University of Southern California Dissertations and Theses

Task-dependent modulation of corticomuscular coherence during dexterous manipulation

(USC Thesis Other)

Task-dependent modulation of corticomuscular coherence during dexterous manipulation

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content TASK-DEPENDENT MODULATION OF CORTICOMUSCULAR
COHERENCE DURING DEXTEROUS MANIPULATION
by
Alexander Reyes
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulllment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BIOMEDICAL ENGINEERING)
December 2015
Copyright 2015 Alexander Reyes
Epigraph
\If you want to nd the secrets of the universe, think in terms of energy, frequency and
vibration."
- Nikola Tesla
ii
Dedication
To my mother and father.
iii
Acknowledgements
Over the years, there have been many individuals who have helped me pursue the goals
I had set out to do many years ago. Their encouragement, guidance and friendship
has allowed me to achieve my dreams and without their in
uence I would not have the
strength to overcome the obstacles I have encountered.
First and foremost, I would like to thank Francisco for being my adviser and giving
me the opportunity to work in his lab. He has always provided me with the support I
needed throughout my PhD. Because of his teachings, I have become a better scientist.
I have learned many things from observing his interactions with numerous individuals
over the years and the knowledge and in
uence I have gained from Francisco will not be
forgotten.
Jason. Well, Jason has been a great mentor and friend over the years. Witnessing his
transition from being a post-doc in Francisco's lab to having a lab of his own provided
a great deal of inspiration to me. He has helped shape me into the scientist I am today
and has changed my outlook on life. He has been there to make sure that I stay on track
with my PhD and has always encouraged me.
I would like to thank Dr. Gerald Loeb for his input on the direction of my PhD.
He has challenged me over the years to defend my position on each and every scientic
iv
endeavor. One of the rst pieces of advice he gave to me when I started at USC was
to go to the library and perform a rigorous literary search before running to the lab
to conduct experiments. His advice has saved me from countless hours of unnecessary
experimentation. I have the utmost respect for him.
Thanks to Dr. Charles Liu, Dr. Chrisi Heck, Rossana Arreola, and the EEG tech team
at the USC Keck School of Medicine for their assistance and guidance in my research.
Over the years, Josh became one of my best friends. When I arrived at USC, he seemed
to be the only person in the BME graduate school who shared my love of football. We
would watch nearly every sporting event that came on tv. During those those times when
there was nothing good on, we would busy ourselves with countless games of chess and
weekly poker nights. I learned a lot about how to play poker with Josh, including his
tells.
As senior lab members, Manish and Sudarshan helped me to transition into the lab
and provided me with invaluable guidance and support. Emily Lawrence has become a
very good friend of mine. Our numerous outings of oysters and wine tastings provided
relaxing getaways from the stress of research. Sarine Babikian has always been there for
me and I am forever grateful. I would also like to thank Akira, Victor (for going easy
on me in chess), Brendan (for going easy on me in tennis), Evangelos, Heiko, and Nora.
Chris Laine has been a huge help in my PhD. He has helped shape my research into what
it is today. Overall, I have to thank all of my present and past lab members for tolerating
the outrageous nicknames I came up with for them.
George Tsianos and Matteo have been very good friends from the beginning. Sanna
Sundquist has been a dear friend to me over the years and has provided me with lots of
v
feedback throughout the years. I also would like to thank Dr. Khoo, Mischal Diasanta,
Diana Sabogdal for their academic support.
I would like to thank my Texas friends Brian Bowden, Eric Villase~ nor, Royce, Mered-
ith, Gary, Cindy, and Brad and Mark Creel, as well as Leslie and Sandy for their endless
criticism of the USC football program every year.
My sisters, Melissa and Amy, have always been there to support my eorts, and
for that I am eternally grateful. I would also like to thank my nephews and nieces:
Andres, Christina, Thalia, Maya, Gabriella, Isaiah, Anthony. Lastly, none of this would
be possible without the endless support from my mother and father.
vi
Table of Contents
Epigraph ii
Dedication iii
Acknowledgements iv
List Of Tables x
List Of Figures xi
Abstract xvii
Chapter 1: Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Neural Control of Movement . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.1 Sensory Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.2 Motor Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Previous Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.1 Electrophysiological Studies . . . . . . . . . . . . . . . . . . . . . . 6
1.3.2 Measuring Dexterous Ability . . . . . . . . . . . . . . . . . . . . . 8
1.4 Prior Lab Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5 Signicance of Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.6 Dissertation Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.6.1 Chapter 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.6.2 Chapter 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.6.3 Chapter 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.6.4 Chapter 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.6.5 Chapter 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.6.6 Chapter 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Chapter 2: Data Acquisition Box 17
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2.1 Telemetry Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2.2 Receiving Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2.3 Additional Features . . . . . . . . . . . . . . . . . . . . . . . . . . 22
vii
2.2.4 Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Chapter 3: Localization of a Fine-wire Recording Site and its Propagation
Characteristics 34
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2.1 Ethics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2.2 Recordings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2.3 Spike-Triggered Averaging . . . . . . . . . . . . . . . . . . . . . . . 39
3.2.4 Imaging of a Hematoma . . . . . . . . . . . . . . . . . . . . . . . . 40
3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.3.1 MR Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.3.2 STA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Chapter 4: Power Spectral Density Analysis in Phase II Epilepsy Patients
with Implanted Subdural Electrodes 52
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.1 Ethics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.2 Experimental Paradigm . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.3 Electrocorticography (ECoG) . . . . . . . . . . . . . . . . . . . . . 56
4.2.4 Spectral Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Chapter 5: Introduction to Coherence 63
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.2 Oscillations in the Cortex . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.3 Correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.4 Calculation of Coherence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5.5 Multitaper Power Spectral Density Estimation . . . . . . . . . . . . . . . 77
5.6 Data Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.7 Corticomuscular Coherence Review . . . . . . . . . . . . . . . . . . . . . . 81
5.8 Separation of Power and Coherence . . . . . . . . . . . . . . . . . . . . . . 83
Chapter 6: Synchronous Corticomuscular Oscillations During Dynamic
Unstable Manipulation 87
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6.2.1 Subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6.2.2 Experimental Paradigm . . . . . . . . . . . . . . . . . . . . . . . . 92
6.2.2.1 Task 1: Strength-Dexterity (SD) Test . . . . . . . . . . . 92
6.2.2.2 Task 2: Visuomotor Force Tracking . . . . . . . . . . . . 93
viii
6.2.3 Compliant and Rigid Object Characteristics . . . . . . . . . . . . . 94
6.2.4 Recordings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.2.4.1 Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.2.4.2 Electromyography (EMG) . . . . . . . . . . . . . . . . . . 97
6.2.4.3 Electroencephalography (EEG) . . . . . . . . . . . . . . . 98
6.2.5 Trial Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
6.2.6 Coherence Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.2.7 Selection of EEG Electrodes . . . . . . . . . . . . . . . . . . . . . . 102
6.2.8 Linear Mixed-Eects Model . . . . . . . . . . . . . . . . . . . . . . 103
6.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.3.1 SD Test Performance . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.3.2 Muscle Activation . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.3.3 FDI-EEG Coherence . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6.3.4 LME Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
6.3.5 Power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
6.3.6 Root Mean Square Error of Force . . . . . . . . . . . . . . . . . . . 119
6.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
Chapter 7: Conclusions and Future Work 129
Bibliography 131
ix
List Of Tables
2.1 Serial communication arrangement for transmitted data. Start and stop
bits indicate to the receiver when valid data have arrived. Four address
bits are used to indicate channel number and 12 data bits correspond to
sensor voltage. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1 Coordinates of each recording electrode with respect to MRI coordinate
system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.2 Distance from ne-wire electrode to each surface electrode in millimeters. 46
5.1 Event-related synchronization (ERS) and event-related desynchronization
(ERD) in the cortical motor areas in association with specic motor tasks. 68
5.2 Corticomuscular coherence studies across the dierent frequency bands. . 82
6.1 F
max
values for each subject. 15 right-handed subjects participated in this
study, six of which were female. Mean age was 30:3 4:6 years. Mean
F
max
was 2.2 N, median was 2.2 N and range was 1:8 2:8 N. . . . . . . . 106
x
List Of Figures
1.1 Schematic of the sensorimotor process involved in motor control. Initially,
sensory information from the object being held is obtained. The demands
of the task designate an appropriate motor response to the sensory cortex.
Execution of the motor command causes in a change in the state of the
task resulting in new sensory information and the process continues. . . . 3
2.1 Circuit board layout for Data Acquisition Box (DAB) with individual com-
ponents and description of features. The circuit board measures 1.8 x 2.2
inches. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2 Uni-directional load cell and ti-axial accelerometer next to a US penny for
size comparison. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3 Sensitivity of accelerometer to detect accelerations in three perpendicular
directions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.4 Compressible spring with load cells and accelerometers attached at either
end. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.5 Top trace: Force proles for index nger and thumb during spring compres-
sion. Bottom trace: Euclidean norm of index nger and thumb accelerations. 27
2.6 Subject wearing ve accelerometers attached to the thighs, trunk and ankles. 28
2.7 Five accelerometers attached to the thighs, trunk and ankles of a subject
as they start from rest then perform a light jog. . . . . . . . . . . . . . . . 29
2.8 A comparison of the position data of the Kinect system to the integrated
accelerometer recordings. . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.1 Placement of EMG electrodes around forearm. (a) Point of insertion for
ne-wire electrode. Surface electrodes 2 - 4 can be seen on the posterior-
lateral forearm. (b) Surface electrodes 4 - 6 on the anterior-lateral forearm.
