45 (0.49) 16.61 18.24
1= Tulsi & Hermanis, 1993
2 = Panjabi et al., 1993
169

Table 32: Description of spine segments utilized

The specimens were mounted into a Bose Kinematic Spine Testing Device (Bose
Corporation, Minnetonka, MN) for mechanical testing as shown in Figure 86. This device
provided six degrees-of-freedom motion through a combination of five rotational and three
translational actuators (pneumatic or servo-electronic) and was instrumented with two load cells
and local rotational torque cells for each gimbal, which together allowed for essentially pure
moment bending.
The mechanical testing protocol consisted of three cycles of dynamic pure moment
flexion/extension loading (±10 Nm). The rate of loading was 1°/second until the load limit of
±10 Nm was reached (Cholewicki et al., 1996, Goel et al., 2006, Panjabi et al., 1989, Yamamoto
et al., 1989).
Each vertebral body was instrumented with an array of four infrared light emitting diodes
(LED’s). These were used to calculate three-dimensional linear and rotational responses using an
optoelectronic motion capture system (Optotrak 3020, Northern Digital Inc., Waterloo, Ontario,
Canada). The Optotrak 3020 is an active infrared optical position tracking system wherein high
resolution sensors detect the 3 dimensional position of a light emitting diode (IRED) through
triangulation. The IRED's are turned on and off in sequence by a strobe unit permitting the
system control unit (SCU) to identify which IRED is active at any point in time. As a result, it is
not necessary to digitize or manually identify markers nor does the system confuse or interchange
markers as they cross one another. Maximum sampling rates is 3500 Hz. When 3 or more
Spine Age Gender Segments
167
F L1-Sacrum
273 F
L1-Sacrum
170

position markers are oriented on a rigid body (.rig files), the additional 3 rotational DOF can be
determined. The manufacturer reports RMS accuracy of 0.1 mm for X, Y coordinates and 0.15
mm for Z coordinate (towards the camera). Research by States and Pappas (2006) indicated an
average within-trial standard deviation of 0.125 mm for linear and 0.0196 for angular dimensions.
All of the Optotrak target/marker design, construction, system programming, calibration
and validation for the present project and for every other project in the Cedar Sinai spine lab were
completed by this researcher. The local coordinate system origin was selected to be the lower left
corner diode of the 4 diodes that define the fixed rigid body of the Enduratek/Bose Kinematic
Spine Simulator. The 6-DOF motion of the other three rigid bodies is reported relative to the
local machine fixed coordinate system. This is shown in Figure 85. The local right-hand
coordinate system has X positive to the right, Y positive up, and Z-positive out.

171

Figure 85: Optotrak coordinate system and target location

The L4-5 bilateral facet joint capsules were incised approximately 180 degrees along the
posterior edge from the superior apex to the inferior apex to facilitate sensor placement. The new
sensor array was placed typically in the right facet and a dummy array with the same sensor
configurations was placed in the contralateral facet as seen at the posterior of the spine in Figure
86. Both facets were cut to maintain symmetry and avoid any asymmetries in the response
associated with asymmetries in joint capsule integrity.
Optotrak Rigid
Bodies
Origin
+X
+Y
+Z
172

The mechanical testing device, motion analysis system and pressure sensor system were
synchronized off a common time zero signal. Data were collected on the mechanical testing
device and motion analysis system at 50 Hz, and at 250 Hz on the pressure sensor system. Data
collected on the second loading cycle of each loading condition were utilized for data analysis.
Power spectral density analysis of the unfiltered 250 Hz pressure sensor data revealed that
99.935% of the power was below 5 Hz and 99.3% of the power was below 0.5 Hz. To preserve
as much signal power as possible while removing what is likely spurious noise, all data were
filtered with a four pole Butterworth phaseless low pass filter set at 5 Hz.

Figure 86: Anterior view of lumbosacral specimen setup in Bose Kinematic Spine
Simulator
173

5.2.4 Total cost of utilization - hardware, software, sensors and in vitro durability.
Total cost of utilization was examined by considering the cost of sensor elements,
software, hardware and the cost of sensor replacements. A component of cost is durability of the
sensor, which is qualitatively assessed.
5.3 Results
5.3.1 Facet Geometry and Sensor Geometry Constraints
The new sensor array was approximately 0.6 mm in thickness, as measured with a
micrometer, and was scalable to below the nominal 14 mm wide by 16 mm high lumbar spine
facet joint size. A sensor of oval shape 12 mm wide by 15 mm high (Figure 82) was constructed
to insure maximum versatility and fit in the L4-5 facet joint. Pressure vessel testing (Welcher et
al., 2011) found no difference in sensor performance down to radius of curvature of 7 mm and
only a 0.66%+/-0.97% change in sensor sensitivity were observed at a radius of 5.5 mm.
5.3.2 Preliminary In vitro lumbar spine facet testing
Figure 87 shows the typical angular flexion-extension motion of both the entire specimen
and locally of L4 relative to L5 in response to the applied +/- 10 Nm bending moment. Flexion
angles and moments are reported as positive. At maximum flexion the L4-5 segment was flexed
approximately 4.5 degrees and extended approximately -3 degrees at maximum extension.

174

Figure 87 Specimen moment loading and response

As seen in Figure 88, the L4-5 facet joint had decreasing (unloading) facet pressure
relative to the facet pressure at the neutral position when flexed from neutral to full flexion. This
data was measured from sensor position #1 (0 mm, -5 mm) as shown in Figure 91. The mean
pressure of all seven sensors for both spines at the neutral position (0 degrees rotation and 0 Nm
applied moment) was 300.4 kPa and 40.7 kPa at full flexion, corresponding to a 10 Nm applied
flexion moment. This is consistent with other published studies that noted the greatest facet loads
were measured in extension (Prasad et al., 1974, Adams & Hutton, 1980, Yang & King, 1984,
Dunlop et al., 1984, Abumi et al., 1990, Schendel et al., 1993, Sharma et al., 1995, Yamashita et
al., 1996). Therefore, only facet pressure readings for all seven sensors for the two tests at 50%
extension and 100% extension are presented in Figure 89 and Figure 90, respectively.
175

Figure 88: Sensor at 0 mm,-5mm location response to all cycles

Figure 89: Facet pressures at 50% extension
176

Figure 90: Facet pressures at 100% (10 Nm) extension spines 1 and 2

5.3.4 Total cost of utilization - hardware, software, sensors and in vitro durability.
Each individual sensors cost approximately $220 (US$). As the underlying sensing
element of the individual sensor is a strain gauge, standard commercially available data
acquisition systems could be utilized with no specialized viewing software required. Initial in
vitro testing of the sensor array elements found the durability to be extremely poor, but
manufacturing modification and appropriate sensor orientation improved durability. Initial sensor
failure rates started at approximately one sensing element failing per one or two tests, but were
improved to approximately one sensing element failing per eight tests.
5.4 Discussion
To this author’s knowledge, this is the first paper to present data of direct simultaneous
measurement of temporal and spatial lumbar spine facet pressures to pure moment flexion-
extension loading. “A new measuring device is useful if it is able to provide more reliable, more
177

complete or more accurate information than the existing techniques” (Hedman, 1992). It is
believed that this proposed new sensor array satisfies these requirements.
The objective was to develop a new sensor array capable of more accurately quantifying
both the temporal and spatial distributions of intra-articular pressures within the curved lumbar
facet joint. Other existing technology exhibited higher errors when the arrays were curved at
physiologically relevant radii.
The most commonly utilized systems similar to the new sensor array are the FSR-T. The
FSR-T has been shown to exhibit negligible errors at radii down to 32.5 mm but a substantial
jump in errors to approximately 225% at 16.1 mm, 325% at 13.6 mm, 700% at 10.4 mm and
750% at an 8 mm radius (Polliack et al., 2002). Additionally, curving the sensor array added a
substantial offset (non zero reading at zero pressure) and lowered the sensor’s sensitivity by
nearly 200% at the two smallest radii. As seen in Table 31, lumbar facets joints exhibit
curvatures of between 8.7 mm and 18.2 mm (Panjabi et al., 1993, Tulsi & Hermanis, 1993). This
curvature error may explain why prior testing of the FSR-T in a lumbar facet joint exhibited
errors as high as 50% (Martinelli et al., 2006). The less used, but likely more accurate, FSC-N
had very little performance data for curvature in the literature, but it was shown to have an error
of 31%+/-21% at a 17.1 mm radius and a mean accuracy error of 9.96%+/-9.10% for radius
between 29.1 and 99.9 mm (Polliack et al., 2002). Both of these commercially available systems
will likely exhibit substantial error even in much larger joints such as the knee joint (femoral-
tibia) and ankle joint (talocrural) which have been shown to have radii of between 17.8 mm and
25 mm (Howell et al., 2010, Dam, et al., 2007, Nuno & Ahmed, 2003) and 16 mm and 28 mm
(Demirci et al., 2008), respectively. The current sensor array’s performance was unchanged
down to a radius of 7 mm, which was approximately 1.5 mm below the design requirements.
178

The new sensor array system is approximately 0.6 mm thick, with very broad scalability
of total sensor array size, density and orientation. Frequency response was beyond 5 Hz, with
good linearity, minimal drift, and acceptable hysteresis.
As it is believed that this is the first paper to present data of direct simultaneous
measurement of temporal and spatial lumbar spine facet pressures to pure moment bending, there
are no direct comparisons in the published literature. However, a portion of the data can be
compared to previously published data.
The range of peak facet pressure values of 1,210 kPa for Spine 1 and 3,059 kPa for Spine
2 (Figure 90) are generally in the range of the peak pressures reported using Fuji Prescale
pressure sensitive film “Fujifilm” (Lorenz et al., 1983, Wiseman et al., 2005, Dunlop et al., 1984).
Lorenz et al. (1983) measured peak L4-5 facet pressures (average of left and right facets) of 1,667
kPa with a compressive preload of 196.2 N and extension of 6 to 8 degrees. Other facet force
studies that used Fujifilm demonstrated mean peak pressure of 4,030 kPa ±2,640 kPa with a 15
Nm applied moment and 700 N follower load (Wiseman et al., 2005), and 5,755 kPa (Dunlop et
al., 1984), with a 1000 N compressive load and 200 to 400 N of shear force .
The pressure distributions within the facet joint are consistent with that presented in the
literature (Lorenz et al., 1983, Wiseman et al., 2005 Schendel et al., 1993, Schmidt et al., 2008) in
that the highest measured facet pressures in extension were located in the inferior portion of the
facet.
A near threefold increase in pressure from 812 kPa to 2,411 kPa was seen in the most
inferior sensor in Spine 2. This would be consistent with the frequently described case where the
superior facet bottoms out on the inferior facet (Yang & King, 1984, Dunlop et al., 1984,
Schendel et al., 1993). This would also be consistent with the downward migration of peak
179

pressures with increasing extension moments reported by others (Lorenz et al., 1983, Schmidt et
al., 2008).
As standard commercially available data acquisition and data visualization systems (such
as Excel or Matlab) for strain gauges can be utilized, additional system costs beyond standard lab
equipment are limited to the $220 ($US) per sensor element cost.
The most limiting attribute of the sensor array is durability. The actual sensor, as shown
on the left side of Figure 81, is composed of a solder tab extending to a domed structure.
Examination of multiple failed sensors under a microscope demonstrated failures with the solder
tab cracking or breaking as it entered the domed structure. This failure was immediately evident
by off-scale pressure sensor readings and saturation of the data channel. The second mode of
failure was that the epoxy that covered the wires at the solder tab would crack and either break
the wires at the solder tab or allow the wires to short, also resulting in an obvious pronounced
failure. While conducting initial porcine and then human cadaveric lumbar spine facet testing, a
typical seven sensing element sensor array would last for approximately one to two tests before at
least one sensing element failed and required replacement. Discussions with the manufacturer
resulted in slight changes in the manufacturing process, reducing the amount of epoxy on the
solder tab. Orienting the individual sensing elements so that the long axis of the each sensor
element was as perpendicular as possible to the anticipated loading direction also greatly
improved sensor durability. Durability was extended from one sensor failing every one to two
test to one sensor failing approximately every eight tests. Durability could still be substantially
improved and remains a weakness of this system. However the componentized nature of the
sensor array allows easy replacement of the individual sensor elements within a given array if one
or more sensors fail. This characteristic is beneficial especially when compared to other
commonly available arrays such as those from FSR-T and FSC-N which are constructed by the
180

manufacturer and do not allow either redistribution or replacement of individual sensor elements
within an array.
There are several limitations to this pilot study, notably a small sample size of only two
spines, the facet capsules were cut bilaterally to facilitate sensor insertion, and a sensor of finite
thickness was inserted in what was likely only a virtual space. Future testing will be conducted
with additional spines to examine specimen variability and repeatability, and under different
loading conditions such as pure moments with an axial compressive follower load. Cutting the
facet capsule does change the response however this has been shown to have only “a minor
effect” up to flexion-extension moments of 11.2 N-m (Tencer et al., 1982, Tencer & Mayer,
1983). At a thickness of 0.6 mm, inserting the sensor element will likely induce between 10 and
20 N of joint load (Hedman, 1992).
Because the sensor array satisfies the requirements of being able to measure the time
varying and spatial distribution of pressure within the highly curved lumbar facet joints, it should
be extremely useful in understanding the overall lumbar spine intra-articular mechanics.
Implementing this technology will advance the understanding of general spine biomechanics,
assist in validation of finite element models, and provide data for prosthetic design and testing. It
is reasoned that satisfying the design requirements for the relatively demanding lumbar facet joint
ensures viability of the sensor array in less geometrically demanding joints and surface interfaces
such as the knee and ankle (Howell et al., 2010, Dam, et al., 2007, Nuno & Ahmed, 2003).