(c) Surface electrodes 6 and 7 on the anterior forearm. . . . . . . . . . . . 39
xi
3.2 Path of hypodermic ned used to insert ne-wire electrodes. The left panel
shows the path from the perspective of the sagittal plane. The right panel
tracks the path from the transverse plane. . . . . . . . . . . . . . . . . . . 42
3.3 Insertion point for ne-wire electrode. Structural MR images of forearm
indicating a hematoma representing the location of a ne-wire electrode
within the extensor carpi radialis brevis muscle. . . . . . . . . . . . . . . . 43
3.4 Spike-triggered average of 840 individual motor unit action potentials in a
ne-wire recording. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.5 Comparison between the calculated transfer function coecients from ne-
wire to electrode 3 (blue trace) and an ideal low-pass lter (orange trace). 45
3.6 Spike-triggered average of ne wire and surface electrodes with approxi-
mate location. Peak of ne wire MUAP used as trigger. White traces
indicate STA. Yellow traces are estimates of STA. Electrode 8 is the STA
of the ne-wire channel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.7 System identication and equalization for use in source localization. . . . 50
4.1 Manipulation tasks performed during electrocorticographic recordings. Nor-
mal forces at the point of contact were recorded using a uni-axial load cell.
From left the right the tasks increase in dexterity demand from static hold
to slow movements and nally to unstable object manipulation. (2) Three-
ngered static grasp of a 300 g object using the thumb, index and middle
ngers. (b) Three-ngered rotation using the 300 g object which was os-
cillated back and forth in a twisting motion at a rate of approximately 1
Hz. (c) Strength-Dexterity test in which the subject compressed a slender
spring as much as possible using a precision pinch. . . . . . . . . . . . . . 56
4.2 Approximate electrode grid layout for Subjects (a) 1 and (b) 2. . . . . . . 57
4.3 ERS and ERD. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.4 Electrocorticographic power associated with dexterity. . . . . . . . . . . . 60
5.1 Two cosine signals and frequency representation with a perfect linear cor-
relation. (a) The blue trace is a pure cosine wave with a frequency of 15
Hz and an amplitude of 1. The red trace is a 15 Hz cosine wave with an
amplitude of 0.5. The correlation between the two signals is = 1. (b)
Frequency domain representation of the cosine signals in (a). The peak
frequency is at 15 Hz for both traces and their magnitude directly relate
to the amplitude of their respective cosine waves. . . . . . . . . . . . . . . 70
xii
5.2 The eects of phase shifting a signal on correlation. (a) The primary sig-
nal (blue trace) is a cosine with an amplitude of 1 and frequency of 15 Hz.
The next three signals share the same frequency but are shifted by =5
(red trace), =2 (yellow trace) and (purple trace), resulting in correla-
tion coecients of = 0:81, = 0 and =1. (b) Frequency domain
representation of the cosine signals in (a). The main trace is represented
in blue with magnitude 1 and the three shifted waves overlap each other
and have magnitude 0.5 at 15 Hz. . . . . . . . . . . . . . . . . . . . . . . . 71
5.3 Eect of frequency component magnitude on correlation. (a) The primary
wave (shown in blue) consists of two frequencies: a 15 Hz component with
unit amplitude and a 250 Hz component with amplitude 0.01. The sec-
ond signal consists of the same frequency components, however the 15 Hz
component has an amplitude of 0.5 and a 250 Hz component of 0.03. The
correlation between the signals is nearly 0 (b) Frequency domain represen-
tation of signals in (a). The 15 Hz components for both signals are much
larger than the 250 Hz components. (c) The two signals have similar fre-
quency components as in (a) however, the 250 Hz components have been
amplied by 100. The correlation is now = 0:7. (d) Frequency spectra
of the signals in (c). The 250 Hz components now dominate the 15 Hz
components. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.4 First three Slepian multitapers. . . . . . . . . . . . . . . . . . . . . . . . . 79
5.5 Eect on the randomization of signal phase on coherence. (a) Two signals
which share frequency components at 5, 12 and 20 Hz are created, each
with phase relationships that vary after each second. (a) The amplitude of
the 5, 12 and 20 Hz components are 5, 12 and 1, respectively. The phase
of the 5 Hz component is randomly varied between18

, the 12 Hz com-
ponent varies by180

and the 20 Hz component contains an unchanging
phase with each passing second. (b) Since the phase of the 5 Hz compo-
nents was bounded within a small range, the coherence at this frequency
remains relatively strong. Because the phase of the 20 Hz component varies
drastically from second to second, the coherence is extremely low. Lastly,
the 20 Hz signal, although it had the smallest amplitude, had the strongest
coherence due to the consistency in the phase throughout the duration of
the signals. Taken from Nunez et al. (1997). . . . . . . . . . . . . . . . . . 85
6.1 Spring used in the quantication of hand dexterity. (a) Typical precision
pinch hand posture used in the Strength-Dexterity test. Endcaps at either
end with a load cell attached to the index nger side of the spring. (b)
Close-up of spring and force sensor next to a ruler. . . . . . . . . . . . . . 95
6.2 (a) Wooden dowel with uni-axial load cell attached. (b) Dowel with load
cell and end caps next to a ruler. . . . . . . . . . . . . . . . . . . . . . . . 96
xiii
6.3 Intrinsic muscles of the hand that were recorded. (a) First dorsal in-
terosseous. (b) Abductor policis brevis. . . . . . . . . . . . . . . . . . . . . 97
6.4 eegosports EEG cap. (a) Front view. (b) Left view. (c) Top view. (d) 2-D
layout of all channels. The ground electrode, AFz, is shown in red and the
reference electrode, CPz, is shown in green. . . . . . . . . . . . . . . . . . 99
6.5 Sample force prole during the Strength-Dexterity test. The average max-
imal compression force for this subject was 2.6 N (red dotted line). As the
spring is compressed, it becomes unstable and dicult to control, resulting
in the subject dropping the spring. These drops are clearly shown as sud-
den decreases in the force prole. 40% and 80% of F
max
were calculated
to be 1.0 N and 2.1 N, respectively, and shown as the purple and green
dotted lines. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.6 Visuomotor paradigm and force proles (a) Spring-low precision pinch task.
Subjects squeezed a small spring between their index nger and thumb.
Normal index nger forces and bipolar surface EMG from several muscles
of the hand were recorded (only FDI shown here). (b) Typical force trace
for a representative subject during the visuomotor task. Black dashed line
is the target force, which, for this subject are 1.0 N and 2.1 N for the low
and high target forces, respectively. Red lines indicate tolerance limits of
15 N. The grey area represents valid hold data. The criteria were that
the force had to be within the tolerance limits and be held within that
range for a minimum of ve seconds. In the last SH condition, it can be
seen that the force fell out of range and thus this data was not included
in the analysis. (c) The spring object is replaced by a wooden dowel. (d)
Force proles for the visuomotor force tracking task with the same force
level performance criteria as in (b). . . . . . . . . . . . . . . . . . . . . . . 108
6.7 Muscle coordination patterns for the FDI and APB across all conditions. 110
xiv
6.8 Scatter plots showing the muscle activation for the FDI and APB for
matched force levels of the objects. The rst principal component for
each condition is shown to capture the direction of the maximum variance.
(a) Rest condition. During the period between the 40% and 80% F
m
ax
target force compressions, the activation the muscles were calculated. The
PCs corresponding to the spring and dowel objects are aligned in this task
demonstrating that the FDI and APB muscle are activated similarly dur-
ing rest. (b) Low force condition. As in the resting condition, the rst
PCs for each object are aligned, but with a slightly lower slope than in
(a). (c) High force condition. During high compression with the dowel,
the activation of the APB muscle is minimal and is dominated by FDI
activity. Compression at the high force of the spring object shows a slop
approximately equal to one, suggesting that there is an equal contribution
from both muscles in order to maintain a constant force on the spring. . . 111
6.9 Results for the dowel-low task. (a) Grand average Z-transformed coherence
head map for the FDI muscle to all EEG electrodes for the DL task. A
99% Bonferonni corrected Z-score threshold was applied to the head map
to account for multiple comparisons based on the number of EEG channels.
The four electrodes with signicant coherence above the threshold were C1,
C3, CP1, and CP3 with respective Z-transformed coherence values of 4.81,
5.26, 4.66, and 4.54. (b) Average coherence spectra for the four electrodes
shown in (a). The beta frequency range (15 - 30 Hz) is shown as the grey
shaded area. Peak coherence for the average was 6.7 at a frequency of
18.15 Hz. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.10 Coherence spectra for the four EEG electrodes with average beta range
coherence values over the threshold limit. In each plot, all four conditions.
The respective condition and trace color are as follows: DL - blue, SH - red,
DH - yellow, and SH - purple. Red dashed line in each gure corresponds
to the 99% Bonferroni corrected threshold value and the grey shaded ar-
eas indicate the beta frequency range (15 - 30 Hz). Individual coherence
spectra for each condition for electrode (a) C3, (b) C1, (c) CP3, and (d)
CP1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6.11 Results of the linear mixed-eects model. The model was constructed
to predict mean beta range coherence using Condition as the xed-eect
and Participant as the random eect. In each bar graph, the mean beta
range CMC is shown on the vertical axis and condition is on the horizontal
axis. Standard error bars are included for each condition and the indicators
above the bars represent the statistical dierence in the linear mixed eects
coecients as determined using an F-test. n.s. indicates that there was no
signicant dierence in eect between two conditions and indicates
that the p-value was less than 0.001. Linear mixed-eect models for the
prediction of (a) FDI-EEG beta coherence and (b) APB-EEG beta coherence.116
xv
6.12 Average beta power for the FDI and APB muscles for the DL, SL, DH
and SH conditions. (a) FDI power for all four conditions. The largest beta
power for the FDI was apparent during the SH task. The power during
the SH was signicantly higher than for all other conditions. (b) . . . . . 117
6.13 Average beta power for the EEG electrode C3 for the DL, SL, DH and SH
conditions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.14 Root mean square error of compression force to target force. . . . . . . . . 120
6.15 Model of cortical drive to hand muscles during stable and unstable tasks.
In the stable domain, ares of the cortex representing the FDI and APB are
driven by underlying neural oscillators. . . . . . . . . . . . . . . . . . . . . 122
6.16 EMG to EMG coherence between the rst dorsal interosseous and the
abductor pollicis brevis. The SL condition is shown as the yellow trace
with a peak coherence of 10.5 at 23.3 Hz. The DL (blue trace) and DH
(red trace) have similar peak coherence values within the beta range at 5.7
and 6.6 at 24.8 and 24.6 Hz, respectively. The SH condition (purple trace)
has the lowest overall beta range coherence with a max value of 2.3 at 24.4
Hz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.17 Grand average FDI-EEG beta coherence head maps during the low force
conditions. (a) DL condition. Peak CMC appears over contralateral M1.
(b) SL condition. Peak coherence exists over contralateral M1 with greater
magnitude than in the DL condition. Coherence extends medially into the
supplementary motor area (i.e. electrodes Cz and FCz). . . . . . . . . . . 125
6.18 Change in gamma coherence across subjects and the rest, low and high
conditions in electrodes C3 (over sensorimotor) and Cz (over SMA). Using
a Wilcoxon rank sum test, it was determined that there was no statistical
dierence in the change in gamma coherence in any of the C3 electrodes
for matched force conditions. However in the SMA, there existed a signif-
icant increase in the average gamma coherence for the spring-high task as
compared to the dowel-high task. . . . . . . . . . . . . . . . . . . . . . . . 127
xvi
Abstract
This dissertation focuses on characterizing cortical involvement during low force dexter-
ous manipulation. Cortical oscillations in the beta frequency range (15 - 30 Hz) are
synchronous with contralateral muscular activity during static precision pinch, indicative
of strong cortico-muscular coupling. However, it is poorly understood how the cortex
modulates the control of ngertip forces during a time-critical dexterous task. The goal
of this research was to examine the functional connectivity between cortex and muscle
during a force tracking precision pinch task using a rigid wooden dowel and a compli-
ant unstable spring at two force levels. At the low force level for both objects and at
the high force level with the dowel, the diculty in maintaining a steady compression
was minimal. However, at the high force level with the unstable spring, the dexterity
requirements to maintain a steady force compression were signicantly more challenging
and required heightened sensorimotor integration. Using this novel paradigm, we showed
that increases in sensory feedback and dexterity demand disrupt consistent descending
commands seen in stable grasps and are re
ected as a reduction in beta corticomuscular
coherence. Despite the fact that the force levels were kept constant for both objects, these
ndings suggest that for precision force control there exist functionally dierent cortical
circuits that are highly dependent on the temporal demands of the task.
xvii
Chapter 1
Introduction
1.1 Background
Human dexterous manipulation is characterized by the ability to precisely control the
forces we exert on everyday objects with our ngers and hands. Throughout the day we
interact with our surroundings in a variety of ways, ranging from large force production
tasks, such as when holding a hammer to strike a nail, to ne manipulation involving low
forces, such as typing, playing the piano or buttoning a shirt button. Fine manipulation
involving the pads of the index nger and thumb is commonly referred to in literature as
a precision pinch. Numerous studies have utilized this paradigm to assess many aspects
of hand function including the examination of forces exerted on an object (Johansson
& Westling 1984, McDonnell, Ridding, Flavel & Miles 2005), muscle strategies using
electromyography (Maier & Hepp-Reymond 1995), the eects of transcranial stimulation
(Davare, Lemon & Olivier 2008), cortical power (Murthy & Fetz 1992), and cortical
synchrony with hand muscles (Baker, Olivier & Lemon 1997) to name a few. While the
1
literature is rich with static precision pinch analysis, this encapsulates a small fraction of
daily hand-object interactions.
This dissertation focuses on characterizing dexterous function by utilizing well es-
tablished electrophysiological recordings and force measurements and introducing the
element of instability. The goal here was to tax the nervous system with a dexterously de-
manding task to dierentiate between how the brain communicates with peripheral hand
muscles when performing simple versus dicult manipulation tasks. Coherence analysis
was used to assess these dierences by identifying cortical areas with known functional
projections onto spinal motor neurons controlling hand muscles and by describing the
preferred frequencies of communication.
1.2 Neural Control of Movement
The sensorimotor process involved in performing a motor task is depicted as a three
step process as shown in Fig. 1.1. During skilled manipulation, tactile information
about the physical properties of the object in hand (e.g., shape, weight, smoothness,
etc.) are relayed to the cortex. The brain incorporates this information and subsequently
provides an appropriate motor command, thereby changing the state of the task. With
this new state, sensory information is updated and the process continues. While this
model provides a general conception of the sensorimotor process, the true mechanisms
involved are far more intricate.
2
Motor
Command
Previous State
State
Context
Sensory
Feedback
Context
Motor
Command
Context
[previous state, motor command, context] state [state, motor command, context] sensory feedback
[task, state, context] motor command
Forward
Dynamic
Model
Forward
Sensory
Model
Inverse
Model
CNS Internal
Representations
Wolpert and Ghahramani (2000). “Computational principles of movement neuroscience. ” Nature Neuroscience.
Figure 1.1: Schematic of the sensorimotor process involved in motor control. Initially,
sensory information from the object being held is obtained. The demands of the task
designate an appropriate motor response to the sensory cortex. Execution of the motor
command causes in a change in the state of the task resulting in new sensory information
and the process continues.
1.2.1 Sensory Processing
The importance of sensory feedback in manipulation cannot be overstated. This is evident
in the large somatotopic representation of the hand in the sensory and motor homunculus.
3
Indeed, the success of homo sapiens is largely due to the co-evolution of cognitive devel-
opment and the ability to use tools (Faisal, Stout, Apel & Bradley 2010). The abundance
of mechanoreceptors in the glabrous (i.e. hairless) skin of the hand which relay critical
aerent sensory information such as pressure, vibration, static touch, and proprioception
(Johansson & Flanagan 2009) emphasize the importance of sensory feedback in object
manipulation.
Mechanoreceptors in the ngers and hands transmit sensory information to dorsal
root neurons in cervical segments C6 and C7 through A neurons. The diameter of these
heavily myelinated group II axons range from 6 - 12 m and permit the transmission
of cutaneous information at conduction velocities of 35 - 75 meters per second (Bear,
Connors, Paradiso, Bear, Connors & Neuroscience 1996), second only in speed to A
neurons for proprioception. Axons from the nuclei in the dorsal root ascend through
the dorsal column medial lemniscus pathway, decussate in the medulla and synapse with
neurons in the thalamus. These neurons then project onto cells in the primary sensory
cortex (S1), secondary sensory cortex (S2) and the posterior parietal cortex.
The methods by which the sensory information is processed is dependent on the spe-
cic goals of the task. For example, a feedforward strategy is applied to reach for an
object based on an internal model (Kawato 1999) and no sensory integration is necessary.
However, when the task requires manipulation, a feedback control strategy must be imple-
mented to integrate sensory information to correct for errors (Desmurget & Grafton 2000).
Neurons of the primary sensory area are known to project to areas involved in voluntary
movement and the planning of movement, namely the primary motor cortex (M1) and
the supplementary motor area (SMA) (Martin 2003).
4
1.2.2 Motor Control
Cortical areas involved in the execution of motor task have been heavily studied. Gross
factors in
uencing the contributions from specic areas include velocity, force and posi-
tion (Jancke, Specht, Mirzazade, Loose, Himmelbach, Lutz & Shah 1998, Deiber, Honda,
Ibaez, Sadato & Hallett 1999, Ashe 1997, Thickbroom, Phillips, Morris, Byrnes & Mastaglia
1998, Humphrey, Schmidt & Thompson 1970). In a functional magnetic resonance
imaging (fMRI) study comparing cortical involvement during power grip versus pre-
cision pinch, the researchers showed greater sensorimotor activity when performing a
power grip compared to a precision pinch. However the precision pinch task showed
higher activation in ventral premotor, posterior parietal and prefrontal cortices (Ehrsson,
Fagergren, Jonsson, Westling, Johansson & Forssberg 2000, KuhtzBuschbeck, Ehrsson
& Forssberg 2001). In a separate fMRI study, it was shown that maintaining a pre-
cision pinch on a small object with just enough force to keep it from slipping acti-
vated the supplementary motor area (SMA), whereas stronger isometric forces showed
no SMA activation (KuhtzBuschbeck et al. 2001, Haller, Chapuis, Gassert, Burdet &
Klarhfer 2009, Galla, de Graaf, Bonnard & Pailhous 2005). Furthermore, self-paced
movements of individual ngers revealed prominent blood
ow into the SMA (Roland,
Larsen, Lassen & Skinhoj 1980). In a positron emission tomography (PET) study, it
has been shown that as the diculty of a task is increased, activation in the premotor,
SMA and caudate nucleus was increased (Winstein, Grafton & Pohl 1997). The inter-
pretation from these studies suggests that gross movements and strong contractions can
be controlled directly from the primary motor area (M1), however, low force and ne
5
precision pinch tasks require the additional engagement of the SMA and premotor cortex
for critical control of ne manipulation and high level motor planning.
The control of hand function is mediated by direct cortical projections via the lateral
corticospinal tract (CST) onto alpha motor neurons in the ventral horn of the spinal
cord. The rst evidence for direct corticospinal projections onto contralateral muscles
was suggested by Bernhard et al. in 1953 who investigated skilled hand tasks performed
by macaque monkeys (Bernhard, Bohm & Petersen 1953). Anatomical studies have shown
that primary motor neurons terminate in the spinal cord (Kuypers 1960, Shinoda, Yokota
& Futami 1981). It is now known that, in addition to M1, axons in the CST originate
from the dorsal and ventral premotor cortices, SMA, and cingulate motor cortex (Dum
& Strick 2005, Dum & Strick 1991). Descending commands from the cortex are highly
task specic and the origin, pathway and synaptic input play a role in the execution of
movement (Lemon 2008).