181

CHAPTER 6: LUMBAR FACET SPATIAL AND
TEMPORAL MEASUREMENTS UNDER PURE FLEXION-
EXTENSION LOADING

6.1 Introduction
The precise etiology of low back pain lacks a general consensus. However, facet
arthrosis is a common radiographic finding and has been suggested as a common cause of low
back pain since as early as 1911 (Goldthwait, 1911; Badgley, 1941). A study in 1933 (Ghormley,
1933) first published the term “facet syndrome” and suggested that hypertrophic changes
secondary to osteoarthritis of this joint led to nerve root entrapment which caused low back pain.
Mooney and Robertson (1976) were able to elicit pain in normal volunteers with hypertonic
saline injections that then subsided with local anesthetic injection into the joint. However, a
randomized clinical trial (Lilius et al., 1989) with three groups of unilateral low back pain
patients examined injections of cortisone and steroids into the joint and around the joint, and
injection of saline into the joint and found significant improvement in all groups, independent of
the treatment given. Using facet injections, Griffiths and Parantainen (1993) achieved long-term
pain relief in 50% of their population with immediate relief in 90%. A literature review by
Boswell et al. (2007) found that, consistent with criteria established by the International
Association for the Study of Pain (Merskey & Bogduk, 1994), facet joints were found to be a
possible source of chronic pain in 15% to 45% of patients with chronic low back pain; 36% to
60% of the patients with chronic neck pain; and 34% to 48% of the patients with thoracic pain. A
lower prevalence of facet pain was reported in individuals under 65 years of age (30%) compared
with those greater than 65 years of age (52%) (Manchikanti et al., 2001). Recent studies (Deyo et
al., 2005, 2009, Manchikanti et al., 2009) document a 629% increase in Medicare expenditures
for epidural steroid injections; a 423% increase in expenditures for opioids for back pain; a 307%
182

increase in the number of lumbar magnetic resonance images among Medicare beneficiaries; and
a 220% increase in spinal fusion surgery rates from 1990 to 2001 in the United States. As noted
below, some or all of these increases may be attributable to the failure of many spine prosthetics
to adequately assess facet loading or change in loading with intervention.
The facets have been shown to be significant load bearing elements in the spine. Early
testing examining vertical spine loading (Prasad et al, 1974) showed the L3-4 level facets carried
between 35% and 50% of the total spine load when the spine was subjected to 2.5 to 9 G’s of
vertical accelerations. Finite element models with extension testing and a 1000 N preload found
the L2-3 facets carried approximately 30% of the compressive preload when extended to 6
degrees (Shirazi & Drouin, 1987). Externally mounted strain gauges (Schendel et al., 1993)
demonstrated the L1-2 facets carried no load in flexion, 205 N at a 10 Nm extension moment and
190 N axial load, 65 N at a 10 Nm torsion moment and 150 N axial load, and 78 N at a 3 Nm
lateral bending moment and 160 N axial load. The facets have also found to be a substantial
contributor to lumbar shear stiffness with a 77.7% reduction in anterior shear stiffness and a 79%
reduction in posterior shear stiffness with removal of the posterior elements (Lu & Luk, et al.,
2005). The facet and capsule alone were found to contribute 54% to anterior shear stiffness and
67% to posterior shear stiffness (Skipor et al., 1985).
Measurement of facet load and pressures has steadily evolved. Early attempts (Ewing et
al., 1972, Prasad et al., 1974, Prasad & King, 1974, King et al., 1975, Butterman et al., 1991a,
199b) utilized externally mounted strain gauges that inferentially measured the load within the
joint by external measurements. This method was highly sensitive to sensor placement and
density, often required destructive testing to directly correlate strain to joint load, and has been
shown to be subject to artifactual strain coupling from contralateral facet loading (Zhu et al.,
183

2008). Contemporary utilization of this method demonstrated errors in peak facet force values
between -44% and +42% (Zhu et al., 2008).
One of the first studies to directly measure the facet intra-articular load (Lorenz et al.,
1983) utilized Fujifilm (Fujifilm Corp., Tokyo, Japan), which has many beneficial qualities, but
has no temporal and limited spatial resolution. Additionally, the film has many serious
limitations including variations in performance with load rate
(www.fujifilm.com/products/prescale/guide/feature/spec.html, Rudert et al., 1987), temperature
(www.fujifilm.com/products/prescale/guide/feature/spec.html, Liggins & Finlay, 1992), and
humidity (www.fujifilm.com/products/prescale/guide/feature/spec.html, Liggins & Finlay, 1992),
and it requires interpretation of fine color gradations
(www.fujifilm.com/products/prescale/guide/feature/spec.html, Liggins et al., 1995a, Liggins &
Finlay, 1997, Lia et al., 2001, Hale & Brown, 1992, Caldwell et al., 1993). Fujifilm is also prone
to fluid contamination with protective measures possibly altering the factory calibration curves
(Liggins et al., 1994, Liggins et al., 1995a, Hedman, 1992), has a 50 hour development time
(Liggins & Finlay, 1992), and may be subject to “crinkle” artifacts on bicurvillinear surfaces
(Caldwell et al., 1993).
Newer technologies such as force sensitive resistive arrays from Tekscan (FSR-T)
(Boston, MA, USA) show promise. However, these sensors have been shown to have high
hysteresis (Ferguson-Pell & Cardi, 1992, Ferguson-Pell & Cardi, 1993, Pavlovic et al., 1993,
Woodburn & Helliwell, 1996), high creep (Ferguson-Pell & Cardi, 1992, Ferguson-Pell & Cardi,
1993, Werner et al., 1995, Pitei et al., 1996, Otto et al., 1998, Otto et al., 1999), temperature
sensitivity (Werner et al., 1995, Luo et al., 1983) poor dynamic response (Werner, 1995, Pitei et
al., 1996, Otto et al., 1998 Otto et al., 1999, McPoil et al., 1995, Sumiya et al., 1998),
repeatability problems (Woodburn & Helliwell, 1996, Pitei et al., 1996, McPoil et al., 1995),
184

memory artifacts (Pavlovic et al., 1993, Pitei et al., 1996), and up to 300%-700% errors when the
sensing element is curved to radii of physiological relevance (Polliack et al., 2002, Arndt, 2003).
Additionally, proper calibration of the FSR-T requires a priori knowledge of load and material
interface properties (Pavlovic et al., 1993, Luo et al., 1983). Wilson et al.(2006) utilized the grid
based array FSR-T to quantify L3-4 facet loads and found that the I-Scan FSR-T overestimated
the applied compressive force in the lumbar facet joint by 50+/-9%, 35+/-7% and 18+/-9% at the
relatively small compressive load levels of Pavlovic et al., 1993, 50, and 100 N, respectively.
With modification of the calibration routine (Giacomozzi, 2009), during compression testing in
the L3-4 and L4-5 facet joints, the FSR-T was found to overestimate the applied force by19.36+/-
10% at 25N, 14.00+/-9% at 50 N and underestimate the applied force by 3.7+/-10% at 100 N and
11.27+/-13% at 150N.
A similar but capacitive based sensor array from Novel GMBH (Munich, Germany),
force sensitive capacitive-Novel (FSC-N), has been found to have lower hysteresis (Werner,
1995, Giacomozzi, 2009, Giacomozzi, 2010), better linearity (Pavlovic et al., 1993, Giacomozzi,
2009, Giacomozzi, 2010), and less creep (Werner, 1995, Giacomozzi, 2009, Giacomozzi, 2010,
Arndt, 2003, Martinelli et al., 2006) than the FSR-T systems. However, similar problems with
measurements on physiologically relevant curved surfaces persist (Martinelli et al., 2006, Polliack
et al., 2006) with errors of 31%+/- 21% at a 17 mm radius (Polliack et al., 2006). Very little
intra-articular research has been published with the FSC-N and no examples of spinal facet
testing could be located.
Despite the prevalence of facet mediated spinal pain, the facet joints serving as
significant load bearing elements in the spine, and perhaps due to the above stated difficulties in
measuring facet loads, none of the current standards or published testing protocols specify
requirements for measurement or quantification of facet loads (McNally, 2002, Wilke, et al.,
185

1998, Goel et al., 2006, ASTM F2077-03, ASTM F2423-05, ASTM F2346-05, ASTM F2624-07,
ASTM F2790-10). One standard noted that a purpose of an artificial IVD (intervertebral disc
prosthesis) is long-term restoration of function (ASTM F2423-05), yet this standard has no
requirements for quantification of facet intra-articular loading, static or dynamic. A standard
specifically intended to provide guidance for the static and dynamic testing of Lumbar Total
Facet Prostheses (FP) (ASTM F2790-10) likewise does not address measurement of facet loads or
pressures. A set of test protocols for evaluation of spinal implants published by a group of highly
regarded spine researchers (Goel et al., 2006) acknowledged a need for a basic understanding of
among other things, the facet joints, facet arthritis, and that a focus of testing is quantification of
load-sharing between the device and functional spinal unit structures. However, these authors do
not mention measurement of the load or load sharing on the facets. A book on spinal implants
subtitled “Are we evaluating them appropriately?” (Melkerson et al., 2001) contains no data on
facet loads or how to evaluate them and the word facet is not present in the subject index. These
failures to specify measure or otherwise quantify either the static or dynamic load or pressures
within the facet joints may account for some of the reported clinical failures of motion-
sparing/preserving prosthetics. Perhaps these glaring omissions should be taken as a clarion call
for a meaningful leap forward in the ability to quantify the normal and pathological loading
experienced by these important but neglected spinal joints.
The objective of this study was to accurately quantify the spatial and temporal
distribution of intra-articular pressures within the lumbar L4-5 facet joints utilizing previously
validated technology (Welcher et al., 2011a, 2011b). This chapter presents the results from
testing that directly measured the time-dependent and spatial distribution of intra-articular
pressure within cadaveric lumbar spine L4-5 facet joints subjected to 10 Nm pure flexion-
extension bending moments.
186

6.2 Methods
Sensor construction and performance
Bench top testing and preliminary in-vitro performance of the sensors have been
previously published (Welcher, 2011).
An array of seven sensors was used as shown in Figure 91. Positive X position values are
lateral of the facet surface vertical midline and positive Y position values are superior to the facet
horizontal midline. The sensor array geometry was then approximated as an ellipse to match the
general facet shape (Panjabi et al., 1993, Tischer et al., 2006), centered at the origin (0 mm, 0
mm), with a minor horizontal axis of 12 mm, and major vertical axis of 15 mm. There was
approximately 0.5 mm between each of the five horizontal sensor elements and approximately 3.5
mm separating the three vertical sensor elements. The coordinate system was established such
that the 0 mm, 0 mm sensor position is located at the approximate anatomical center of the
average size facet surface. As the height and width of the sensor array is based upon calculated
means of a large number of facets (Table 31), if the facet under study is wider, there will be a
slight medial shift of the sensor array center relative to the anatomical center. Conversely, a
narrow facet will results in a slight lateral shift of the senor array relative to the anatomical center.
The right side of Figure 91 is a to-scale representation of the sensor size, and array size, and
placement.

187

Figure 91: Seven sensor array (left) and sensor position and coordinate system (right).

Specimen preparation and grading
Six frozen human lumbosacral L1-Sacrum specimens were utilized. There were 3 female
and 3 male specimens with a mean age of 61 years (range 26-76 years) as seen in Table 33.

188

Table 33: Human lumbosacral specimens (n=6)

The extents of disc and facet degeneration have both been shown to influence the response of the
lumbar spine (White & Panjabi, 1990, Niosi & Oxland, 2004). A review of existing grading
systems for lumbar disc and face degeneration (Kettler & Wilke, 2006) recommended based on a
statistical analysis, the Thompson grading method (Thompson et al., 1990) for macroscopic
anatomical grading of lumbar disc degeneration, the Pfirrmann method (Pfirrmann et al., 2001)
Thompson Grading
Pfirmann
Grading
Weishaupt
Grading
Orthopedic
Surgeon
Biomechanical
Engineer
Orthopedic
Surgeon
Orthopedic
Surgeon
Spine Age Gender Segments Disc (Gross) Disc (Gross) Disc (MRI) R Facet (MRI)
1 60 M L1-Sacrum
L3-4 3 3 3
L4-5 4 4 4 0
L5-S1 5 5 5
2 67 F L1-Sacrum
L3-4 3 3 2
L4-5 3 3 3 2
L5-S1 4 3 4
3 76 M L1-Sacrum
L3-4 3 4 3
L4-5 4 4 4 3
L5-S1 5 5 4
4 73 F L1-Sacrum
L3-4 3 3 4
L4-5 4 3 4 2
L5-S1 n/a n/a n/a
5 65 M L1-Sacrum
L3-4 4 4 3
L4-5 3 4 4 1
L5-S1 4 4 4
6 26 F L1-Sacrum
L3-4 3 3 4
L4-5 4 3 4 3
L5-S1 n/a n/a 4
189

for MRI grading of lumbar disc degeneration, and Weishaupt’s method (Weishaupt et al., 1999)
for MRI grading of lumbar facet joints. These methods were used to grade the lumbar discs and
facet joints. The gross disc were graded by a board certified Orthopedic Surgeon and a
Biomechanical Engineer and the disc and facet MRI scans were graded by the Orthopedic
Surgeon. The results are shown in Table 33.
Each specimen was allowed to thaw at room temperature and was meticulously dissected
and cleaned of any surrounding paraspinal musculature, spinal nerves, and spinal cord. Special
attention was directed to preserving all osteoligamentous portions of the lumbosacral spine.
Radiographs of each specimen were examined along with donor history to insure that no tumors,
previous fractures, or metabolic diseases affecting the tissues were present. Each specimen was
anchored in a two-part polyurethane potting solution (BJB Enterprises, Inc., Tustin, CA). The
most superior and inferior vertebral bodies were submerged in the potting solution to the
approximate level of the vertical vertebrae midline, with the spine oriented so that the L4
vertebral body superior endplate was oriented horizontally to the ground, similar to a normal erect
position. Hydration of the specimen was maintained prior to testing by wrapping it with gauze
soaked in isotonic saline and during testing by routinely spraying it with isotonic saline solution.
The specimens were mounted into a Bose Kinematic Spine Testing Device (Bose
Corporation, Minnetonka, MN) for mechanical testing as shown in Figure 92. This device
provided six degrees-of-freedom motion through a combination of five rotational and three
translational actuators (pneumatic or servo-electronic) and was instrumented with two load cells
and local rotational torque cells for each gimbal, which together allowed for essentially pure
moment bending.
The mechanical testing protocol consisted of three cycles of dynamic pure moment
flexion/extension loading (±10 Nm). The loading rate was 1°/second until the load limit of ±10
190

Nm was reached (Cholewicki et al., 1996, Goel et al., 2006, Panjabi et al., 1989, Yamamoto et al.,
1989).
Each vertebral body was instrumented with an array of four infrared light emitting diodes
(LED). These were used to calculate three-dimensional linear and rotational responses using an
optoelectronic motion capture system (Optotrak 3020, Northern Digital Inc., Waterloo, Ontario,
Canada).
The L4-5 bilateral facet joint capsules were incised approximately 180 degrees along the
posterior edge from the superior apex to the inferior apex to facilitate sensor placement. The new
sensor array was placed typically in the right facet and a dummy array with the same sensor
configurations was placed in the contralateral facet as seen at the posterior of the spine in Figure
92. Both facets were cut to maintain symmetry and avoid any asymmetries in the response
associated with asymmetries in joint capsule integrity (Zhu et al., 2008).