1.3 Previous Work
1.3.1 Electrophysiological Studies
The production of static and slow oscillatory precision pinch forces upon an object is
a critical aspect of everyday grasping. Several studies have used time series analysis
techniques such as spike-triggered averaging, transcranial magnetic stimulation and cross-
correlation to assess the relationship between cortical activity and the electromyogram
(EMG) (Lemon, Johansson & Westling 1995, Lemon & Mantel 1989, Muir & Lemon
1983). In addition to these techniques, corticomuscular coherence (CMC), has been
6
used to determine cortical projections from M1 onto spinal neurons (Feige, Aertsen &
Kristeva-Feige 2000, Fetz & Cheney 1980, Mima & Hallett 1999, Halliday, Conway, Farmer
& Rosenberg 1998). Simply stated, CMC measures the consistency of the phase lag
between cortical activity (i.e. EEG) and muscular activity (i.e. EMG). The result is
a coherence spectra describing the correlation between the signals for all frequencies of
interest. Studies have associated coherence in distinct frequency bands with specic motor
tasks. For example, CMC in the alpha frequency range (8 - 12 Hz) has been associated
with brief nger movements (Feige et al. 2000, Ohara, Mima, Baba, Ikeda, Kunieda,
Matsumoto, Yamamoto, Matsuhashi, Nagamine & Hirasawa 2001) and coherent sigma
oscillations (12 - 15 Hz) are associated with the startle re
ex (Grosse & Brown 2003).
Perhaps the most studied frequency range for coherence analysis is the beta band
which extends from 15 - 30 Hz. Corticomuscular coherence studies have shown that
this range is associated with static force production (Murthy & Fetz 1992, Baker et al.
1997, Baker 2007, Conway, Halliday, Farmer, Shahani, Maas, Weir & Rosenberg 1995,
Kilner, Baker, Salenius, Hari & Lemon 2000, Kilner, Fisher & Lemon 2004, Kilner, Baker,
Salenius, Jousmki, Hari & Lemon 1999, Kristeva, Patino & Omlor 2007). CMC has been
used to determine the specic cortical areas and frequencies involved in precision pinch
tasks (Kilner et al. 2000, Riddle & Baker 2006, Fisher, Galea, Brown & Lemon 2002, Chen,
Entakli, Bonnard, Berton & De Graaf 2013). It has also been shown that beta CMC in
the primary motor cortex is modulated by digit displacement (Riddle & Baker 2006) and
object compliance (Kilner et al. 2000). Furthermore, these rhythms are absent during
movement (Baker et al. 1997, Brown 2000, Kilner et al. 1999, Kilner et al. 2000, Kilner
et al. 2004, Feige et al. 2000). Signicant coherence in the beta frequency range during
7
sustained muscular contractions suggests that these rhythms are necessary for stability.
One study has investigated beta coherence in the supplementary motor area to show that
CMC was associated with precision control of ngertip forces (Chen et al. 2013).
Only a handful of groups have investigated the synchrony of gamma oscillations (>
30 Hz) between the cortex and musculature. These studies have shown that gamma
CMC is associated with strong muscular contractions (Mima, Simpkins, Oluwatimilehin
& Hallett 1999, Brown, Salenius, Rothwell & Hari 1998, Hari & Salenius 1999). In more
recent investigations, it was shown that during static force production, peak coherence
appeared in the beta frequency range, however, during slow oscillatory force production,
the coherence spectra shifted into the gamma range (Omlor, Patino, Hepp-Reymond &
Kristeva 2007, Patino, Omlor, Chakarov, Hepp-Reymond & Kristeva 2008).
The limitations of these studies is that they focus primarily on static force production
or slow nger movements. As a result, the functional role of M1 and other cortical areas
remain relatively unexplored for time-sensitive dynamic manipulation. Largely due to
the complex nature involved in the control of the hand, few research eorts have explored
the full details of the neuromechanics involved in dynamic force production. Thus, the
current body of literature lacks adequate coherence analysis where unpredictable and
unstable objects are manipulated.
1.3.2 Measuring Dexterous Ability
Given the ease with which we are able to pick up an object, such as a pen and immediately
begin to write, it is easy to underestimate the complex neural strategies involved in
controlling the large number of degrees of freedom in the hand. Literally, there exist
8
an innite number of muscle coordination patterns that can generate the same force
on an object, a problem known as muscle redundancy (Bernstein 1967). The method
used by the nervous system to select the appropriate pattern are currently unknown
but is of signicant interest. Several studies hypothesize that a strategy is selected which
maximizes or minimizes a cost function. Optimization functions that have been employed
include energy minimization (Alexander 1997), maximizing smoothness of movements
(Flash & Hogan 1985) and minimizing torque (Uno, Kawato & Suzuki 1989). Despite the
implementation of the multitude of control strategies in state-of-the-art robotic hands,
even the most advanced designs pale in comparison to the dexterous capabilities of a
toddler.
Multi-nger dexterous manipulation of an object involves rotation, translation, mak-
ing and breaking of contact surfaces, and adjustments to ngertip endpoint force magni-
tude and direction (Valero-Cuevas, Smaby, Venkadesan, Peterson & Wright 2003, West-
ling & Johansson 1984, Loeb, Brown & Cheng 1999). This process requires continuous
sensorimotor integration to update muscle coordination strategies to produce accurate
hand postures and endpoint nger forces necessary to maintain control of an object. Over
the years, many clinical methods have been developed to measure dexterity in humans
including the Box and Blocks (Mathiowetz, Volland, Kashman & Weber 1985), 9-Hole
Peg (Mathiowetz, Weber, Kashman & Volland 1985), Jebsen-Taylor (Jebsen, Taylor, Tri-
eschmann, Trotter & Howard 1969), and Fugl-Meyer (Duncan, Propst & Nelson 1983)
tests. These quantitative measures of dexterous performance suer from a number of
pitfalls: (1) they rely on repetitive tasks which are scored based on time to completion
and accuracy, (2) they involve gross upper body movements, eectively taking the focus
9
away from individual nger movements, and (3) they do not rely on the integration of
critical sensory feedback and can be performed based on visual feedback alone.
Valero-Cuevas et al. (2003) denes dexterity as the ability to dynamically regu-
late endpoint force magnitude and direction (Valero-Cuevas et al. 2003). This concrete
denition led to the development of the Strength-Dexterity (SD) test (Valero-Cuevas
et al. 2003). In this novel precision pinch paradigm, dexterity is measured by asking par-
ticipants to compress a slender spring prone to buckling to the point of maximal stability.
Subject performance is based on their ability to dynamically regulate their endpoint force
direction and magnitude to stabilize the spring throughout compression (Valero-Cuevas
et al. 2003). Strength, the force necessary to bring the spring to solid length (i.e. where
all the coils of the spring are touching), and dexterity, the ability to dynamically regu-
late endpoint force direction and magnitude, are paramount to task performance. Given
the dynamic nature of the SD test, continuous sensorimotor integration is required to
account for variability, small perturbations and increased proprioceptive/sensory feed-
back from the ngertips. fMRI studies during SD spring compression have found that
specic cortical and subcortical areas become active when dexterity demands of the task
increase (Mosier, Lau, Wang, Venkadesan & Valero-Cuevas 2011, Holmstrom, de Man-
zano, Vollmer, Forsman, Valero-Cuevas, Ullen & Forssberg 2011, Talati, Valero-Cuevas
& Hirsch 2005).
In precision pinch paradigms where static forces are applied to rigid objects, the pre-
dominant muscle strategy is that of co-contraction (Smith 1981). In the SD test, however,
the dynamic nature of the task necessitates a more advanced control strategy. Individ-
ual control of the index nger and thumb are required to maintain a constant force on
10
the spring. As a result, the strategy of the muscles controlling the digits operate in
a more fractionated pattern (Schieber & Santello 2004, Bennett & Lemon 1996, Kilner
et al. 1999). It is known that sensory feedback in precision pinch manipulation is crucial
to task performance (Johansson & Westling 1984, Johansson & Flanagan 2009, Westling
& Johansson 1984, Westling & Johansson 1987) and given the dynamic nature of the SD
test, the sensorimotor loop becomes highly susceptible to neural noise and transmission
delays. By taxing the nervous system with a dexterously demanding task, this precision
pinch paradigm makes it possible to push the limits of sensorimotor integration to de-
termine how the nervous system controls for instability. Building upon this work and
by combining high-temporal electrophysiological recordings, we employ a variation of the
SD paradigm to characterize the neural control of the hand in the context of real world
dynamic manipulation.
1.4 Prior Lab Work
A multipurpose data acquisition system was developed to capture force data mea-
sured from compressible springs and transmit the information wirelessly. The device
was designed with versatility in mind and was able to capture data from a multi-
tude of both passive and active sensors (Reyes & Valero-Cuevas 2013). This device
was used in a number of studies including analysis of body movement on a slack
line, during running, walking and cutting, and as a rehabilitation gaming device
for dexterous manipulation and motor control in children with autism spectrum
disorder.
11
We successfully showed that the precise location of a ne-wire recording site could
be detected in a 3T scanner. Furthermore, spike-triggered averaging was used to
develop transfer function models from a ne-wire EMG to surface EMG recordings.
We recorded electrocorticographic (ECoG) data from a few patients undergoing
monitoring for epileptic seizures to investigate the changes in cortical power during
simple and dexterous tasks involving static and dynamic force production as well
as movements.
We have developed an experimental paradigm to investigate spatiotemporal cor-
ticomuscular relationships between rigid and compliant object manipulation with
the hand. Non-invasive EEG was recorded from ve healthy subjects during a vi-
suomotor force tracking task. For all data collections, we simultaneously recorded
endpoint ngertip force and surface EMG from select muscles that control the in-
dex nger and thumb. We utilized signal processing techniques such as coherence
and power spectral density analysis to demonstrate that signicant dierences arise
in the frequency domain between cortical and muscular activity during tasks of
varying degrees of stability.
1.5 Signicance of Research
This research can provide several signicant contributions to the scientic, clinical and
robotics communities. Scientically, the ideas presented here will help formulate a new
denition of dexterity which includes neurophysiological measurements. The spectral
analysis in this dissertation will help further our understanding of how the brain utilizes
12
specic cortical oscillations to integrate sensory feedback with motor control. From this,
it would be possible to determine how and when the brain becomes dissociated during
complex movements and assigns the task to subcortical and/or spinal circuits. From a
clinical perspective, corticomuscular coherence analysis could provide clinicians with non-
invasive methods for the early detection of nerve pathologies and neuromuscular disorders
aecting dexterous performance and assist in tracking rehabilitation progress. In the
robotics eld, this research could be used in the development of hierarchical controllers
which track synchronous cortical and muscular oscillations to send commands to robotic
and prosthetic limbs based on the desired intent of the user.
This dissertation provides a novel analysis of cortical drive to contralateral hand mus-
cles during dexterous manipulation. Many hand-object interactions throughout the day
involve dynamic movements and the readjustment of ngertip forces, however these com-
plex interactions are rarely addressed in literature. The Strength-Dexterity paradigm
oers a novel method of investigating cortical involvement during dynamic dexterous
tasks. Unfortunately, the investigations utilizing the SD test have been performed under
the temporal constraints of functional MRI. The methods presented here utilize a vari-
ation of the SD paradigm and bypass the temporal limitations of fMRI by using EEG
to characterize the frequency content involved in brain-body communication. Further-
more, this research expands on the corticomuscular literature to incorporate instability
and unpredictability in manipulation tasks in an eort to move away from static grasp
analysis. The results presented here begin to address how the sensorimotor system uti-
lizes time-sensitive tactile feedback in order to accurately control the muscles of the hand
13
during dexterous manipulation and helps to further our understanding of how cortical
bandwidth is utilized in strategic motor tasks.
1.6 Dissertation Outline
1.6.1 Chapter 2
This chapter presents a custom designed low-cost wireless data acquisition system with
user-congurable settings. Originally, the device was designed to capture force data
during the Strength-Dexterity test, however several applications are discussed to highlight
the devices' wireless capabilities as well as the versatility in acquiring data from several
types of active and passive sensors. This work was presented at the 35th Annual American
Society for Biomechanics Conference in 2011 and as a podium presentation at the IEEE
EMBS Special Topic Conference on Point-of-Care Healthcare Technologies in 2012. Dr.
Francisco J. Valero-Cuevas is a co-author.
1.6.2 Chapter 3
This chapter discusses the ability to detect a micro-hematoma formation made during ne-
wire electrode placement in a structural MRI. Additionally, transfer function models were
developed to relate the propagation of individual motor unit action potentials recorded
with a ne-wire electrode to an array of surface electrodes. Part of this work was presented
at the 6th International IEEE/EMBS Conference of Neural Engineering in 2013. Dr.
Krishna Nayak, Dr. Gerald Loeb and Dr. Francisco J. Valero-Cuevas are co-authors.
14
1.6.3 Chapter 4
This chapter investigates dierences in cortical power in epilepsy patients during tasks
of varying diculty. This pilot work was done in collaboration with the Department of
Neurology at the Keck School of Medicine of USC. Part of this work was presented at
the 43rd Annual Meeting of the Society for Neuroscience in 2013 and at the 7th World
Congress of Biomechanics in 2014. Emily L Lawrence, Sarine Babikian, Dr. Christianne
Heck, Dr. Charles Liu and Dr. Francisco J. Valero-Cuevas are co-authors.
1.6.4 Chapter 5
Chapter 5 provides a review of the functional signicance of synchronous oscillations
in the cortex and in the muscle. The calculation and interpretation of coherence are
discussed.
1.6.5 Chapter 6
This chapter examines the eects of modulation of synchronous cortico-muscular oscil-
lations in dexterous tasks. We explore the role of cortical areas other than the primary
motor cortex during dexterous manipulation and extends the analysis of corticomuscular
coherence into the gamma frequency range and the supplementary motor area. This work
was done in collaboration with the Applied Mathematical Physiology Lab (AMPL). This
work was done in collaboration with the Applied Mathematical Physiology Lab (AMPL).
Dr. Christopher M. Laine, Dr. Jason J. Kutch and Dr. Francisco J. Valero-Cuevas are
co-authors.
15
1.6.6 Chapter 7
Chapter 7 discusses the future direction of this research.
16
Chapter 2
Data Acquisition Box
Abstract
As electronic components become smaller and cheaper with each passing year, wearable
technology is rapidly becoming realizable. The ability to wirelessly transmit data from
wearable sensors allows for individuals to perform complex movements and tasks without
being tethered to a computer or conned to an indoor laboratory setting. In this paper,
we describe a novel wireless data acquisition box (DAB) capable of transmitting data from
up to 16 unique sensors to a computer with built-in Bluetooth capabilities. Sample rates,
signal gains and lter cuto frequencies can be preprogrammed as per the requirements of
the study. We have been able to receive stable transmissions from over 100 ft. away even
as the wearer performs rapid movements. Furthermore, the device is compact enough
to t inside a pocket and has the ability to support both passive and low-power active
sensors. Data has been successfully collected with this system on several occasions, a few
of which are mentioned here.
17
2.1 Introduction
The use of wearable wireless systems are an invaluable asset in the laboratory setting as
they allow for the acquisition of biological and kinematic data. However, many commer-
cially available systems are typically expensive, require the use of a USB receiver and
cannot be used in harsh outdoor environments. Furthermore, many Bluetooth systems
cater to a proprietary or specic type or brand of sensor and are application specic, such
as ECG monitoring (Andreasson, Ekstrom, Fard, Castao & Johnson 2002, Proulx, Clif-
ford, Sorensen, Lee & Archibald 2006), glucose monitoring and EMG systems. Current
wireless systems which allow for the acquisition from dierent types of sensors are often
limited to passive sensors. Without a doubt, sensor versatility, long range acquisition and
aordability are paramount for research purposes.
To our knowledge, there does not currently exist a commercially available device which
can sample from several sensors at satisfactory sampling rates using Bluetooth technology
(Cosmanescu, Miller, Magno, Ahmed & Kremenic 2006). The wireless acquisition system
described here circumvents these shortcomings by giving experimentalists the freedom to
collect data from active and passive sensors while also incorporating a novel encoding
algorithm to maximize sampling rates. We have employed o-the-shelf electronic compo-
nents to bring low-cost technology to bring versatility and aordability to the laboratory
setting. In addition to its wireless capabilities and ease-of-use, the device is small enough
to t into the pocket of the user, giving participants the freedom to explore their en-
vironment while data from up to 16 sensors are telemetered to a computer. This rst
18
implementation of the device works in tandem with a custom designed graphical user
interface giving testers visual conrmation of data quality.
2.2 Methods
2.2.1 Telemetry Details
The data acquisition box (DAB), shown in Fig. 2.1, supports inputs from up to 16 types
of sensors including accelerometers, load cells, temperature transducers, etc. Interfacing
connectors provide +3.3 V and +5.0 V to power active sensors. Three of the DAB in-
puts have been congured in dierential mode while the remaining 13 are single-ended.
Dierential inputs reject common noise with a CMRR of 105 dB before conversion to
single-ended inputs. Data are then amplied, ltered and sequentially sampled at an
adjustable rate. 12-bit analog-to-digital conversion is achieved with a peripheral inter-
face controller (PIC) microcontroller (Microchip, Chandler, AZ) operating at 30 million
instructions per second (MIPS). Each sample is prepended by a 4-bit address to ensure
proper decoding by the receiving algorithm. The DAB employs Universal Asynchronous
Receiver/Transmitter (UART) protocol to transmit data and address as two 8-bit pack-
ets. Each data packet is enclosed within a start (low) and stop (high) bit. These bits
indicate to the receiver when a valid data packet is being transmitted. Data that do not
contain the correct transmission sequence are discarded. The result is a dual packet serial
transmission of 20 bits for single sample and is shown in Table 2.1.
19
Connector:
13-pins
5 devices per connector
Easily accessible
Bluetooth Module:
RN-41
921,600 Bps transmission rate
Communication using UART
Frequency hopping for security
Microcontroller:
dsPIC30F5011
16 analog data inputs
64-pin TQFP
10 mm x 10 mm
30 MHz operating
frequency
Programming connector:
Quick software updates
ON/OFF Switch:
Simply ip on to start
collecting data

Status LEDs:
Green - ON
Red - Low Battery

Board dimensions:
1.8” x 2.2”
Figure 2.1: Circuit board layout for Data Acquisition Box (DAB) with individual com-
ponents and description of features. The circuit board measures 1.8 x 2.2 inches.
Although the PIC microcontroller can sample data at high frequencies, the maximal
sampling rate per channel is limited by the transmission rate of the Class 1 Bluetooth
Module (Roving Networks, Los Gatos, CA), which can achieve stable transmission rates
up to 921,600 bps. Given this, the calculated maximal sampling frequency for a channel
is Fs
max
= 46080=N Hz/channel, where N is the number of channels in use. As a result,
when only a single channel is in use, the maximal sampling rate is approximately 46 kHz.
When all 16 channels are in use, the DAB can sample at 2.88 kHz per channel. While
20
these are the maximal sampling rates that can be achieved, the user is able to sample at
lower sampling rates.
Byte 1
start d8 d9 d10 d11 a0 a1 a2 a3 stop
Byte 2
start d0 d1 d2 d3 d4 d5 d6 d7 stop
Table 2.1: Serial communication arrangement for transmitted data. Start and stop bits
indicate to the receiver when valid data have arrived. Four address bits are used to
indicate channel number and 12 data bits correspond to sensor voltage.
2.2.2 Receiving Data
Since many laptop computers and portable devices come with built-in Bluetooth connec-
tivity, we have eliminated the need for receiving hardware (such as a receiving antenna),
which may take up a USB port and require the installation of drivers. We take advan-
tage of the widespread use, ease of connectivity and low power consumption associated
with Bluetooth technology, making this the ideal choice for the DAB over other wireless
standards such as WiFi and RF. Once the device has been paired with the host com-
puter, a virtual communication port is created through which all data are transmitted
and received. Using a custom Matlab GUI, users simply push a start and stop button to
begin and end data collection. This feature allows users to stream data for an indenite
amount of time, limited only by disk space and battery life.
When the start button is pushed, the algorithm opens Realterm (Broadcast Equip-
ment Ltd., Auckland, New Zealand), an open-source hyper terminal program, to establish
the serial communication link. A request for data is sent from the computer to wake the
device up from a power-saving sleep mode to begin transmission. After the user has
21
collected data for the desired time, communication terminates and the device re-enters
sleep mode.