Figure 92: Measurement sensor in right L4-5 facet and dummy array in left facet.
191

The mechanical testing device, motion analysis system and pressure sensor system were
synchronized off a common time zero signal. Data were collected on the mechanical testing
device and motion analysis system at 50 Hz, and at 250 Hz on the pressure sensor system. Data
collected on the second loading cycle of each loading condition were utilized for data analysis.
Based on prior power spectral density analysis (Welcher, 2011b) of the raw 250 Hz pressure
sensor data, all data were filtered with a four pole Butterworth phaseless low pass filter set at 5
Hz.

6.3 Results

The sensor array measured a general reduction in joint contact pressure from the neutral
position to full measured flexion. Conversely, an increase in joint contact pressures was
measured from neutral to full measured extension. Figure 93 represents a typical data set
showing the total L1-Sacrum flexion-extension rotation, rotation of L4 relative to L5, and facet
contact pressure at sensor #5 (at -2 mm, 0 mm. see Figure 91) due to the applied flexion extension
bending moment.
192

Figure 93: Typical system response to applied bending moment (positive moments and
angles are flexion). (in 28 USC k004 – Spine w39)

The non-linear increase in bending moment during a relatively linear increase in L4-5
relative rotation demonstrates the increased flexural stiffness of the L4-5 level beyond
approximately 2-3 degrees of flexion or extension rotation. Facet contact pressure increases
relatively linearly in proportion to applied bending moment up to approximately 2-3 degrees of
L4-5 extension and approximately 7-8 Nm of extension moment (Figure 94). Relatively little
increase in facet intra-articular pressure or L4-5 rotations were observed beyond 8 Nm of applied
moment (Figure 94 and Figure 97).

193

Figure 94: Relationship of applied moment to L4-5 rotation and facet pressure (in K004 –
Spine, w78)

The greatest increases in facet pressures relative to the neutral position values were noted
with increasing extension moments and displacements. As such, facet contact pressures for all
seven sensors in the sensor array are only graphically presented for positions of 50% maximum
L4-5 extension and 100% L4-5 extension (Figure 95 and Figure 96). However mean pressure and
standard deviations for 100% flexion, neutral, 50% extension and 100% extension are given for
each sensor location (Table 34). Data at 7.5 Nm is also included in Table 34 to allow comparison
to a number of other published studies that used a 7.5 Nm bending moment protocol.

194

Figure 95: Spatial pressure distribution at 50% of maximum L4-5 extension.

Figure 96: Spatial pressure distribution at 100% of maximum L4-5 extension.

Relative to the neutral position (Table 34), L4-5 facet pressures across all sensor
locations were on average 3.2 times greater at 50% extension and 9.3 times greater at 100%
195

extension. Average peak facet pressures across all spines at 100% extension ranged from 65 kPa
at sensor #3 to 1,088 kPa at sensor #5 (range from 0 kPa to 3,328 kPa). The highest average
maximum pressures were found in the midline sensor 2 mm medial of the midpoint and in the
most inferior midline sensor (Sensors #5 and #1, Figure 91, respectively). The most superior
sensor (Sensor #3, Figure 91) always had the lowest average peak pressures during extension.
The data did exhibit relatively large standard deviations.
Table 34: Average maximum L4-5 pressures under various loading conditions.

The center of pressure (COP) within the facet joint (Table 35 and Figure 97) started very
near the anatomical center of the facet and migrated medially (toward sensor #5) and inferior
(toward sensor #1) under increasing extension moments. The average COP position value was
never greater than 1.2 mm from the anatomical center in either direction (Figure 97). However,
relatively large standard deviations were also observed in the COP data.

Table 35: Center of pressure location under various loading conditions

Loading Condition Average Maximum Pressure [Standard Deviation] (kPa)
Sensor 1 Sensor 2 Sensor 3 Sensor 4 Sensor 5 Sensor 6 Sensor 7
100% flexion 136.3 [165.2] 12.7 [14.2] 67.5 [96.5] 15.4 [19.8] 100.8 [202.1] 127.7 [210.7] 79.5 [161.2]
Neutral 82.7 [149.1] 87.6 [142.5] 23.9 [31.6] 36.3 [57.5] 86.8 [163.9] 165.5 [209.9] 123.5 [209.4]
50% extension 250.6 [329.3] 183.4 [191.3] 41.9 [41.0] 224.6 [153.4] 299.0 [292.0] 249.5 [395.3] 569.7 [829.8]
7.5 Nm 745.1 [803.7] 448.0 [436.0] 47.3 [39.2] 621.5 [376.8] 820.8 [730.1] 374.1 [542.5] 722.9 [1087.6]
100% extension (10 Nm) 973.0 [843.5] 611.1 [691.8] 65.3 [70.5] 786.2 [586.2] 1087.9 [1188.7] 396.7 [543.9] 861.3 [1184.8]
Loading Condition Center of Pressure Location (mm)
X position Y position
Min Max Mean [S.D.] Min Max Mean [S.D.]
100% flexion -0.23 1.97 0.30 [0.83] -3.10 0.70 -0.72 [1.51]
Neutral -2.18 1.49 0.18 [1.25] -1.06 0.81 0.00 [0.73]
50% extension -2.48 1.67 -0.28 [1.61] -1.27 0.28 -0.36 [0.62]
7.5 Nm -2.76 1.33 -0.57 [1.45] -2.66 -0.12 -1.10 [0.94]
100% extension (10 Nm) -2.75 0.83 -0.61 [1.26] -3.66 -0.16 -1.20 [1.27]
196

Figure 97: Mean center of pressure migration under extension loading from neutral
position to 50% extension, 7.5 Nm and 100% extension (10 Nm).

6.4 Discussion
The objective of this study was to accurately quantify the spatial and temporal
distribution of intra-articular pressures within the lumbar L4-5 facet joints utilizing previously
validated technology described in Chapters 3, 4 and 5 (Welcher et al., 2011a, 2011b). This new
measurement method has overcome many of the shortcomings of previous methods such as strain
gauges (Zhu, et al., 2008), Fujifilm (www.fujifilm.com/products/prescale/guide/feature/spec.html,
Rudert et al., 1987, Liggins & Finlay, 1992, Liggins et al., 1994, Liggins et al., 1995a, 1995b,
Liggins & Finlay, 1997, Lia et al., 2001, Hale & Brown, 1992, Caldwell et al., 1993), and thin
film sensors from Tekscan (Ferguson-Pell & Cardi, 1992, Ferguson-Pell & Cardi, 1993, Pavlovic
et al., 1993, Woodburn & Helliwell, 1996, Werner, 1995, Pitei et al., 1996, Otto et al., 1998, Otto
197

et al., 1999, Luo et al., 1983, McPoil et al., 1995, Sumiya et al., 1998, Ferguson-Pell et al., 2000,
Buis & Convery, 1997, Wilson et al., 2006) and Novel (Pavlovic et al., 1993, Werner, 1995,
Giacomozzi, 2009, Giacomozzi, 2010, Arndt, 2003, Martinelli et al., 2006, Polliack et al., 2006).
Most notably, relative to other technology, the new sensor array is capable of accurately
measuring temporal and spatial distribution of intra-articular pressure with the sensor array being
curved within the relatively tightly curved lumbar spine facets without sensor induced errors
(non-zero offset when curved under no load) or alterations in sensor sensitivity (Welcher, 2011).
Sensor induced errors and alteration in sensor sensitivity with curvature are critical, especially
with sensors calibrated on flat surfaces and then used to measure in curved surfaces, and are
frequently unaddressed shortcomings of many other technologies when used for intra-articular
measurement.
There are several limitations to this study, notably a small sample size of only six spines
spread over a relatively wide range of ages (26-76 years). The rather large standard deviations in
the data indicated high variability between specimens. Although all specimens were screened for
severe pathology, the possibility that localized discontinuities within the facet joint, which may
be caused by small areas of increased degeneration or bony growth, resulted in highly localized
stress concentrations could not be excluded. The facet capsule was cut bilaterally to facilitate
sensor insertion. Cutting the facet capsule does change the response; however, this has been
shown to have only a minor effect up to flexion-extension moments of 11.2 N-m (Tencer, 1982,
Tencer & Mayer, 1983). Additionally a sensor of finite thickness was inserted in what was likely
only a virtual space. At a thickness of 0.6 mm, inserting the sensor element will likely induce
between 10 and 20 N of joint load artifact merely by the sensor's presence (Hedman, 1992).
As it is believed that this is one of the first papers to present data of direct simultaneous
measurement of temporal and spatial lumbar spine facet pressures to pure moment bending, there
198

are no direct comparisons in the published literature. However, portions of the data can be
compared to previously published data as both are shown in Table 36. The reader is cautioned to
critically review prior research in light of the stated limitations in these technologies, most
notably curvature effects relative to calibration methods.
Table 36: Comparison of facet loads to prior research

The range of peak facet pressure values at 100% extension (-10 Nm) spanned from a low
of approximately 0 kPa in Spine 1, sensor #2 to a high of 3,238 kPa in Spine 5, sensor #5 (Figure
95 and Figure 96) are generally in the range of the peak pressures reported using Fuji pressure
sensitive film (Lorenz et al., 1983, Wiseman et al., 2005, Dunlop et al., 1984). Lorenz et al. et al.
(Lorenz et al., 1983) measured peak L4-5 facet pressures (average of left and right facets) of
1,667 kPa with a compressive preload of 196.2 N and extension of 6 to 8 degrees. Other facet
force studies that used Fujifilm demonstrated mean peak pressure of 4,030 kPa ±2,640 kPa with a
Reference-Author (year) Method of assessment Lumbar Level
Number of
facets
Capsule
intact Loading Mode
Facet Force
(N)
Peak Facet
Pressure (kPa)
Fiorini & McCammond, 1976 Math model L3-4 1 Y Sitting, 756 N body weight 48.9 490
Adams & Hutton, 1980 Experimental L1-2 through L5-S1 13 Y 460 N-1000 N comp. 8.5
Lorenz et al., 1983 Fuji Film L4-5 1 N 196 N comp. 89.05 1667.5
Yang & King, 1984 Experimental T12-L2 and L3-5 3 Y 1112 N comp 192.376
Dunlop et al., 1984 Fuji Film L1-2 through L5-S1 24 N
1000 N comp., 200-400 N
shear 5755.2
Shirazi-Adl & Drouin, 1987 FEA L2-3 2 y 10 Nm ext., 190 N comp. 110
Goel, et al., 1987 Experimental L2-L3 and L4-5 5 Y 6.9 Nm ext. 51.9
El-Bohy et al., 1989 Needle Pressure Gauge L1-2 and L4-5 6 Y 532 N comp. 160
Hedman, 1992 FSR-Interlink/Fuji Film L2-3 and L4-5 5 N 1000 N comp. 50-600
Schendel et al., 1993 External surface strain L1-L2 1 Y 10 Nm ext., 190 N comp. 205
Wiseman et al., 2005 Fuji Film L4-5 7 N 15 Nm ext., 700 N comp. 72.5
4030±2640 (ave
900±380)
Goel et al., 2005 FEA L4-5 1 Y 10.6 Nm ext., 400 N comp. 325.6
Rousseau et al., 2006 Tekscan FSR-T L5-S1 24 N 650 N comp., 550 N shear 51.6
Niosi et al., 2008 Tekscan FSR-T L3-4 10 N 7.5 Nm ext. 13.5
Sawa and Crawford, 2008 External surface strain L1-L2 Y 7.5 Nm ext. 51.5
Schmidt et al., 2008 FEA L4-5 1 Y 7.5 Nm ext., 500 N comp 50
Zhu et al., 2008 External surface Strain L3-4 20 N 7.5 Nm ext. 28.6
Zhu et al., 2008 Tekscan FSR-T L3-4 20 N 7.5 Nm ext. 13.3
Natarajan et al., 2008 FEA L4-5 1 Y
Task 2-extending from
flexed while lifting 13.6 kg 350
Ramruttun et al., 2009 Tekscan FSR-T L4-5 1 N 10 Nm ext., 300 N comp. 50
Chen et al., 2009 FEA L3-4 1 Y 10 Nm ext., 400 N comp. 240
Kuo et al., 2010 FEA L4-5 2 N 10 Nm ext., 300 N comp. 95
Botolin et al., 2010 Tekscan FSR-T L3-4 7 N "Pure moment", 400 N comp. ~70 ~1500
Welcher et al., 2011a, 2011b
(current study) Pressure sensor L4-5 6 N Pure moment, 10 Nm F-E n/a 3238
199

15 Nm applied moment and 700 N follower load (Wiseman et al., 2005), and 5,755 kPa (Dunlop
et al., 1984), with a 1000 N compressive load and 200 to 400 N of shear force. A recent study
using the FSR-T and pure moment loading with a 400 N follower load reported L3-4 peak
pressure of approximately 1,500 kPa in extension although the exact moments were not specified
(Botolin et al., 2010).
The pressure distributions within the facet joint are consistent with that presented in the
literature (Lorenz et al., 1983, Wiseman et al., 2005, Schmidt et al., 2008) in that the highest
measured facet pressures in extension were located on the midline and inferior portions of the
facet. The superior sensor #3 had the lowest average peak facet pressure.
The downward migration of the COP would also be consistent with the downward
migration of peak pressures with increasing extension moments reported by others (Lorenz et al.,
1983, Schmidt et al., 2008) however this is the first study to report the additional medial
migration.
As seen in Figure 94 and Figure 97, the rotational stiffness was relatively constant out to
8 Nm as evidenced by the small increase in approximately linear increase facet pressure and
relative L4-5 rotation up to 8 Nm of applied moment. Future research should be conducted
utilizing at least a 10 Nm bending moment as in the present study, as this covers the entire linear
rotational stiffness range and includes the regions of increased stiffness beyond 8 Nm. Testing at
or above 10 Nm also captures data along the way that allows comparison to prior research
conducted at 7.5 Nm and newer studies conducted at 10 Nm.
This research demonstrated a new method to more accurately measure the temporal and
spatial characteristics of intra-articular L4-5 facet pressure under pure moment loading. This
technology overcomes most of the limitations that reduced the accuracy of prior technology. The
pressure magnitude, distribution and temporal characteristics of intra-articular L4-5 facet
200

pressures under pure +/- 10 Nm loading were presented with center of pressure data at neutral,
50% extension, 7.5 Nm moment, and at 100% extension (10 Nm).
Future studies need to be conducted with additional spines under additional loading
conditions such as with compressive follower loads, lateral bending, and rotation. These studies
should include more precise assessment of the facet topography such as with high resolution CT
scans. This may assist in explaining spine to spine variation in intra-articular pressures. Given
the modular nature of the individual sensor, future studies could optimize sensor density and
distribution to more specifically quantify effects.
6.5 Conclusions
This is the first study to present direct simultaneous measurement of the spatial and
temporal characteristics of facet intra-articular pressure.
The greatest increases in facet pressures relative to the neutral position values were noted
with increasing extension moments and displacements. The L4-5 facet exhibited very little
change in intra-articular pressure when flexed from neutral to a bending moment of 10 Nm
(100% flexion). When extended to -10 Nm (100% extension) average peak intra-articular facet
increased by an average factor of 9.3 over neutral position values.
Average peak facet pressures across all spines at 100% extension ranged from 65 kPa at
the most superior midline sensor (Sensor #3, Figure 91) to 1,088 kPa at the midline sensor 2 mm
medial of the midpoint (Sensor #5, Figure 91). Peak pressures ranged from 0 kPa to 3,328 kPa.
The highest average maximum pressures were found in the midline sensor 2 mm medial of the
midpoint and in the most inferior midline sensor (Sensors #5 and #1, Figure 91, respectively).
The most superior sensor (#3) always had the lowest average peak pressures during all phases of
extension.
201

The center of pressure within the facet joint started near the anatomical center of the facet
and migrated medially and inferior under increasing extension moments. The average COP
position value was never greater than 1.2 mm from the anatomical center in either the vertical or
horizontal direction.

202

CHAPTER 7: SUMMARY AND CONCLUSIONS

Back pain is an enormously costly and growing problem with current estimated total
costs for back pain exceeding $100 billion annually. Many studies attest to the high frequency of
back complaints in modern society, with 60–85% of all people experiencing back pain at some
time in their lives. In the United States between the periods of 1971 to 1986, the number of
people disabled from back pain increased 168%, a rate 14 times the rate of population growth
(Haldeman, 1990). Recent studies (Deyo et al., 2005, 2009, Manchikanti et al., 2009) document a
629% increase in Medicare expenditures for epidural steroid injections; a 423% increase in
expenditures for opioids for back pain; a 307% increase in the number of lumbar magnetic
resonance images among Medicare beneficiaries; and a 220% increase in spinal fusion surgery
rates from 1990 to 2001 in the United States. A substantial portion of back pain has been shown
to be attributable to the facet joints, which have also been shown to be major load bearing
elements.
The specific aims of the current study as articulated in Chapter 1 were:
1. Critically review the applicability and accuracy of all currently published methods
and techniques for measurement of intra-articular force, pressure and/or contact area.
2. Build, validate and utilize a bench top test apparatus for performance evaluation of
existing and potential new technologies intended to measure the spatial and temporal
pressure distributions within human lumbar facets joint under load.
3. Bench top performance validation of the new sensor array with physiologically
relevant test criteria including: geometric constraints of length, width, thickness and
sensor spatial resolution; and the effects of sensor array curvature, frequency
response, linearity, drift, hysteresis, repeatability, and total system cost.
203

4. Perform preliminary in vitro testing to identify any performance issues associated
with in vitro testing compared to bench top testing.
5. Provide more accurate and more robust experiments to gather data on lumbar spine
mechanics by directly measuring the in vitro intra-articular temporal and spatial
distributions of lumbar facet joint contact pressures when spine specimens are
subjected to pure flexion-extension bending moments.
Chapter 2 outlines the enormity of the back pain problem, how this may be attributable to
the facet joint, and that the facet is a significant load bearing element in the spine. Basic lumbar
spine anatomy is discussed and the unique geometric constraints of the lumbar facets are
reviewed. This segues into Chapter 3 with an introduction describing that much of the described
data taken with various methods in the literature have methodological flaws rendering the data
tenuous at best. This is supported by the subsequent extensive literature review. Chapter 3
culminates with specific evaluation and ultimate rejection of what was initially deemed the most
viable technologies.
Chapter 4 addresses specific aims 2 and 3 by discussing the construction of a bench top
test apparatus to evaluate the sensors in a controlled environment and the results of the bench top
testing. With the new sensors geometric constraints of length, width, thickness and sensor spatial
resolution were evaluated. Additionally, the effects of sensor array curvature, frequency
response, linearity, drift, hysteresis, repeatability, and total system cost were assessed. The new
sensor array was approximately 0.6 mm in thickness, scalable to below the nominal 12 mm wide
by 15 mm high lumbar spine facet joint size, offered no inherent limitations on the number or
spacing of the sensors with less than 1.7% cross talk with sensor immediately adjacent to one
another. Perhaps the most important finding, no difference was observed in sensor performance
down to a radius of curvature of 7 mm and a 0.66±0.97% change in sensor sensitivity was
204

observed at a radius of 5.5 mm. The sensor array had less than 0.07 dB signal loss up to 5.5 Hz,
linearity was 0.58±0.13% full scale (FS), drift was less than 0.2% FS at 250 s and less than 0.6%
FS at 700 s, and hysteresis was 0.78±0.18%. Repeatability was excellent with a coefficient of
variation less than 2% at pressures between 0 and 1.000 MPa. Total system cost was relatively
small as standard commercially available data acquisition systems could be utilized, with no
specialized software, and individual sensors within an array can be replaced as needed.
Preliminary in vitro data is presented in Chapter 5. This demonstrated the viability of the
new sensor array in quantifying temporal and spatial distributions of pressure within the L4-5
facet joint. This preliminary testing demonstrated that in vitro durability was problematic with
very high initial sensor failure rates. Manufacturing changes and orientation optimization of each
individual sensor relative to load direction improved durability. Durability was improved and
considered acceptable in light of the ease and relatively low cost of individual sensor
replacement.
Chapter 6 represents the ultimate application of the previously developed and validated
technology and addresses aim number 5. The utility of the sensor in more accurately quantifying
spatial and temporal changes in lumbar spine facet intra-articular pressure was demonstrated with
six fresh-frozen human cadaveric lumbosacral specimens tested under pure moment bending
(±10Nm). L4-5 facet contact pressures were continuously measured at seven locations within the
facet. Center of pressure at various phases of loading was calculated. The data demonstrated an
increase in facet pressure with increasing extension moments and displacements. Facet contact
pressure increased relatively linearly in proportion to applied bending moment up to
approximately 2-3 degrees of L4-5 extension and approximately 7-8 Nm of extension moment.
The highest average maximum pressures of 1,087 kPa were found in the midline sensor 2 mm
medial of the midpoint and 973 kPa in the most inferior midline sensor. The most superior
205

midline sensor always had the lowest average peak pressures during extension. The center of
pressure started very near the anatomical center of the facet and migrated medially and inferior
under increasing extension moments.
In summary, completion of specific aims 1-4 culminated in the development of a viable
test technology that was assessed and shown to be capable of accurately measuring the in vitro
intra-articular temporal and spatial distributions of lumbar facet joint contact pressures under
load. The ultimate goal of this study, which is to provide more accurate and more detailed data
on intra-articular facet loads, is articulated in specific aim number 5 and was achieved as
demonstrated in Chapter 6.
Because the sensor array satisfies the requirements of being able to measure the time
varying and spatial distribution of pressure within the highly curved lumbar facet joints, it should
be extremely useful in understanding the overall mechanics of the lumbar spine. Implementing
this technology will advance the understanding of general spine biomechanics, assist in validation
of finite element models, and provide data for prosthetic design and testing.
Although this research is targeted at designing a sensor to function in the lumbar facet
joint, it is reasoned that satisfying the design requirements for the relatively demanding lumbar
facet joint ensures viability of the sensor array in less geometrically demanding joints and surface
interfaces such as the knee and ankle (Howell, et al., 2010, Dam et al., 2007, Nuno & Ahmed,
2003, Welcher et al., 2011b). Additionally much of the validation data provided in Chapter 4 also
has general applicability to non intra-articular usage of this technology such as a source of tactile
feedback for both prosthetic limb design and prosthetic fit, foot pressure distribution in podiatric
applications and gait analysis.

206

REFERENCES

[1] Abumi K, Panjabi MM, Kramer KM, Duranceau J, Oxland T, Crisco JJ. Biomechanical
evaluation of lumbar spinal stability after graded facetectomies. Spine 1990;
15(11):1142-7.

[2] ACI 214R-02 Evaluation of Strength Test Results of Concrete. American Concrete
Institute Report by ACI Committee 214, 2002.

[3] Adams MA, Hutton WC. Cadaver lumbar intervertebral joints. Spine 1980; 5(5), 483–
484.

[4] Adams M, Hutton W. The effect of posture on the role of the apophysial joints in
resisting intervertebral compressive forces. Journal of Bone and Joint Surgery 1980; 62-
B (3), 358-362.

[5] Adams MA, Hutton WC. The relevance of torsion to the mechanical derangement of the
lumbar spine. Spine 1981; 6(3), 241–248.

[6] Adams MA, Hutton WC. The mechanical function of the lumbar apophyseal joints. Spine
1983; 8(3), 327–330.

[7] Afoke NP, Byers PD, Hutton WC. Contact pressures in the human hip joint. Journal of
Bone and Joint Surgery 1987. 69B (4):536-541.

[8] Ahmed AM. A pressure distribution transducer for in-vitro static measurements in
synovial joints. Journal of Biomechanical Engineering 1983; 105(3):309-14.

[9] Ahroni JH, Boyko EJ, Forsberg R. Reliability of F-scan in-shoe measurements of plantar
pressure. Foot & Ankle International 1998; 19(10):668-73.

[10] Alexander IJ, Chao EY, Johnson KA. The assessment of dynamic foot-to-ground contact
forces and plantar pressure distribution: a review of the evolution of current techniques
and clinical applications. Foot & Ankle 1990; 11(3):152-67.

[11] Anderson GB. The biomechanics of the posterior elements of the lumbar spine,
introductory comments. Spine 1983; 8:326.

[12] Andersson GB. The epidemiology of spinal disorders. The Adult Spine: Principles and
Practice 1997; 93–141.

[13] Andersson GB. Epidemiological features of chronic low-back pain. Lancet. 1999;
354(9178):581-5.

[14] Aritomi H, Morita M, Yonemoto K. A simple method of measuring the foot sole pressure
of normal subjects using Prescale pressure-detecting sheets. Journal of Biomechanics
1983; 16(2):157-165.
207

[15] Arndt A. Correction for sensor creep in the evaluation of long-term plantar pressure data.
Journal of Biomechanics 2003; 36(12):1813-7.

[16] ASTM Standard F2077-03, “Test Methods for Intervertebral Body Fusion Devices,”
ASTM International, West Conshohocken, PA, 2003.

[17] ASTM Standard F2346-05, “Standard Test Methods for Static and Dynamic
Characterization of Spinal Artificial Discs,” ASTM International, West Conshohocken,
PA, 2005.

[18] ASTM Standard F2423-05, “Standard Guide for Functional, Kinematic, and Wear
Assessment of Total Disc Prostheses,” ASTM International, West Conshohocken, PA,
2005.

[19] ASTM Standard F2624-07, “Standard Test Methods for Static, Dynamic, and Wear
Assessment of Extra-Discal Spine Motion Preserving Implants,” ASTM International,
West Conshohocken, PA, 2007.

[20] ASTM Standard F2790-10, “Standard Practice for Static and Dynamic Characterization
of Motion Preserving Lumbar Total Facet Prostheses,” ASTM International, West
Conshohocken, PA, 2010.

[21] Avramov AI, Cavanaugh JM, Ozaktay CA, Getchell TV, King AI. The effects of
controlled mechanical loading on group-II, III, and IV afferent units from the lumbar
facet joint and surrounding tissue. An in vitro study. J Bone Joint Surg Am 1992;
74:1464-1471.

[22] Bachus KN, DeMarco AL, Judd KT, Horwitz DS, Brodke DS. Measuring contact area,
force, and pressure for bioengineering applications: using Fujifilm and Tekscan systems.
Medical Engineering & Physics 2006; 28(5):483-8.

[23] Badgley CE. The articular facet in relation to low back pain and sciatic radiation. J Bone
Joint Surg Am 1941; 23:481-496.

[24] Barnett S, Cunningham JL, Hunter G, McDowell N, West S. A comparison of vertical
force and temporal parameters produced by an in-shoe pressure measuring system and a
force platform. Clinical Biomechanics 1999; 14(8): 552.

[25] Barnett S, Cunningham JL, West S. A comparison of vertical force and temporal
parameters produced by an in-shoe pressure measuring system and a force platform.
Clinical Biomechanics 2000; 15(10), 781-785. (Also re-published Clinical Biomechanics
2001; 16(4):353-7.)

[26] Basmajian, JV. Grant’s Method of Anatomy 8
th
Edition 1971; p.18.

208

[27] Baumhauer JF, Wervey R, McWilliams J, Harris GF, Shereff MJ. A comparison study of
plantar foot pressure in a standardized shoe, total contact cast, and prefabricated
pneumatic walking brace. Foot & Ankle International 1997; 18(1):26-33.

[28] Bay BK, Hamel AJ, Olson SA, Sharkey NA. Statically equivalent load and support
conditions produce different hip joint contact pressures and periacetabular strains.
Journal of Biomechanics 1997; 30:193-196.

[29] Bierung-Sorenson F. Physical measurements as risk indicators for low back trouble over
a one year period. Spine 1984;9: 106-19.

[30] Boswell MV, Colson JD, Sehgal N, Dunbar EE, Epter R. A systematic review of
therapeutic facet joint interventions in chronic spinal pain. Pain Physician 2007;
10(1):229-53.

[31] Botolin S, Puttlitz C, Baldini T, Petrella A, Burger E, Abjornson C, Patel V. Facet joint
biomechanics at the treated and adjacent levels after total disc replacement. Spine 2011.
36(1):E27-32.

[32] Boyd LA, Bontrager EL, Mulroy SJ, Perry J. The reliability and validity of the Novel
Pedar system of in-shoe pressure measurement during free ambulation. Gait & Posture
1997; 5(2):165. (Abstract Podium presentation Session 16: Foot ankle II).

[33] Brand PW, Ebner JD. Pressure sensitive devices for denervated hands and feet: A
preliminary communication. J Bone Joint Surg Am 1969; 51:109-116.

[34] Brimacombe JM, Anglin C, Hodgson AJ, Wislon DR. Validation of calibration
techniques for Tekscan pressure sensors. ISB XXth Congress – ASB 29
th
Annual Meeting,
July 31-August 5, Cleveland, Ohio 2005; p. 263.

[35] Brimacombe JM, Wilson DR, Hodgson AJ, Ho KC, Anglin C. Effect of calibration
method on Tekscan sensor accuracy. Journal of Biomechanical Engineering 2009;
131(3):034503.