For this initial design, incoming data are not displayed in real time, but are instead
written to a text le. This was done to reduce computational demands on the receiving
unit and minimize potential data bus and CPU con
icts across platforms. Once transmis-
sion has completed, the program opens the text le, decodes the data using the address
bits and populates the data matrix according to the appropriate channel. The time re-
quired to decode the data stream will depend on the recording length and CPU speed.
All data are saved as .mat les with subject identication and trial number stored for
recall when necessary in the Matlab GUI.
2.2.3 Additional Features
The printed circuit board (PCB) is contained within a custom wearable rectangular enclo-
sure made from acrylonitrile butadiene styrene (ABS) material and manufactured using
a 3-D printing machine. The entire device measures 5.08 cm 5.84 cm 1.27 cm and
weighs less than 100 g, making it smaller and lighter than an average cell phone. It uses
a standard low-prole +3.7 V Li-ion rechargeable battery with a capacity of 1000 mAh.
The battery is recharged by an external wall charger and takes approximately 30 minutes
to complete. Future versions of the device will incorporate a recharging circuit to elim-
inate the need to remove the battery. Noise and crosstalk on the PCB were minimized
by optimizing component placement to create direct signal traces and by incorporating
a ground plane layer. Two onboard LEDs indicate status: green when the device is on
and red when the battery is low and needs to be recharged. A red LED on the Bluetooth
22
module lets users know that there is no communication between the device and computer.
When communication is established, the red
ashing LED ceases and a second greed LED
turns on and remains illuminated for the duration of the data transmission. Preliminary
tests show that the DAB can receive data from up to 100 ft. away and with a few hours
of use every day, the device can operate for two weeks without the need to recharge the
battery.
2.2.4 Sensors
Here we describe two types of sensors that have been congured for use by the data
acquisition box. To measure ngertip forces exerted upon hand held objects, we used a
strain gauge based single-axis load cell (Measurement Specialties, Hampton, VA) capable
of measuring up to 10 lbs. of force. This sensor uses a Wheatstone bridge conguration
to measure changes in resistance and produces a dierential output proportional to the
applied normal forces. The circular load cell measures 1.27 cm in diameter and 0.41 cm
in height, requires a +5:0 V supply and draws 2.0 mA of current.
For movement analysis, we used a custom designed low-prole tri-axial accelerometer
(STMicroelectronics, Geneva, Switzerland) mounted on a custom designed printed circuit
board measuring 15 mm in diameter. The miniature accelerometer measures 4 mm
4 mm 1 mm, has a detection range of2 g, sensitivity of about 1 V/g, powered by
either +3:3 V or +5.0 V, and requires approximately 850 A of current. The load cell
and accelerometer in comparison to a U.S. penny are shown in Fig. 2.2.
23
Figure 2.2: Uni-directional load cell and ti-axial accelerometer next to a US penny for
size comparison.
2.3 Results
We rst tested the sensitivity and reliability of this device by mounting the tri-axial
accelerometer on a sliding tray that moves along a single axis. The sensor was attached
using tape and moved back and forth causing the sensor to experience acceleration in a
single direction. After a few trial movements, the accelerometer was remounted so that
the movement direction aligned with a dierent acceleration axis. This procedure was
repeated for all three axes.
Fig. 2.3 shows data collected from an accelerometer mounted on a moving tray at a
sample rate of 400 Hz. Initially, the direction of the movement was aligned with the x-axis
of the accelerometer, then moved to align with the y-axis and nally the z-axis. The top,
middle and bottom graphs correspond to accelerations along the x, y and z directions,
respectively. As expected, the output of the accelerometer is sensitive only to movements
along the aligned axis while the other channels produce no activity. At approximately
21 and 42 seconds, the accelerometer was repositioned, causing the sensor to experience
24
accelerations on all channels. The results obtained in this test illustrate that the device
can accurately collect rapid dynamical data from several channels without cross talk.
x-axis acceleration
Sensor Voltage (V)
-0.1
0
0.1
y-axis acceleration
Time (s)
10 20 30 40 50 60
z-axis acceleration
-0.1
0
0.1
-0.1
0
0.1
Figure 2.3: Sensitivity of accelerometer to detect accelerations in three perpendicular
directions.
Next, we captured ngertip force dynamics during a sensorimotor tasks that measures
strength and dexterity during manipulation of a small deformable spring (Valero-Cuevas
et al. 2003). The device records from force and acceleration sensors attached to either
end of a compressible spring using double-sided tape, as shown in Fig. 2.4. Subjects
were asked to squeeze the spring between their thumb and index nger in an attempt to
compress it fully without it buckling. Subjects were asked to curl in the remaining three
ngers to avoid any contact with the spring. Once they reached a comfortable level of
compression, they were asked to maintain that posture for a few seconds.
25
Figure 2.4: Compressible spring with load cells and accelerometers attached at either
end.
Fig. 2.5 shows a sample of data collected from two load cells and two tri-axial ac-
celerometers, a total of 8 channels, sampled at 400 Hz for 30 seconds. The top gure
shows normal forces applied to a compressible spring squeezed between the thumb and
index nger. Sudden drops in the sensor voltage indicate when the spring buckled and
nger contact was lost. The bottom gure shows the Euclidean norm of the x, y and z
accelerations from accelerometers placed on the ends of the spring. This was performed
to show two representative acceleration traces rather than all six channels.
26
Force (N)
0.5
1
1.5
2
Finger Forces
Time (s)
0 5 10 15 20 25 30
Sensor Voltage (V)
-0.1
-0.05
0
0.05
0.1
Finger Accelerations
Figure 2.5: Top trace: Force proles for index nger and thumb during spring compres-
sion. Bottom trace: Euclidean norm of index nger and thumb accelerations.
Analysis of the force data can be used to obtain the maximal compression force and the
sustained compression, dened as the period in which the compression force is bounded
by one standard deviation of the mean force (Dayanidhi, Hedberg, Valero-Cuevas &
Forssberg 2013). Using these metrics, a dexterity score can be given to the subject based
on their performance (Valero-Cuevas et al. 2003).
To showcase the wearability of the device, ve tri-axial accelerometers were attached
to a subject as they ran, walked, made cutting maneuvers, and jumped in an open eld.
Accelerometers were attached using medical tape to the knees, ankles and trunk as shown
in Fig. 2.6.
27
Figure 2.6: Subject wearing ve accelerometers attached to the thighs, trunk and ankles.
Fig. 2.7 shows ten seconds of acceleration data captured from a subject who started
from rest and then began to jog. Data from 15 channels were captured at a rate of 300
Hz from the subject to show the kinematics involved during a light jog. The time series
data show that each channel is independent of the others and there are no indications
of erroneous decoding or lost data while collecting from multiple sensors. During our
recording session, we were able to receive reliable data transmissions at distances up to
15 m, giving an active collection area of approximately 706 m
2
. All data were collected
using a PC running Windows 7 and Realterm 2.0.0.70.
28
Time (s)
0 5 10
Acceleration (m/s
2
)
-30
-20
-10
0
10
20
30
Left Thigh
Time (s)
0 5 10
Acceleration (m/s
2
)
-20
-10
0
10
20
Right Thigh
Time (s)
0 5 10
Acceleration (m/s
2
)
-20
-10
0
10
20
Trunk
Time (s)
0 2 4 6 8 10
Acceleration (m/s
2
)
-30
-20
-10
0
10
20
Left Ankle
Time (s)
0 2 4 6 8 10
Acceleration (m/s
2
)
-20
-10
0
10
20
Right Ankle
a
x
a
y
a
z
Figure 2.7: Five accelerometers attached to the thighs, trunk and ankles of a subject as
they start from rest then perform a light jog.
By analyzing the sensor data during these exercise and movement tasks, it will be
possible to compare kinematic data between healthy subjects and individuals at varying
stages of rehabilitation from anterior cruciate ligament (ACL) injuries. Deviations from
healthy kinematic data will provide physical therapists with a measurement tool to help
them gauge the severity of the injury and allow them to develop customized rehabilitation
programs for ACL patients.
Lastly, the device was used as a gaming controller to promote gross motor skill de-
velopment in children with autism spectrum disorder (ASD). The device was attached to
the trunk of the child using an adjustable belt and had an accelerometer attached which
29
would mimic the child's jumping actions to control an on-screen character's vertical posi-
tion. The side-scrolling game consisted of a character whose goal was to avoid obstacles
by jumping over them.
During the prototyping phase of the game, the acceleration captured from the DAB
was integrated twice to yield position data. This was then compared to a Kinect gaming
system to determine if it was possible to obtain accurate position data from our custom
designed accelerometers and DAB. In this test, the subject performed gradually larger
jumps within a 10 second interval. Figure 2.8 shows the vertical acceleration, velocity and
position of the trunk sensor obtained through the integration of the acceleration captured
by the DAB. In the last panel of the gure, the position data captured from the Kinect
system is plotted to show the comparison of the true vertical position against our twice
integrated acceleration. The oset in the trunk trace of the Kinect data arises from the
fact that the Kinect system uses the center hip marker as the datum (i.e. at rest, the
trunk marker is approximately 0.15 m above the hip).
30
Acceleration (m/s
2
)
-20
-10
0
10
Trunk Acceleration
Velocity (m/s)
-1
0
1
2
Trunk Velocity
Time (s)
0 2 4 6 8 10 12 14 16 18 20
Position (m)
-0.2
0
0.2
0.4
Trunk Position
Accelerometer
Kinect
Figure 2.8: A comparison of the position data of the Kinect system to the integrated
accelerometer recordings.
31
2.4 Discussion
The multipurpose wireless data acquisition box presented here is congurable, wearable,
robust, low cost, and easy to use. The device can be tailored to collect analog data from
a wide variety of active and passive sensors, giving it the potential for use in countless
laboratory, clinical and industrial applications. With this device, subjects are permitted
to freely move around in their environment while data is reliably transmitting and dis-
playing data on a user-friendly GUI. Aside from being a valuable laboratory device, the
system was also designed with clinical and in-home rehabilitation and monitoring appli-
cations in mind. Given the economical pricing of the wireless transmitter, we hope to
distribute these devices to parents and their children in need of physical therapy for hand
function and gross motor skills. Our in-home rehabilitation comes in the form of custom
designed video games that use the device and attached sensors as the controlling input.