[36] Brown TD, Shaw DT. A technique for measuring instantaneous in vitro contact stress
distributions in articular joints. Journal of Biomechanics 1982; 15(4):329-33.

[37] Buis AW, Convery P. Calibration problems encountered while monitoring stump/socket
interface pressures with force sensing resistors: techniques adopted to minimize
inaccuracies. Prosthetics & Orthotics International 1997; 21(3):179-82.

[38] Butler D, Trafimow JH, Andersson GB, McNeil TW, Huckman MS. Discs degenerate
before facets. Spine 1990; 15:111–113.

[39] Butterman G, Kahmann R, Lewis J, Bradford D. An experimental method for measuring
force on the spinal facet joint: Description and application of the method. Journal of
Biomechanical Engineering 1991a; 113: 375-386.

209

[40] Butterman GR, Schendel MJ, Kahmann RD, Lewis JL, Bradford DS. In vivo facet joint
loading of the canine lumbar spine. Spine 1991b; 17(1): 81-92.

[41] Caldwell NJ, Hale JE, Rudert JM, Brown TD. Technical note: An algorithm for
approximate crinkle artifact compensation in pressure-sensitive film recordings. J
Biomech 1993; 26:1001–1009.

[42] Cavanagh PR, Hewitt FG, Perry JE. In-shoe plantar pressure measurement: A review.
The Foot 1997; 2: 185-194.

[43] Cavanaugh JM, Ozaktay AC, Yamashita HT, King AI. Lumbar facet pain: biomechanics,
neuroanatomy and neurophysiology. Journal of Biomechanics 1996; 29(9):1117-29.

[44] Cavanaugh JM, Ozaktay AC, Yamashita T, Avramov A, Getchell TV, King AI.
Mechanisms of low back pain: a neurophysiologic and neuroanatomic study. Clinical
Orthopaedics & Related Research 1997; 335:166-80.

[45] Cavanaugh JM, Lu Y, Chen C, Kallakuri S. Pain generation in lumbar and cervical facet
joints. J Bone Joint Surg Am 2006; 88:63-67.

[46] Chen B, Bates B. Comparison of F-Scan in-sole and AMTI forceplate systems in
measuring vertical ground reaction force during gait. Physiotherapy Theory and Practice
2000; 16:43–53.

[47] Chen MW, Park C, Lee K, Lee S. In situ contact analysis of the prosthesis components of
Prodisc-L in lumbar spine following total disc replacement. Spine 2009; 34(20):E716-
E723.

[48] Choi HS. Park YJ. Lee SM. Kim KM. Functional characteristics of a new electrolarynx
"Evada" having a force sensing resistor sensor. Journal of Voice 2001; 15(4):592-9.

[49] Cholewicki J, Crisco JJ, Oxland TR, Yamamoto I, Panjabi MM. Effects of posture and
structure on three-dimensional coupled rotations in the lumbar spine. A biomechanical
analysis. Spine 1996; 21:2421-8.

[50] Cobb J, Claremont DJ. Transducers for foot pressure measurement: survey of recent
developments. Medical & Biological Engineering & Computing 1995; 33(4):525-32.

[51] Collier JP, Mayor MB, McNamara JL, Surprenant VA, Jensen RE. Analysis of the failure
of 122 polyethylene inserts from uncemented tibial knee components. Clinical
Orthopaedics 1991; 273:232-242.

[52] Collis WJ, Jayson MI. Measurement of pedal pressures. An illustration of a method.
Annals of the Rheumatic Diseases 1972; 31(3):215-7.

210

[53] Coronado JL, Uylloa-Perez A, Garcia-Cordova F, Villalba-Fernandez JJ. Experimental
development for thumb-index pinch force measurements during object manipulation with
application to robotic hands. 1999 IEEE International Conference on Systems, Man, and
Cybernetics, 1999. IEEE SMC '99 Conference Proceedings, Tokyo, Japan 1999; 6:786-
791.

[54] Cramer GD. Cantu JA. Dorsett RD. Greenstein JS. McGregor M. Howe JE. Glenn WV.
Dimensions of the lumbar intervertebral foramina as determined from the sagittal plane
magnetic resonance imaging scans of 95 normal subjects. Journal of Manipulative &
Physiological Therapeutics 2003; 26(3):160-70.

[55] Cranmer HH, Patterson RP. The accuracy of the sensor array used in the FSCAN system.
Proceedings of the RESNA International Conference. Toronto, Ontario, Canada June 6-
11, 1992; p. 7.

[56] Dam EB, Folkesson J, Pettersen PC, Christiansen C. Automatic morphometric cartilage
quantification in the medial tibial plateau from MRI for osteoarthritis grading.
Osteoarthritis & Cartilage 2007; 15(7):808-18.

[57] Demirci S, Jubel A, Andermahr J, Koebke J. Chondral thickness and radii of curvature of
the femoral condyles and talar trochlea. International Journal of Sports Medicine 2008;
29(4):327-30.

[58] Denis F. The three column spine and its significance in the classification of acute
thoracolumbar spinal injuries. Spine 1983; 8(8):817-31.

[59] Denis F. Spinal instability as defined by the three-column spine concept in acute spinal
trauma. Clinical Orthopaedics & Related Research 1984; 189:65-76.

[60] Deyo RA, Gray DT, Kreuter W, Mirza S, Martin BI. United States trends in lumbar
fusion surgery for degenerative conditions. Spine 2005; 30:1441-1445.

[61] Deyo RA, Mirza SK, Turner JA, Martin BI. Overtreating chronic back pain: Time to back
off?. J Am Board Fam Med 2009; 22:62-68.

[62] Drewniak EI, Crisco JJ, Spenciner DB, Fleming BC. Accuracy of contact area
measurements with thin-film pressure sensors. American Society of Biomechanics.
Annual Meeting, Blacksburg, Virginia 2006; 114.

[63] Drewniak EI, Crisco JJ, Spenciner DB, Fleming BC. Accuracy of circular contact area
measurements with thin-film pressure sensors. Journal of Biomechanics 2007;
40(11):2569-72.

[64] Dreyer SJ, Dreyfuss PH. Low back pain and the zygapophyseal (facet) joints. Arch
Physiology Medical Rehabilitation 1996; 77:290–300.

211

[65] Dunlop RB, Adams MA, Hutton WC. Disc space narrowing and the lumbar facet joints.
Journal of Bone & Joint Surgery - British Volume 1984; 66(5):706-10.

[66] El-Bohy AA. A comprehensive analysis of the facet joint in relation to low back pain.
Diss. Wayne State University 1988; 169 pages, AAT 8910322.

[67] El-Bohy AA, Yang KH, King AI. Experimental verification of facet load transmission by
direct measurement of facet lamina contact pressure. J Biomech 1989; 22:931–41.

[68] El-Bohy AA. Sensory innervation of the lumbar facet joint capsule and its possible
relation to low back pain. Diss. Wayne State University 1990; 175 pages: AAT 9029619.

[69] Ellis SJ, Hillstrom H, Cheng R, Lipman J, Garrison G, Deland JT. The development of an
intraoperative plantar pressure assessment device. Foot & Ankle International 2009;
30(4):333-40.

[70] Emborg J, Spaich E, Andersen OK. Vertical unloading produced by electrically evoked
withdrawal reflexes during gait: preliminary results. IFMBE Proceedings Volume 16,
11th Mediterranean Conference on Medical and Biomedical Engineering and Computing
MEDICON 2007, 26-30 June 2007 Ljubljana, Slovenia; 669-672.

[71] Emborg J, Spaich EG, Andersen OK. Withdrawal reflexes examined during human gait
by ground reaction forces: site and gait phase dependency. Medical & Biological
Engineering & Computing 2009; 47(1):29-39.

[72] Ewing CL, King AI, Prasad P. Structural consideration of the human vertebral column
under +Gz impact acceleration. Office of Naval Research. Journal of Aircraft 1972;
9(1):84-90. (Office of Naval Research. Contract N00014-69-A-0235, March 20, 1973)

[73] Ferguson-Pelll MW. Design criteria for the measurement of pressure at body/support
interfaces. Engineering in Medicine 1980; 9(4):209-214.

[74] Ferguson-Pell M, Cardi M. Pressure Mapping Systems. Center for Rehabilitation
Technology, Helen Hayes Hospital, NY. Team Rehab Report 1992; 3(7):28-32.

[75] Ferguson-Pell M, Cardi M. Prototype Development and Comparative Evaluation of
Wheelchair Pressure Mapping System. Center for Rehabilitation Technology, Helen
Hayes Hospital, West Haverstraw, NY 1993; 78-91.

[76] Ferguson-Pell M, Hagisawa S, Bain D. Evaluation of a sensor for low interface pressure
applications. Medical Engineering & Physics 2000; 22(9):657-63.

[77] Fick R., von Bardeleben KH., Handbuch der Anatomie und Mechanik der Gelenke, 1904;
Jena.

[78] Fiorini GT, McCammond D. Forces on lumbo-vertebral facets. Annals of Biomedical
Engineering 1976; 4:354-363.

212

[79] Fraden J. Handbook of Modern Sensors: Physics, Designs, and Applications. New York:
SpringerVerlag, 2003.

[80] Frisina W, Lehneis HR. Pressure mapping: a preliminary report. Journal of Biomechanics
1970; 3(5):526.

[81] Frymoyer JW, Pope MH, Clements JH, Wilder DG, MacPherson B, Ashikaga T. Risk
factors in low-back pain. An epidemiological survey. J Bone Joint Surg Am 1983;
65:213-8.

[82] Fuji Film Features- Specifications. Retrieved February 28, 2010, from
www.fujifilm.com/products/prescale/guide/feature/spec.html

[83] Fukubayashi T, Kurosawa H. The contact area and pressure distribution pattern of the
knee: A study of normal and osteoarthrotic knees. Acta Orthopaedica Scandinavica 1980;
51:871-879.

[84] Ghormley RK. Low back pain with special reference to the articular facets with
presentation of an operative procedure. JAMA 1933; 101(23):1773-1777.

[85] Giacomozzi C. Appropriateness of plantar pressure measurement devices in a research
context: a comparative technical assessment. International Society of Biomechanics
Congress XXII, Cape Town, South Africa 2009; Abstract 134.

[86] Giacomozzi C. Appropriateness of plantar pressure measurement devices: a comparative
technical assessment. Gait & Posture 2010; 32(1):141-4.

[87] Giacomozzi C. (editor). Istituto Superiore di Sanità Conference. Assessment of pressure
measurement devices (PMDs) for their appropriate use in biomechanical research and in
the clinical practice. Istituto Superiore di Sanità. Abstract book 2010; 27: ISTISAN
Congressi 10/C3.

[88] Giles LG. Taylor JR. The effect of postural scoliosis on lumbar apophyseal joints.
Scandinavian Journal of Rheumatology 1984; 13(3):209-20.

[89] Goel VK, Winterbottom JM, Weinstein JN, Kim YE. Load sharing among spinal
elements of a motion segment in extension and lateral bending. J. Biomechanical
Engineering 1987; 109(4):291–297.

[90] Goel VK, Grauer JN, Patel TC, Biyani A, Sairyo K, Vishnubhotla S, Matyas A, Cowgill
I, Shaw M, Long R, Dick D, Panjabi MM, Serhan H. Effects of Charité artificial disc on
the implanted and adjacent spinal segments mechanics using a hybrid testing protocol.
Spine 2005; 30(24):2755-64.

[91] Goel VK, Panjabi MM, Patwardhan AG, Dooris AP, Serhan H. Test protocols for
evaluation of spinal implants. J Bone Joint Surg Am 2006; 88(2):103-9.

213

[92] Goldberg HI, Deyo RA, Taylor VM, Cheadle AD, Conrad DA, Loeser JD, Heagerty PJ,
Diehr P. Can evidence change the rate of back surgery? A randomized trial of
community-based education. Effective Clinical Practice 2001; 4(3):95-104.

[93] Goldthwait JE. The lumbosacral articulation: an explanation of many cases of lumbago,
sciatica and paraplegia. Boston Medical Surgical Journal 1911; 164:365.

[94] Goodfellow J, Hungerford DS, Zindel M. Patello-femoral joint mechanics and pathology.
1. Functional anatomy of the patello-femoral joint. J Bone Joint Surg Br 1976; 58B:287-
290.

[95] Gotfried Y, Bradford DS, Oegema TR: Facet joint changes after chemonucleolysis-
induced disc space narrowing. Spine 1986 11:944-950.

[96] Graf PM. The EMED System of foot pressure analysis. Clinics in Podiatric Medicine &
Surgery 1993; 10(3):445-54.

[97] Griffiths HJ, Parantainen H. Disease of the lumbosacral facet joints. Neuroimaging Clin
North Am 1993; 3:567-575.

[98] Guo HR, Tanaka S, Halperin W, Cameron L. Back pain prevalence in US industry and
estimates of lost workdays. American Journal of Public Health 1999; 89(7):1029-1035.

[99] Gurney JK, Kersting UG, Rosenbaum D. Between-day reliability of repeated plantar
pressure distribution measurements in a normal population. Gait & Posture 2008;
27(4):706-9.

[100] Ha K, Schendel M, Lewis J, Ogilvie J. Effect of immobilization and configuration on
lumbar adjacent-segment biomechanics. Journal of Spinal Disorders 1993; 6(2): 99-105.

[101] Hachisuka K, Takahashi M, Ogata H, Ohmine S, Shitama H, Shinkoda K. Properties of
the flexible pressure sensor under laboratory conditions simulating the internal
environment of the total surface bearing socket. Prosthetics & Orthotics International
1998; 22(3):186-92.

[102] Hakim NS, King AI. A three dimensional finite element dynamic response analysis of a
vertebra with experimental verification. Journal of Biomechanics 1979; 12(4):277-285,
287-292.

[103] Hakim NS, King AI. Static and dynamic articular facet loads. Proc. 20th Stapp Car
Crash Conf 1976; 607–639.

[104] Hakim NS King AI. Programmed replication of in situ (whole-body) loading conditions,
during in vitro (substructure) testing of a vertebral column segment, J. Biomech 1976;
9:629–632.

[105] Hakim S, King AI. An experimentally validated 3-D finite element model of a vertebra,
Adv. Bioengineering. ASME 1976; 11–12.
214

[106] Hakim S, King AI. Study of the strain distribution over the cortical surface of lumbar
vertebrae. Proc. 23rd Int. Instrum. Symp1977; 145–153.

[107] Haldeman S. Presidential Address, North American Spine Society: Failure of the
Pathology Model to Predict Back Pain. Spine 1990; 15(7): 718-724.