We believe this approach will serve as a fun and eective motivational tool for children
to engage in daily rehabilitation. Currently, the system is undergoing several updates
to the hardware, GUI and rmware to make the device integratabtle, user-friendly and
fully congurable. The device, PCB and enclosure will be upgraded by adding strain
relief to the connectors and will incorporate a rechargeable battery circuit. Additionally,
the device will undergo a complete redesign of the enclosure to make it more robust and
stylish. Firmware will be updated to reduce power consumption for the PIC to give the
device a longer battery life. Lastly, the GUI software will be updated to present the data
in real time and we will work towards making the device compatible with cell phones,
tablets and other Bluetooth enabled devices.
32
Acknowledgments
This work is funded in part by grants EFRI-COPN 0836042 and OPTT-RERC 84-
133E2008-8 to FVC. The author would also like to thank Gary Lin for his assistance
in the development of the rmware for data collection and Brendan Holt for his work on
the interface and gaming design for the DAB.
33
Chapter 3
Localization of a Fine-wire Recording Site and its
Propagation Characteristics
Abstract
The ability to extract information about muscle activation and control at the scale of
individual motor units from surface EMG depends critically on the spatial distribution
of the electrodes with respect to the source within the muscle. Despite this importance,
few studies have used the precise location of an intramuscular recording site to relate
frequency characteristics, such as ltering eects, to those of non-invasive surface record-
ings. In this study, we recorded from a single ne-wire EMG electrode inserted into the
extensor carpi radialis brevis (ECRB) of the right forearm arm of a single volunteer along
with seven surface electrodes placed equidistantly around the proximal forearm in the
same transverse plane as the ne-wire electrode. Following a series of hand and nger
manipulation tasks, the ne-wire electrode was removed and each surface electrode was
replaced with a ducial marker. We then acquired structural images of the forearm of
the subject with a 3T scanner. The objectives of this study were two-fold: (1) show that,
34
with a structural MRI, it is possible to locate the exact ne-wire recording site based on
the presence of a micro-hematoma formed during needle insertion and (2) characterize
transfer function characteristics from ne-wire to surface electrodes using spike-triggered
averaging (STA) of individual motor unit action potentials (MUAPs). Our ndings in-
dicate hematoma formations showing the path of the needle as it moved deeper and
proximally into the muscle of interest. Secondly, we developed input-to-output charac-
teristics and extracted hints of MUAPs buried in the noise
oor which would otherwise
be overlooked. These results pave the way for the optimization of system identication
models, validation of volume conduction models and may lead to methods of solving the
inverse problem of source localization.
3.1 Introduction
Intramuscular recordings have long served as the gold standard for obtaining muscular
activity from individual motor units (MUs). Fine-wire and needle electrodes have been
used in numerous human and animals studies for use in understanding biokinesiology
during movement and force production (Cianchetti & Valero-Cuevas 2010, Burgar, Valero-
Cuevas & Hentz 1997), assisting in the clinical diagnosis of neuromuscular diseases (Oaube
1991), and the assessment of motor unit conduction velocities (Farina, Arendt-Nielsen,
Merletti & Graven-Nielsen 2002) to name a few.
Despite the numerous advantages of intramuscular recordings, many pitfalls exist.
Fine-wire recordings are obtained using a hypodermic needle that pierces the skin to insert
either one or two thin-gauged stripped wires in the muscle of interest. Once placement
35
has been carefully validated through either stimulation or visual inspection of the EMG
signal, the hypodermic needle is removed leaving the wires embedded within the muscle.
Placement of needle electrodes is a similar procedure, however the needle remains within
the muscle throughout the duration of the study. Certication is required for insertion
of intramuscular electrodes in human subjects. Potential complications associated with
intramuscular EMG recordings include muscle spasms, numbness, soreness and may cause
the subject to feel faint (Jonsson, Omfeldt & Rundgren 1967).
Surface EMG (sEMG) oers a non-invasive alternative to intramuscular recordings
at the cost of limited bandwidth. Bandwidth for sEMG ranges up to approximately 600
Hz whereas invasive techniques range up to 1.0 kHz (Rash & Quesada 2003). Studies
have investigated the relationship between sEMG to those of ne wire (Rajaratnam 2014,
Semciw, Neate & Pizzari 2014), however the experimentalist is never precisely sure of the
exact location of the intramuscular electrode recording site nor the spatial relationship
between the target muscle and the surface electrode(s).
An action potential in a motor neuron depolarizes the sarcolemma in all the muscle
bers that it innervates. The action potential enters the innervation zone and travels along
the muscle in both direction. The interference pattern of the depolarized muscle cells make
up the electromyogram. Many factors in
uence the quality of the surface EMG, includ-
ing electrode design, contact impedance and electrode placement (Fuglevand, Winter,
Patla & Stashuk 1992). However, even under ideal conditions, one is still faced with the
problem of degraded signal-to-noise ratio caused by the spatial low-pass ltering eects
of tissues and signal attenuation from muscle to electrode. Understanding the spatio-
temporal spectral relationship that exists between intramuscular and surface electrode
36
recordings would serve as an invaluable tool for describing muscle activation within a vol-
ume conductor. To date, signal propagation has often been described with computational
models using symmetrical congurations and estimating electromagnetic propagation pa-
rameters (Plonsey 1974, Farina, Mesin, Martina & Merletti 2004, Farina 2001, Lowery,
Stoykov, Ta
ove & Kuiken 2002, Gootzen, Stegeman & Van Oosterom 1991, Roeleveld,
Blok, Stegeman & Van Oosterom 1997). By knowing the exact location of an intramus-
cular ne-wire recording site within the muscles of the forearm, we can maximize the
information obtained from spatially known surface EMG recordings (Farina 2001).
To our knowledge, there have been no reports which accurately localize an intramus-
cular recording site within a muscle belly in relation to sEMG sites. We present here, for
the rst time, the detection of a ne-wire recording site from a structural MRI from the
formation of a hematoma created during needle insertion. Secondly, we relate both spatial
distances and signal propagation characteristics from the intramuscular recording site to
generate accurate transfer functions. These results will be used to fully characterize the
electrical parameters of biological tissues to experimentally validate volume conduction
models.
3.2 Methods
3.2.1 Ethics
We collected data from a right-handed consenting volunteer (male, 28 years old). The
participant had no known history of neurological conditions and had no history of hand
37
surgery. All aspects of this study were approved by the Institutional Review Board (IRB)
at the University of Southern California.
3.2.2 Recordings
A pair of ne-wire electrodes and seven bipolar surface electrodes were used to record
EMG from the right forearm of a single consenting volunteer (male, 28 years old) who
performed isometric and dynamic force generation manipulation tasks with the hand,
wrist and ngers. The seven bipolar surface EMG electrodes (Delsys, Boston, MA) were
placed circumferentially around the proximal third at the widest part of the right forearm.
The electrodes were arranged equidistantly with no particular objective muscle in mind.
Each electrode was labelled with a number from one to seven for reference. A large ground
electrode was placed on the right olecranon. The ne-wire electrode was inserted using a
26-gauge needle into the belly of the extensor carpi radialis brevis (ECRB) muscle between
surface electrodes 3 and 4 at an angle of approximately 25

relative to the surface of the
arm. The target muscle was conrmed using structural landmarks, muscle palpation
and visual inspection of EMG response. Figure 3.1 shows the placement of the surface
electrodes and Fig. 3.1a shows the insertion point for the hypodermic needle used to
place the ne-wire electrodes. All EMG recordings were captured at a sample rate of 4
kHz.
38
(a) (b) (c)
Figure 3.1: Placement of EMG electrodes around forearm. (a) Point of insertion for
ne-wire electrode. Surface electrodes 2 - 4 can be seen on the posterior-lateral forearm.
(b) Surface electrodes 4 - 6 on the anterior-lateral forearm. (c) Surface electrodes 6 and
7 on the anterior forearm.
3.2.3 Spike-Triggered Averaging
Motor unit action potentials represent the basic driving force for producing full scale mus-
cular contractions. Fine-wire recordings often record from either one or a few individual
units. Here, we investigated synchronous time-locked surface EMG activity centered
on the peak of each ne-wire MUAP using spike-triggered averaging (STA) (Murthy &
Fetz 1996b). In general, the STA can be written as
STA () =
1
N
N
X
i=1
x (t
i
) (3.1)
39
whereN is the total number of spikes to average, t
i
is the temporal location of event
i for signal x (t) and is the window size for the average. STA was calculated using the
peak height of a detected MUAP over a 30 ms window (5 ms before and 25 ms after
peak). EMG data were notch ltered at 60 Hz and high pass ltered from from 2 Hz to
remove any slow trends in the data.
To describe the characteristic of signal propagation in biological tissue, a transfer
function model was estimated using a 50th order fast transversal least means square
(LMS) adaptive lter (Slock & Kailath 1991). For a sample frequency of 4 kHz, a fty-
tap FIR lter will include latencies up to 12.5 ms, allowing deep MUAPs to propagate to
surface electrodes.
3.2.4 Imaging of a Hematoma
To determine if it would be possible to image the exact location of the ne-wire recording
site, we had made preparations to capture structural MR images immediately following
the manipulation portion of the study. Within 30 minutes, the ne-wire electrodes were
removed and each surface electrode was replaced by an MR compatible ducial marker
(i.e. vitamin E tablet). We obtained structural MR images of the subject's forearm in
a clinical 3T scanner (GE Healthcare, Waukesha, WI) with the arm fully pronated to
maintain the same posture and muscle-electrode relationship as during the experiment.
The scanner was congured to acquire fast HD 3D spoiled gradient recalled (SPGR) with
1.5 mm slices.