[108] Hale JE, Brown TD. Contact Stress Gradient Detection Limits of Pressensor Film. ASME
J. Biomech 1992; 114:352–357.

[109] Harkness EF, Macfarlane GJ, Silman AJ, McBeth J. Is musculoskeletal pain more
common now than 40 years ago?: Two population-based cross-sectional studies.
Rheumatology (Oxford) 2005; 44:890-895.

[110] Harris ML, Morberg P, Bruce WJ, Walsh WR. An improved method for measuring
tibiofemoral contact areas in total knee arthroplasty: a comparison of K-scan sensor and
Fujifilm. Journal of Biomechanics 1999; 32(9):951-8.

[111] Hart LG, Deyo RA, Cherkin DC. Physician office visits for low back pain. Frequency,
clinical evaluation, and treatment patterns from a U.S. national survey. Spine 1995; 20:
11–19.

[112] Hartmann JM, Rudert MJ, Pedersen DR, Baer TE, Goreham-Voss CM, Brown TD.
Compliance-dependent load allocation between sensing versus non-sensing portions of a
sheet-array contact stress sensor. Iowa Orthop. J 2009; 29:43–47.

[113] Harvey WL, Hatch RA, Osborne JW. Computerized occlusal analysis: an evaluation of
the sensors. Journal of Prosthetic Dentistry 1991; 65(1):89-92.

[114] Hasler EM, Herzog W, Fick GH. Appropriateness of plane pressure-sensitive film
calibration for contact stress measurements in articular joints. Clin Biomech (Bristol,
Avon) 1996; Sep;11(6):358-360.

[115] Haut RC, Ile TM, DeCamp CE. Mechanical responses of the rabbit patello-femoral joint
to blunt impact. Journal of Biomechanical Engineering 1995; 117:402-408.

[116] Hedman TP. A New Transducer for Facet Force Measurement in the Lumbar Spine:
Benchmarks and In Vitro Test Results. Journal of Biomechanics 1992; 25(1): 69-80.

[117] Hedman TP. Mechanical response of human lumbar spines to seated postural loads.
Dissertation University of Strathclyde (Scotland) 1995; 118 pages

[118] Hedman TP, Fernie GR. Mechanical response of the lumbar spine to seated postural
loads. Spine 1997; 22(7):734-43.

215

[119] Hillstrom H, Giacomozzi C, Lenhoff M, Zifchock R, Vanadurongwan B, Gross D,
McIennan C, Hannan M. Accuracy of the Novel, EMED X and Tekscan Matscan Plantar
Pressure Measurement Systems. Proceedings of the XI Emed Scientific Meeting (ESM)
2008; 87.

[120] Hoffmann K, Decker K. Inaccuracies in measurement of contact stresses due to the
measuring grid of a foil sensor. Proceedings of the First International Conference on
Sensing Technology, Palmerston North, New Zealand 2005; p. 696-699.

[121] Hollinger A., Wanderley MW. Evaluation of Commercial Force-Sensing Resistors. Input
Devices and Musical Interaction Laboratory (IDMIL) Report, January 2006.

[122] Holzapfel GA, Stadler M. Role of facet curvature for accurate vertebral facet load
analysis. European Spine Journal 2006; 15(6):849-56.

[123] Howell SM, Howell SJ, Hull ML. Assessment of the radii of the medial and lateral
femoral condyles in varus and valgus knees with osteoarthritis. J Bone Joint Surg Am,
2010; 92(1):98-104.

[124] Hsiao H, Guan J, Weatherly M. Accuracy and precision of two in-shoe pressure
measurement systems. Ergonomics 2002; 45(8):537-555.

[125] Huberti HH, Hayes WC. Patellofemoral contact pressures; the influence of Q-angle and
tendofemoral contact. Journal of Bone and Joint Surgery 1984; 66A (5):715-723.

[126] Hughes J, Pratt L, Linge K, Clark P, Klenerman L. Reliability of pressure measurements:
the EMED F system. Clin Biomech 1991; 6:14–18.

[127] Hurkmans HL, Bussmann JB, Selles RW, Horemans HL, Benda E, Stam HJ, Verhaar JA.
Validity of the Pedar Mobile system for vertical force measurement during a seven-hour
period. Journal of Biomechanics 2006; 39(1):110-8.

[128] Hurkmans HL, Bussmann JB, Selles RW, Horemans HL, Benda E, Stam HJ, Verhaar JA.
Validity of the Pedar Mobile system for vertical force measurement during a long-term
period. Proceedings of the IX Emed Scientific Meeting (ESM), Leeds, UK 2004; 60.

[129] Hurkmans HL, Bussmann JB, Benda E, Verhaar JA, Stam HJ. Accuracy and repeatability
of the Pedar Mobile system in long-term vertical force measurements. Gait & Posture
2006; 23(1):118-25.

[130] ISO/IEC Guide International Vocabulary of Metrology — Basic and General Concepts
and Associated Terms, VIM. 2007. 99-120.

[131] ISO 18192-1:2011 Implants for surgery -- Wear of total intervertebral spinal disc
prostheses -- Part 1: Loading and displacement parameters for wear testing and
corresponding environmental conditions for test.

216

[132] ISO 18192-2:2010 Implants for surgery -- Wear of total intervertebral spinal disc
prostheses -- Part 2: Nucleus replacements.

[133] Jayson, M. Compression stress in the posterior elements and pathologic consequences.
Spine 1983; 8(3):338-339.

[134] Jensen TR, Radwin RG, Webster JG. A conductive polymer sensor for measuring
external finger forces. Journal of Biomechanics 1991; 24(9):851-8.

[135] Johanson NA, Cerynik DL, Pasquale M. Measuring patellofemoral forces and pressures
in a simulated operating room environment. The Journal of Arthroplasty 2011;
26(1):137-43.

[136] Kalpen A, Seitz P. Comparison between the force values measured with the Pedar system
and Kistler platform. Proceedings of the Fourth EMED User Meeting, Ulm. Gait &
Posture 1994; 2:236–7.

[137] Katz JN. Lumbar disc disorders and low back pain: socioeconomic factors and
consequences. J Bone Joint Surgery 2006; 88(2):21–4.

[138] Kelly M. Physical changes in the prolapsed disc. Lancet 1958; 2:584.

[139] Kettler A. Wilke HJ. Review of existing grading systems for cervical or lumbar disc and
facet joint degeneration. European Spine Journal. 15(6):705-18, 2006 Jun.

[140] King AI, Prasad P, Ewing CL. Mechanism of spinal injury due to caudocephalad
acceleration. Orthopedic Clinics of North America 1975; 6(1): 19-31.

[141] Kong Y, Freivalds A, Kim SE. Evaluation of handles in a maximum gripping task.
Ergonomics 2004; 47(12):1350-1364.

[142] Kuo C, Hu H, Lin R, Huang K, Lin P, Zhong Z, Hseih M. Biomechanical analysis of the
lumbar spine on facet joint force and intradiscal pressure - a finite element study. BMC
Musculoskeletal Disorders 2010; 11:151.

[143] Kwak SD, Colman WW, Ateshian GA, Grelsamer RP, Henry JH, Mow VC. Anatomy of
the human patellofemoral joint articular cartilage: surface curvature analysis. Journal of
Orthopaedic Research 1997; 15(3):468-72.

[144] LaBar G. A battle plan for back injury prevention. Occup Hazards 1992; 54:29–33.

[145] Lai CH, Li-Tsang CW. Validation of the Pliance X System in measuring interface
pressure generated by pressure garment. Burns 2009; 35(6):845-51.

[146] Lee TQ, Anzel SH, Bennett KA, Pang D, Kim WC. The influence of fixed rotations
deformities of the femur on the patellofemoral contact pressures in human cadaver knees.
Clinical Orthopedics 1994; 302:69-74.

217

[147] Liau J, Hu C, Cheng C, Huang C, Lo W. The influence of inserting a Fuji pressure
sensitive film between the tibiofemoral joint of knee prosthesis on actual contact
characteristics. Clinical Biomechanics 2001; 16:160-166.

[148] Liggins AB, Finlay JB. Recording Contact Areas and Pressures in Joint Interfaces.
Proceedings of the International Conference on Experimental Mechanics: Technology
Transfer between High Tech Engineering and Biomechanics 1992.

[149] Liggins AB, Hardie WR, Finlay JB. Sterilization of Fuji pressure-sensitive film. Medical
Engineering & Physics 1994; 16(6):496-500.

[150] Liggins AB, Surry K, Finlay JB. Sealing Fuji prescale pressure sensitive film for
protection against fluid damage: the effect on its response. Strain 1995a; 31(2):57-62.

[151] Liggins AB, Hardie WR, Finlay JB. The spatial and pressure resolution of Fuji pressure
sensitive film. Experimental Mechanics 1995b; 35(2):166-173.

[152] Liggins AB, Finlay JB. Image-Averaging in the analysis of data from pressure-sensitive
film. Experimental Techniques 1997; 21(3):19-22.

[153] Lilius G, Laasonen EM, Myllynen P, Harilainen A, Grönlund G. Lumbar facet joint
syndrome. A randomized clinical trial. J Bone Joint Surg Br.1989; 71(4):681-4.

[154] Lord M. Foot pressure measurement: a review of methodology. Journal of Biomedical
Engineering 1991; 3(2):91-9.

[155] Lord M, Reynolds DP, Hughes JR. Foot pressure measurement: a review of clinical
findings. Journal of Biomedical Engineering 1986; 8(4):283-94.

[156] Lorenz M, Patwardhan A, Vanderby R. Load-bearing characteristics of lumbar facets in
normal and surgically altered spinal segments. Spine 1983; 8(2):122-130.

[157] Lu Y, Chen C, Kallakuri S, Patwardhan A, Cavanaugh JM. Development of an in vivo
method to investigate biomechanical and neurophysiological properties of spine facet
joint capsules. European Spine Journal 2005a; 14(6):565-72.

[158] Lu WW, Luk KD, Holmes AD, Cheung KM, Leong JC. Pure shear properties of lumbar
spinal joints and the effect of tissue sectioning on load sharing. Spine 2005b; 30(8):E204.

[159] Luo ZP, Buttermann GR, Lewis JL. Determination of spinal facet joints from extra-
articular strains-a theoretical validation. Journal of Biomechanics 1996; 29(6):785-790.

[160] Luo ZP, Berglund LJ, An KN. Validation of F-Scan pressure sensor system: a technical
note. Journal of Rehabilitation Research & Development 1998; 35(2):186-91.

[161] Maalej N, Bhat S, Zhu H, Webster JG, Tompkins WJ, Wertsch JJ, Bach-y-Rita P. A
conductive polymer pressure sensor. Engineering in Medicine and Biology Society, 1988.
Proceedings of the Annual International Conference of the IEEE 1988; 2:770-711.
218

[162] Maalej N, Webster JG, Tompkins WJ, Wertsch JJ. A conductive polymer pressure sensor
array. Engineering in Medicine and Biology Society. Images of the Twenty-First Century.
Proceedings of the Annual International Conference of the IEEE 1989; 4: 116 - 1117.

[163] Mak AF, Zhang M, Boone DA. State-of-the-art research in lower-limb prosthetic
biomechanics-socket interface: a review. Journal of Rehabilitation Research &
Development 2001; 38(2):161-74.

[164] Maldjian C, Mesgarzadeh M, Tehranzadeh J. Diagnostic and therapeutic features of facet
and sacroiliac joint injection. Anatomy, Pathophysiology, and Technique. Radiologic
Clinics of North America 1998; 36(3).

[165] Manchikanti L, Pampati VS, Fellows B, Bakhit CE. Prevalence of facet joint pain in
chronic low back pain. Pain Physician 1999;2: 59–64.

[166] Manchikanti L, Pampati V, Rivera J, Fellows B, Beyer C, Damron K. Role of facet joints
in chronic low back pain in the elderly: a controlled comparative prevalence study. Pain
Practice 2001; 1(4):332-7.

[167] Manchikanti L, Boswell MV, Singh V, Pampati V, Damron KS, Beyer CD. Prevalence of
facet joint pain in chronic spinal pain of cervical, thoracic, and lumbar regions. BMC
Musculoskeletal Disorders 2004; 5:15.

[168] Manchikanti L, Singh V, Datta S, Cohen S, Hirsch J. Comprehensive review of
epidemiology, scope, and impact of spinal pain. Pain Physician 2009; 12:E35-E70.

[169] Maness WL, Chapman RJ, Dario LJ. Laboratory evaluation of a direct reading digital
occlusal sensor. Journal of Dental Research 1985; 64: Abstract 1199.

[170] Maness WL. Comparison of the duration of occlusal contacts during habitual closure
using the digital occlusal sensor. Journal of Dental Research 1986; 65: Abstract 141.

[171] Maness WL, Benjamin M, Podoloff R, Bobick A, Golden R. Computerized occlusal
analysis: a new technology. Quintessence Int. 1987; 18(4):287–292.

[172] Manouel M, Pearlman HS, Belakhlef A, Brown TD. A miniature piezoelectric polymer
transducer for in vitro measurement of the dynamic contact stress distribution. Journal of
Biomechanics 1992; 25(6):627-35.

[173] Marder RA, Swanson TV, Sharkey NA, Duwelius PJ. Effects of partial patellectomy and
reattachment of the patellar tendon on patellofemoral contact areas and pressures. Journal
of Bone and Joint Surgery 1993; 75A (1):35-45.

[174] Martinelli L, Hurschler C, Rosenbaum D. Comparison of capacitive versus resistive joint
contact stress sensors. Clin. Orthop. Relat. Res 1986; 447:214–220.

219

[175] Martinelli, L., Hurschler, C., and Rosenbaum, D. Comparison of Capacitive Versus
Resistive Joint Contact Stress Sensors, Clin. Orthop. Relat. Res. 2006; 447, pp. 214–220.

[176] McCutcheon CW. The frictional properties of animal joints. Wear 1962; 5:1–17.

[177] McDermott ID, Lie DT, Edwards A, Bull AM, Amis AA. The effects of lateral meniscal
allograft transplantation techniques on tibio-femoral contact pressures. Knee Surgery,
Sports Traumatology, Arthroscopy 2008; 16(6):553-60.

[178] McNally DS. The objectives for the mechanical evaluation of spinal instrumentation have
changed. European Spine Journal 2002; 11 Suppl. 2:S179-85.

[179] McPoil TG, Cornwall MW, Yamada W. A comparison of two in-shoe plantar pressure
measurement systems. The Lower Extremity 1995; 2(2): 95-103.

[180] Melkerson MN, Kirkpatrick JS, Griffith SL, eds. Symposium on Spinal Implants, Are We
Evaluating Them Appropriately? (2001: Dallas, Tex.) Spinal implants: are we evaluating
them appropriately? ASTM special technical publication: 1431 2003. West
Conshohocken, PA: ASTM International.

[181] Merskey H, Bogduk N. Classification of chronic pain. Descriptions of Chronic Pain
Syndromes and Definition of Pain Terms, 2nd ed. IASP Press, Seattle, WA 1994; 180-
181.

[182] Miller JA, Haderspeck KA, Schultz AB. Posterior element load in lumbar motion
segments. Spine 1983; 8(3): 331-337.

[183] Mooney V, Robertson J. The facet syndrome. Clinical Orthopedics 1976; 115:149-156.

[184] Moro-oka T, Matsuda S, Miura H, Nagamine R, Urabe K, Kawano T, Higaki H, Iwamoto
Y. Patellar tracking and patellofemoral geometry in deep knee flexion. Clinical
Orthopedics & Related Research 2002; (394):161-8.

[185] Morin EL, Bryant JT, Reid SA, Whiteside RA. Calibration Issues of Tekscan Systems for
Human Pressure Assessment. Paper presented at the RTO HFM Specialists’ Meeting on
“Soldier Mobility: Innovations in Load Carriage System Design and Evaluation”, held in
Kingston, Canada, 27-29 June 2000, and published in RTO MP-056. ADP011007.

[186] Morris JM, Lucas DB, Bresler B. Role of the trunk stability of the spine. J Bone Joint
Surgery Am 1961; 43:327-351.

[187] Morris AS. Measurement & Instrumentation Principles. Third Edition. Oxford:
Butterworth Heinemann, 2001.

[188] Muller O, Parak WJ, Wiedemann MG, Martini F. Three-dimensional measurements of
the pressure distribution in artificial joints with a capacitive sensor array. Journal of
Biomechanics 2004; 37(10):1623-5.

220

[189] Muriuki MG, Gilbertson LG, Harner CD. Characterization of the performance of a
custom program for image processing of pressure sensitive film. Journal of
Biomechanical Engineering 2009; 131(1):014503.

[190] Nachemson A. Lumbar lntradiscal Pressure. Acta orthop. Scandinav 1960; Suppl. 43.

[191] Nachemson A. The influence of spinal movements on the lumbar intradiscal pressure and
on the tensile stresses in the annulus fibrosus. Acta orthop. Scandinav 1963; 33(1-4):
183-207.

[192] Nagata H, Schendel M, Transfeldt E, Lewis J. The effects of immobilization of long
segments of the spine on the adjacent and distal facet force and lumbosacral motion.
Spine 1993; 18(24): 2471-2479.

[193] Natarajan RN, Lavender SA, An HA, Andersson GB. Biomechanical response of a
lumbar intervertebral disc to manual lifting activities: a poroelastic finite element model
study. Spine 2008; 33(18):1958-65.

[194] Netter F. Atlas of Human Anatomy. CIBA-GEIGY Corporation, Summit, New Jersey
1993.

[195] Nikonovas A, Harrison AJ, Hoult S, Sammut D. The application of force-sensing resistor
sensors for measuring forces developed by the human hand. Proc Inst Mech Eng H 2004;
218(2):121-6.

[196] Niosi CA, Oxland TR. Degenerative Mechanics of the Lumbar Spine. The Spine Journal
2004; 4(6) S1:S202-S208.

[197] Niosi CA, Wilson DC, Zhu Q, Keynan O, Wilson DR, Oxland TR. The effect of dynamic
posterior stabilization on facet joint contact forces: an in vitro investigation. Spine 2008;
33(1):19-26.

[198] Nuno N, Ahmed AM. Three-dimensional morphometry of the femoral condyles. Clinical
Biomechanics 2003; 18(10):924-32.

[199] Orlin MN, McPoil TG. Plantar pressure assessment. Physical Therapy 2000; 80(4):399-
409.

[200] Otto JK, Brown TD, Heiner AD, Pedersen DR, Callaghan JJ. Characterization of the
dynamic response of a piezoresistive contact stress sensor. 44
th
Annual Meeting,
Orthopedic Research Society, March 16-19, New Orleans 1998; 23: 80.

[201] Otto JK, Brown TD, Callaghan JJ. Static and dynamic response of a multiplexed-array
piezoresistive contact sensor. Expt Mech 1999; 39:317–323.

[202] Palmer KT, Walsh K, Bendall H, Cooper C, Coggon D. Back pain in Britain:
comparison of two prevalence surveys at an interval of 10 years. Br Med J 2000;
320:1577–1578.
221

[203] Panjabi MM, Oxland T, Takata K, Goel V, Duranceau J, Krag M. Articular facets of the
human spine. Quantitative three-dimensional anatomy. Spine 1993; 18(10):1298-310.

[204] Panjabi M, Yamamoto I, Oxland T, Crisco JJ. How does posture affect coupling in the
lumbar spine? Spine 1989; 14:1002-11.

[205] Papaioannou G, Protopappas VC, Panagiotos T, Mitrogiannis C, Nianios G, Tashman S.
A new method for pressure sensor equilibration and conditioning. Brazilian Journal of
Biomotricity 2008; 2(3): 176-195.

[206] Papaioannou G, Nianios G, Mitrogiannis C, Fyhrie D, Tashman S, Yang KH. Patient-
specific knee joint finite element model validation with high-accuracy kinematics from
biplane dynamic Roentgen stereogrammetric analysis. Journal of Biomechanics 2008;
41(12):2633-8.

[207] Park D, Yun MH, An application of psychophysical ratings to external force estimation.
Computers & Industrial Engineering 1996; Volume 31, Issues 3-4, Pages 675-680.
[208] Park MC, Pirolo JM, Park CJ, Tibone JE, McGarry MH, Lee TQ. The effect of abduction
and rotation on footprint contact for single-row, double-row, and modified double-row
rotator cuff repair techniques. American Journal of Sports Medicine 2009; 37(8):1599-
608.

[209] Pavlovic JL, Takahashi Y, Bechtold JE, Gustilo RB, Kyle RE. Can the Tekscan sensor
accurately measure dynamic pressures in the knee joint?" Proceedings of the American
Society of Biomechanics 1993; 135-136.

[210] Pax RA, Webster JG, Radwin RG. A conductive polymer sensor for the measurement of
palmar pressures. Proc. A. ht. Conf IEEE Eng Med. Biol. 1989; 11:1483-1484.

[211] Pfirrmann CW. Metzdorf A. Zanetti M. Hodler J. Boos N. Magnetic resonance
classification of lumbar intervertebral disc degeneration. Spine 2001; 26(17):1873-8.

[212] Pitei DL, Ison K, Edmonds ME, Lord M. Time-dependent behavior of a force-sensitive
resistor plantar pressure measurement insole. Proceedings of the Institution of
Mechanical Engineers. Part H - Journal of Engineering in Medicine 1996; 210(2):121-5.

[213] Polliack AA, Sieh RC, Craig DD, Landsberger S, McNeil DR, Ayyappa E. Scientific
validation of two commercial pressure sensor systems for prosthetic socket fit.
Prosthetics & Orthotics International 2000; 24(1):63-73.

[214] Polliack AA, Craig DD, Sieh RC, Landsberger S, Mcneal DR. Laboratory and clinical
tests of a prototype pressure sensor for clinical assessment of prosthetic socket fit.
Prosthetics & Orthotics International 2002; 26(1):23-34.

[215] Pollintine P, Przybyla AS, Dolan P, Adams MA. Neural arch load-bearing in old and
degenerated spines. Journal of Biomechanics 2004; 37(2):197-204.
222

[216] Pope MH, Andersson GBJ, Frymoyer JW, Chaffin DB, editors. Occupational low back
pain: Assessment, treatment and prevention. St. Louis: Mosby, 1991.

[217] Potthast W, Lersch C, Segesser B, Koebke J, Bruggemann GP. Intraarticular pressure
distribution in the talocrural joint is related to lower leg muscle forces. Clinical
Biomechanics 2008; 23(5):632-9.

[218] Praemer A, Furnes S, Rice DP. Musculoskeletal conditions in the United States.
Rosemont: Am Acad Orthop Surg, 1999: 1–99.

[219] Prasad P, King AI, Ewing CL. The role of the articular facets during +Gz acceleration.
Journal of Applied Mechanics 1974; 41: 321-326.

[220] Prasad P, King AI. An experimentally validated dynamic model of the spine. Journal of
Applied Mechanics 1974; 41(E3):546-550.

[221] Putti V. New conceptions in the pathogenesis of sciatic pain. Lancet 1927; 2:53-60.

[222] Putti AB, Arnold GP, Cochrane L, Abboud RJ. The Pedar in-shoe system: repeatability
and normal pressure values. Gait & Posture 2007; 25(3):401-5.

[223] Putti AB, Arnold GP, Cochrane LA, Abboud RJ. Normal pressure values and
repeatability of the Emed ST4 system. Gait & Posture 2008; 27(3):501-5.

[224] Quaney B, Meyer K, Cornwall MW, McPoil TG. A comparison of the dynamic
pedobarograph and EMED systems for measuring dynamic foot pressures. Foot & Ankle
International 1995; 16(9):562-6.

[225] Quesada PM, Rash GS, Jarboe N. Assessment of Pedar and F-Scan Revisted. Clinical
Biomechanics 1997; 12(3):S15. (Abstracts of the 5
th
EMED User’s Meeting, Penn State).

[226] Quesada PM, Rash GS. Simultaneous PEDAR and F-scan plantar pressure measurements
during walking. Gait & Posture 1997; 5(2):164-165. (Abstracts Abstract Podium
presentation Session 16: Foot ankle II).

[227] Quesada PM, Rash GS. Quantitative assessment of simultaneous capacitive and resistive
plantar pressure measurements during walking. Foot Ankle International 2000;
21(11):928-34.

[228] Ramruttun AK, Wong HK, Goh JC, Ruis JN. An improved methodology for measuring
facet contact forces in the lumbar spine. IFMBE Proceedings 13th International
Conference on Biomedical Engineering ICBME 2008; 23:1805-1809.

[229] Rikli DA. Honigmann P. Babst R. Cristalli A. Morlock MM. Mittlmeier T. Intra-articular
pressure measurement in the radioulnocarpal joint using a novel sensor: in vitro and in
vivo results. Journal of Hand Surgery - American Volume 2007; 32(1):67-75, 2007.

223

[230] Ronsky JL, Herzog W, Brown TD, Pederson DR, Grood ES, Butler DL. In vivo
quantification of the cat patellofemoral joint contact stresses and areas. Journal of
Biomechanics 1995; 28:977-983.

[231] Rosenbaum D, Becker HP. Plantar pressure distribution measurements. Technical
background and clinical applications. Foot and Ankle Surgery 1997; 3(1):1-14.

[232] Rousseau MA, Bradford DS, Bertagnoli R, Hu SS, Lotz JC. Disc arthroplasty design
influences intervertebral kinematics and facet forces. Spine 2006a; 6(3):258-66.

[233] Rousseau MA, Bradford DS, Hadi TM, Pedersen KL, Lotz JC. The instant axis of
rotation influences facet forces at L5/S1 during flexion/extension and lateral bending.
European Spine Journal 2006b; 15(3):299-307.

[234] Rudert MJ, Baratz ME, Rehak DC, Fu FH. Loading characteristics of Pressensor: a
pressure sensitive film. Proceedings of the 11th Annual Meeting of the American Society
of Biomechanics 1987; 185-186.

[235] Sangeorzan BJ, Tencer AF, Wagner U, Harrington RM. Contact pressure in the hindfoot.
Journal of Rehabilitation Research and Development 1990; 28:53-54.

[236] Sangeorzan BJ, Wagner UA, Harrington RM, Tencer AF. Contact characteristics of the
subtalar joint: The effect of talar neck misalignment. Journal of Orthopaedic Research
1992; 10(4):544-551.

[237] Sawa AG, Crawford NR. The use of surface strain data and a neural networks solution
method to determine lumbar facet joint loads during in vitro spine testing. Journal of
Biomechanics 2008; 41: 2647-2653.

[238] Schendel MJ. Experimental measurements of lumbar spine biomechanics to validate a
finite element model of the lumbar spine. Diss. University of Minnesota 1990; 189: AAT
9100973.

[239] Schendel MJ, Wood KB, Buttermann GR, Lewis JL, Ogilvie JW. Experimental
measurement of ligament force, facet force, and segment motion in the human lumbar
spine. Journal of Biomechanics 1993; 26(4-5):427-38.

[240] Schmidt H, Heuer F, Wilke HJ. Interaction between finite helical axes and facet joint
forces under combined loading. Spine 2008; 33(25):2741-8.

[241] Schwartz RP, Heath AL. The definition of human locomotion on the basis of
measurement: With description of oscillographic method. J Bone Joint Surg Am 1947;
29: 203-214.

[242] Sehgal N, Shah RV, McKenzie-Brown A, Everett CR. Diagnostic utility of facet
(zygapophyseal) joint injections in chronic spinal pain: a systematic review of evidence.
Pain Physician 2005; 8:211-224.

224

[243] Serhan HA, Varnavas G, Dooris AP, Patwadhan A, Tzermiadianos M. Biomechanics of
the posterior lumbar articulating elements. Neurosurgical Focus 2007; 22(1):E1.

[244] Sharma M, Langrana NA, Rodriguez J. Role of ligaments and facets in lumbar spinal
stability. Spine 1995; 20(8):887-900.

[245] Shirazi-Adl A, Drouin G. Load-bearing role of facets in a lumbar segment under sagittal
plane loadings. Journal of Biomechanics 1987; 20(6):601-13.

[246] Shirazi-Adl A. Finite-element evaluation of contact loads on facets of an L2–L3 lumbar
segment in complex loads. Spine 1991; 16(5), 533–541.

[247] Shirazi-Adl A. Nonlinear stress analysis of the whole lumbar spine in torsion –
mechanics of facet articulation. Journal of Biomechanics 1994; 27(3), 289–299.

[248] Sicard J.A. Nevrodocites et funiculites. Press, med. 1918. V.2.-P. 9-11.

[249] Simon WH. Scale effects in animal joints. I. Articular cartilage thickness and
compressive stress. Arthritis & Rheumatism 1970; 13(3):244-56.

[250] Singerman RJ, Pederson DR, Brown TD. Quantitation of pressure-sensitive film using
digital image scanning. Experimental Mechanics 1987; 27 (1): 99-105.

[251] Skipor AF, Miller JA, Spencer DA Schultz AB. Stiffness properties and geometry of
lumbar spine posterior elements. Journal of Biomechanics 1985; 18(11):821-30.

[252] Slingsby LS, Jones A, Waterman-Pearson AE. Use of a new finger-mounted device to
compare mechanical nociceptive thresholds in cats given pethidine or no medication after
castration. Research in Veterinary Science 2001; 70(3):243-6.

[253] Smith BP, Hudson WB. The use of Force Sensing Resistors for muscle force
measurements. Biomedical Sciences Instrumentation 1994; 30:153-8.

[254] Smith BT, Coiro DJ, Finson R, Betz RR, McCarthy J. Evaluation of force-sensing
resistors for gait event detection to trigger electrical stimulation to improve walking in
the child with cerebral palsy. IEEE Transactions on Neural Systems & Rehabilitation
Engineering 2002; 10(1):22-9.

[255] Sowa G. Facet-mediated pain. Disease-A-Month 2005; 51(1):18-33.

[256] Spurling, RG. Lesions of the lumbar intervertebral disc. C. Thomas Publ., Springfield, Ill.
1953.

[257] States RA. Pappas E. Precision and repeatability of the Optotrak 3020 motion
measurement system. Journal of Medical Engineering & Technology 2006;. 30(1):11-6.

[258] Stone KT, Vaughan CL. A new transducer technology for measurement of plantar
pressures. International Society Biomechanics XII Congress 1990; Abstract 100.
225

[259] Stormont TJ, An KN, Morrey BF, Chao EY. Elbow joint contact study: Comparison of
techniques. Journal of Biomechanics 1985; 18(5):329-336.

[260] Stuhmiller JH, Sih BR, Modeling for military operational medicine scientific and
technical objective (Improving accuracy of the F-Scan sensor). Jaycor Technical Report
3150.32-01-143. US Arm Medical Research and Material Command Contract No.
DAMD17-00-C-031, June 2001

[261] Suezawa HA, Jacob C, Bernoski FP. The mechanical response of the neural arch of the
lumbosacral vertebra and its clinical significance. International Orthopaedics 1980; 4:
205-209.

[262] Sumiya T, Suzuki Y, Kasahara T, Ogata H. Sensing stability and dynamic response of the
F-Scan in-shoe sensing system: a technical note. Journal of Rehabilitation Research &
Development 1998; 35(2):192-200.

[263] Svensson H, Vedin A, Wihelmsson C, Andersson GBJ. LBP in relation to other diseases
and cardiovascular risk factors. Spine 1983;8: 277-85.

[264] Szivek JA, Anderson PL, Benjamin JB. Average and peak contact stress distribution
evaluation of total knee arthroplasties. Journal of Arthroplasty 1996; 11:952-963.

[265] Tanno I, Oguma H, Murakami G, Sato S, Yamashita T. Which portion in a facet is
specifically affected by articular cartilage degeneration with aging in the human lumbar
zygapophyseal joint? Okajimas Folia Anatomica Japonica 2003; 80(1):29-34.

[266] Taylor VM, Deyo RA, Cherkin DC, Kreuter W. Low-back pain hospitalization: recent
United States trends and regional variations. Spine 1994; 9(11): 1207–13.

[267] Taylor JR. Twomey LT. Age changes in lumbar zygapophyseal joints. Observations on
structure and function. Spine 1986; 11(7):739-4

[268] Tencer AF, Ahmed AM, Burke DL. Some static mechanical properties of the lumbar
intervertebral joint, intact and injured. Journal of Biomechanical Engineering 1982;
104:193-201.

[269] Tencer AF, Mayer TG. Soft tissue strain and facet face interaction in the lumbar
intervertebral joint--part II: calculated results and comparison with experimental data.
Journal of Biomechanical Engineering 1983; 105(3):210-215.

[270] Tencer AF, Viegas SF, Cantrell J, Chang M, Clegg P, Hicks C, O'Meara C, Williamson
JB. Pressure distribution in the wrist joint. Journal of Orthopaedic Research 1988;
6(4):509-517.

[271] Thompson JP. Pearce RH. Schechter MT. Adams ME. Tsang IK. Bishop PB. Preliminary
evaluation of a scheme for grading the gross morphology of the human intervertebral
disc. Spine 1990; 15(5):411-5.
226

[272] Tischer T, Aktas T, Milz S, Putz RV. Detailed pathological changes of human lumbar
facet joints L1-L5 in elderly individuals. European Spine Journal 2006; 15(3):308-15.

[273] Tochigi Y, Rudert MJ, Saltzman CL, Amendola A, Brown TD. Contribution of articular
surface geometry to ankle stabilization. J Bone Joint Surg Am 2006; 88(12):2704-13.

[274] Townsend PR, Rose RM, Radin EL, Raux P. The biomechanics of the human patella and
its implications for chondromalacia. Journal of Biomechanics 1977; 10(7):403-7.

[275] Tura A, Lamberti C, Davalli A, Sacchetti R. Experimental development of a sensory
control system for an upper limb myoelectric prosthesis with cosmetic covering. Journal
of Rehabilitation Research & Development 1998; 35(1):14-26.

[276] Tulsi RS, Hermanis GM. A study of the angle of inclination and facet curvature of
superior lumbar zygapophyseal facets. Spine 1993; 18(10):1311-7.

[277] Urry S. Plantar pressure-measurement sensors. Meas. Sci. Technol 1999: 10:16-32.

[278] Valdevit A, Ortega-Garcia J, Kambic H, Kuroda R, Elster T, Parker R. Characterization
and application of thin film pressure sensors. Bio-Medical Materials and Engineering
1999; 9: 81-88.

[279] van Schaik JP. van Pinxteren B. Verbiest H. et al. The facet orientation circle. A new
parameter for facet joint angulation in the lower lumbar spine. Spine 1997; 22(5):531-6

[280] Van Schaik JP. Verbiest H. Van Schaik FD. The orientation of laminae and facet joints
in the lower lumbar spine. Spine 1985; 10(1):59-63.

[281] Vernon-Roberts B, Pirie CJ: Degenerative changes in the intervertebral discs of the
lumbar spine and their sequelae. Rheumatol Rehabill 1977; 16:13-21.

[282] von Eisenhart-Rothe R, Eckstein F, Muller-Gerbl M, Landgraf J, Rock C, Putz R. Direct
comparison of contact areas, contact stress and subchondral mineralization in human hip
joint specimens. Anatomy and Embryology 1997; 195:279-288.

[283] von Eisenhart-Rothe R, Adam C, Steinlechner M, Muller-Gerbl M, Eckstein F.
Quantitative determination of joint incongruity and pressure distribution during simulated
gait and cartilage thickness in the human hip joint. Journal of Orthopaedic Research
1999; 17:532-539.

[284] Wagner UA, Sangeorzan BJ, Harrington RM, Tencer AF. Contact characteristics of the
subtalar joint: Load distribution between the anterior and posterior facets. Journal of
Orthopaedic Research 1992; 10(4):535-543.

[285] Wang CL, Cheng CK, Chen CW, Lu CM, Hang YS, Liu TK. Contact areas and pressure
distributions in the subtalar joint. Journal of Biomechanics 1995; 28:269-279.

227

[286] Wasielewski RC, Galat DD, Komistek RD. An intraoperative pressure-measuring device
used in total knee arthroplasties and its kinematics correlations. Clinical Orthopedics &
Related Research 2004; 427:171-8.

[287] Wasielewski RC, Galat DD, Komistek RD. Correlation of compartment pressure data
from an intraoperative sensing device with postoperative fluoroscopic kinematic results
in TKA patients. Journal of Biomechanics 2005; 38(2):333-9.

[288] Wasserman D, Wasserman J, Ahn JI. Instrumentation for measuring coupling forces of
hand-held tools. Sound and Vibration 2001; p. 22-25.

[289] Weishaupt D. Zanetti M. Boos N. Hodler J. MR imaging and CT in osteoarthritis of the
lumbar facet joints. Skeletal Radiology 1999; 28(4):215-9.

[290] Welcher JB, Popovich JM, Tawackoli W, Hedman TP. Development of a versatile intra-
articular pressure sensing array. Med Eng Phys 2011a;
doi:10.1016/j.medengphy.2011.03.004.

[291] Welcher JB., Popovich JM, Tawackoli W., Hedman TP. A new transducer for intra-
articular pressure measurement: preliminary in vitro lumbar facet testing. Journal of
Biomechanics 2011b; (submitted for publication)

[292] Werner FW, Green JK, Fortino MD, Mann KA, Short WH. Evaluation of a dynamics
intra-articular Contact pressure sensing system. Trans. Orthop. Res. Soc 1995; 20:705.

[293] Werner FW, Short WH, Green JK, Evans PJ, Walker JA. Severity of scapholunate
instability is related to joint anatomy and congruency. Journal of Hand Surgery -
American Volume 2007; 32(1):55-60.

[294] White AA. Panjabi MM. Physical properties and functional biomechanics of the spine.
Clinical Biomechanics of The Spine. New York: Lippincott Williams & Wilkins, 1990

[295] Williams JR, Natarajan RN, Andersson GB. Inclusion of regional poroelastic material
properties better predicts biomechanical behavior of lumbar discs subjected to dynamic
loading. Journal of Biomechanics 2007; 40(9):1981-7.

[296] Wilson DC, Niosi CA, Zhu QA, Oxland TR, Wilson DR. Accuracy and repeatability of a
new method for measuring facet loads in the lumbar spine. Journal of Biomechanics
2006; 39(2):348-53.

[297] Wilson DR, Eichler MJ, Hayes WC. Accuracy of the I-scan Pressure Measurement
System. RTO HFM Specialists' Meeting on "Soldier Mobility: Innovations in Load
Carriage System Design and Evaluation ", held in Kingston, Canada 2000; and published
in RTO AIP-056 2001.

[298] Wilson DR, Apreleva MV, Eichler MJ, Harrold FR. Accuracy and repeatability of a
pressure measurement system in the patellofemoral joint. J Biomech 2003; 36:1909–
1915.
228

[299] Wiseman CM, Lindsey DP, Frederick AD, Yerby SA. The effect of an interspinous
process implant on facet loading during extension. Spine 2005; 30(8): 903-907.

[300] Woodburn J, Helliwell PS. Observations on the F-Scan in-shoe pressure measuring
system. Clinical Biomechanics 1996; 11(5): 301-304.

[301] Woodburrn J, Helliwell PS. Observations on the F-Scan in-shoe pressure measuring
system. Clinical Biomechanics 1997; 12(3): S16.

[302] Woodburne RT, Burkel WE. Essentials of Human Anatomy – 8
th
Edition. Oxford
University Press, New York 1988.

[303] Wright Medical Technology. Advance Knee Systems. Downloaded product literature for
Advance Knee Systems,
(http://www.wmt.com/TotalKnee/Physicians/surgicaltechniques/knees/PrimaryTotalKnee
Arthroplasty/ADVANCE®%20Medial-Pivot%20Double-High%20MK118-
504%20R6.04.pdf, 2010).

[304] Wu JZ, Herzog W, Epstein M. Effects of Inserting a Pressensor Film Into Articular Joints
on the Actual Contact Mechanics. Journal of Biomechanical Engineering 1998;
120(5):655-659.

[305] Yamamoto I, Panjabi MM, Crisco T, Oxland T. Three-dimensional movements of the
whole lumbar spine and lumbosacral joint. Spine 1989; 14:1256-60.

[306] Yamashita T, Cavanaugh JM, El-Bohy AA, Getchell TV, King AI. Mechanosensitive
afferent units in the lumbar facet joint. J Bone Joint Surgery 1990; 72-A: 865-70.

[307] Yamashita T, Minaki Y, Ozaktay AC, Cavanaugh JM, King AI. A morphological study
of the fibrous capsule of the human lumbar facet joint. Spine 1996; 21(5):538-43.

[308] Yang K, King A. Mechanism of facet load transmission as a hypothesis for low-back
pain. Spine 1984; 9(6): 557-565.

[309] Yaniger SI. Force sensing resistors: A review of the technology. Electro. Int 1991; p.
666–668.

[310] Yun MH, Cannon D, Freivalds A, Thomas G. An instrumented glove for grasp
specification in virtual-reality-based point-and-direct telerobotics. IEEE Transactions on
Systems, Man, & Cybernetics, Part B: Cybernetics 1997; 27(5):835-46.

[311] Zhu QA, Park YB, Sjovold SG, Niosi CA, Wilson DC, Cripton PA, Oxland TR. Can
extra-articular strains be used to measure facet contact forces in the lumbar spine? An in-
vitro biomechanical study. Proceedings of the Institution of Mechanical Engineers: Part
H - Journal of Engineering in Medicine 2008; 222(2):171-84.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

A generalized scale for the measurement of attitudes toward international issues

PDF

Magnesium-based photocathodes for triggering back-lighted thyratrons

PDF

On the move and in the moment: community formation, identity, and opportunity in South central Los Angeles, 1945-2008

PDF

Convergent morphological evolution in hyenas and dogs: the relationship between form and function

PDF

The effects of various factors on the academic achievement of Vietnamese students at Gold Coast College

PDF

Differential acceptance of a new role for pharmacists

PDF

Identity change in a therapeutic community

PDF

Realism in the fiction of Frances Burney

PDF

Noise robust speech recognition based on fundamental functional principals of the brain

PDF

Student authoring in the American grain

PDF

Making it to the top: a study of Black partners in major law firms

PDF

The effects of centralization and board membership reform on health occupational licensing policies

PDF

Earnings differences among Black, White and Hispanic males and females: the impact of overeducation, undereducation and discrimination

PDF

A comparison of attitudes of university seniors and sales executives toward outside selling, as revealed by clinical interview and questionnaire methods

PDF

An experimental study of the effect of "human interest" factors on listenability

PDF

Probabilistic analysis of power dissipation in VLSI systems

PDF

The functional gene expression dynamics of Shewanella oneidensis MR-1 under various external voltage potentials

PDF

Effects of an intervention approach on the medication prescribing behavior of physicians

PDF

An argumentation-based approach to negotiation in collaborative engineering design

PDF

Design of a portable infrared spectrometer: application to the noninvasive measurement of glucose

Asset Metadata

Creator Welcher, Judson B. (author)

Core Title Development, validation and testing of a new sensor array for intra-articular pressure measurement: in-vitro human lumbar spine intra-articular facet testing

School Graduate School

Degree Doctor of Philosophy

Degree Program Biomedical Engineering

Degree Conferral Date 2011-08

Tag OAI-PMH Harvest

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC1359890

Unique identifier UC1359890

Legacy Identifier etd-WelcherJud-239

Document Type Dissertation