40
3.3 Results
3.3.1 MR Images
Figure 3.2 shows the progression of the hematoma formation through the muscular tissue
as shown by the blue arrows. As the hypodermic needle pierced the skin during ne-wire
electrode insertion, small hematoma formations appear along the path. In the scanner,
these small blood formations are revealed as dark spots in the MR scan. The left panel
shows the hematoma view from the coronal plane moving in the dorsal to volar direction.
The right panel shows the progression of the hematoma as viewed in the transverse plane
moving deeper into the forearm. Bright dots around the forearm correspond to the ducial
markers representing the locations of the surface electrodes.
41
2
4
2
4
2
4
2
4
2
4
2
4
3
2
1
7 6
5
4
R
U
3
2
1
7 6
5
4
R
U
3
2
1
7 6
5
4
R
U
3
2
1
7 6
5
4
R
U
3
2
1
7 6
5
4
R
U
10mm
3
2
1
7 6
5
4
R
U
Distal
Proximal
Dorsal
Volar
Left Right
Figure 3.2: Path of hypodermic ned used to insert ne-wire electrodes. The left panel
shows the path from the perspective of the sagittal plane. The right panel tracks the
path from the transverse plane.
42
Figure 3.3 shows the sagittal, coronal, and transverse planes that best represent the
nal detectable location of the micro-hematoma. From the series of images, it was de-
termined that the needle reached a depth of approximately 8 mm into the tissue. These
measurements were consistent with measured depth of the needle and video taken during
the electrode placement. Cross-validation with anatomical landmarks and similar scans
containing clearly labeled muscle groups further conrmed that the micro-hematoma was
in the ECRB muscle.
1
2
3
4
5
6 7
R
U
2
4
Sagittal Plane Coronal Plane Transverse Plane
10mm
Distal
Proximal
Figure 3.3: Insertion point for ne-wire electrode. Structural MR images of forearm indi-
cating a hematoma representing the location of a ne-wire electrode within the extensor
carpi radialis brevis muscle.
3.3.2 STA
Following strong muscular contractions during the dexterous tasks, a motor unit action
potential (MUAP) train appeared in the intramuscular recording, however there was no
visible movement in the hand or wrist. Figure 3.4 shows the spike-triggered average of
840 MUAP events triggered on the peak of the waveform from the ne-wire recording.
43
Nine separate MUAP bursts appeared throughout the data and consisted of a smaller
MUAP with a peak of 0.12 V and a larger MUAP with a peak voltage of 0.4V. The
peak detection algorithm to capture the spike times was adjusted to capture both sets of
MUAPs.
Time (ms)
-5 0 5 10 15 20 25
EMG (V)
-0.2
-0.1
0
0.1
0.2
STA of MUAP
N = 840
Figure 3.4: Spike-triggered average of 840 individual motor unit action potentials in a
ne-wire recording.
Figure 3.5 shows an example transfer function calculated using the LMS method from
the ne-wire recording to surface electrode 3 and is represented as the blue discretized
trace. Here, the coecients were normalized to have a maximum value of 1 and the entire
trace was inverted from its true transfer function. Inversion of the trace does not aect the
cuto frequency, it merely changes the sign of the resulting waveform. The orange trace
in Fig. 3.5 is a 50th-order sinc function centered at 25. Convolving a sinc function with
a time domain signal, x (t), eectively low-pass (LP) lters the signal. For an innitely
long sinc function, the frequency domain transform yields a rectangular waveform (i.e.
an ideal low-pass lter) as shown in the following Fourier transform pair
44
rect

t

! sinc

!
2

(3.2)
where is the width of the rectangle function, and 2= is the cuto frequency.
The similarity in shape of these signals demonstrates that the transfer function for this
input-to-output relationship assumes the general form of a LP lter.
Coefficient Number
5 10 15 20 25 30 35 40 45 50
Normalized Filter Coefficient
-1
-0.5
0
0.5
1
LMS TF for Electrode 3
Ideal LPF TF
Figure 3.5: Comparison between the calculated transfer function coecients from ne-
wire to electrode 3 (blue trace) and an ideal low-pass lter (orange trace).
45
Table 3.1 describes the relative locations of the surface and ne-wire electrodes accord-
ing to the scanner coordinate system. The last column of Table 3.1 indicates the muscle(s)
directly under the surface electrodes as determined using anatomical landmarks. Elec-
trodes 1 through 7 are the surface electrodes labelled in 3.1 and electrode 8 is the ne-wire
recording. The proximity of the ne-wire recording sight to surface electrode 3 and the
fact that the surface electrode was at least in part recording from the ECRB helps to
explain the relative cleanliness of the transfer function estimation.
Electrode
Coordinates (mm)
Muscle
x y z
1 -87.36 -41.92 43.21
exor digitorum profundus (FDP)
2 -78.58 -7.76 25.14
extensor digiti minimi (EDM)
and extensor carpi ulnaris (ECU)
extensor carpi radialis brevis (ECRB)
3 -54.25 -0.30 6.91
and extensor digitorum (ED)
4 -24.57 -10.03 -2.74 brachioradialis (BR)
5 -17.48 -38.62 -14.09 brachioradialis (BR)
6 -37.59 -67.35 -4.74
palmaris longus (PL)
and
exor carpi radialis (FCR)
7 -66.75 -68.45 28.29
exor carpi radialis (FCR)
8 -44.33 -8.95 1.04 extensor carpi radialis brevis (ECRB)
Table 3.1: Coordinates of each recording electrode with respect to MRI coordinate system.
Electrode
Distance from
ne-wire electrode (mm)
1 68.68
2 41.91
3 14.41
4 20.15
5 42.78
6 59.04
7 69.18
Table 3.2: Distance from ne-wire electrode to each surface electrode in millimeters.
46
Figure 3.6 shows the computed STAs shown as white traces for each surface (Elec-
trodes 1 - 7) and the ne-wire (Electrode 8) channels overlaid on the transverse view of
the right forearm.
The yellow traces in each surface electrode subgure are the result of convolving the
ne-wire STA of the 50 LMS adaptive lter coecients calculated from each input-to-
output relationship. Each subplot is arranged to appear next to its respective electrode
around the arm, indicated by the numbered red dots. The red `x' labeled as number 8
indicates the location of the ne-wire recording and its trace is shown at the bottom of
the gure.
47
1
2
3
4
5
6 7
R
U
Time (ms)
-5 0 5 10 15 20 25
STA (V)
×10
-4
-4
-2
0
2
Electrode 1
Time (ms)
-5 0 5 10 15 20 25
STA (V)
×10
-3
-2
0
2
Electrode 2
Time (ms)
-5 0 5 10 15 20 25
STA (V)
×10
-3
-4
-2
0
2
4
Electrode 3
Time (ms)
-5 0 5 10 15 20 25
STA (V)
-0.01
0
0.01
Electrode 4
Time (ms)
-5 0 5 10 15 20 25
STA (V)
×10
-3
-2
0
2
Electrode 5
Time (ms)
-5 0 5 10 15 20 25
STA (V)
×10
-4
-4
-2
0
2
Electrode 6
Time (ms)
-5 0 5 10 15 20 25
STA (V)
×10
-4
-2
0
2
Electrode 7
Time (ms)
-5 0 5 10 15 20 25
STA (V)
-0.1
0
0.1
0.2
Electrode 8
N = 840
STA
LMS Estimate
x
8
Figure 3.6: Spike-triggered average of ne wire and surface electrodes with approximate
location. Peak of ne wire MUAP used as trigger. White traces indicate STA. Yellow
traces are estimates of STA. Electrode 8 is the STA of the ne-wire channel.
3.4 Discussion
In this, the rst study of its kind, we combined magnetic resonance imaging with precision
electrophysiological recordings to detect the precise location of an invasive EMG electrode
and map its spatial relationship within the muscle. We then continued to investigate
signal transformations that occur when MUAPs propagate through biological tissue to
surface electrodes. While invasive electromyography will likely remain the sole method
48
of recording high quality MUAPs, these results make it possible to begin to tease apart
electrical parameters of skin, fat, muscle and bone.
The location of the recording electrode relative to the source of the ne-wire recording
site is separated by a so-called volume conductor made up of biological tissues. It is well
known that these tissues act as a spatial low-pass lter to both attenuate and spread
the frequency of the signal (Day 1997). The extent of the ltering eects are dependent
on the electrical properties of the medium (i.e conductivity and prematurity), distance
from source to recording site, muscle ber size, etc. (Roth, Gielen & Wikswo 1988, Blok,
Stegeman & van Oosterom 2002, Farina et al. 2004).
In an anisotropic and inhomogeneous medium, such as skeletal muscle, factors such
as ber direction, appellation angle and frequency dispersion play an important role in
how signals are conducted (Lowery, Stoykov, Dewald & Kuiken 2004, Roth et al. 1988,
Roeleveld et al. 1997). In our analysis, system identication was best estimated for
Electrodes 2-5 in Fig. 3.6. The waveforms for these electrodes were preserved in these
channels due to their relative proximity to the ne-wire recording site. Peak latencies
in these channels were consistent with values obtained from conduction velocity studies
(Farina et al. 2002).
Low-pass ltering eects of the tissues becomes evident as the width of the action
potential gradually widens at surface electrodes farther away, with the most prominent
eect seen at electrode 5. The magnitude of the STAs for electrodes 2 - 5 are buried deep
within the noise
oor and are not detectable in the raw traces. In Electrodes 1, 6 and
7, the surface electrodes appear to be too far from the ne-wire site, resulting in poor
estimates of the STAs for those channels. However, by the number of events used in the
49
STA will improve both the SNR and temporal resolution (Farina et al. 2002). In this
study, we averaged 840 action potentials to generate the averages, but with thousands of
events, it could be possible to expose faint underlying signals in distant electrodes.
With the addition of high-resolution MR images, piecewise linear models incorporating
boundary conditions between two media can be developed to increase the accuracy of the
invasive recording site as well as the spatial distribution of tissues and surface recording
sites. Higher density surface EMG arrays would only further enhance the ability to
quantify electrical parameters such as conductivity and permittivity for each tissue (i.e
muscle, bone, skin, and fat). Signal processing techniques such as system identication,
independent components analysis and equalization techniques can be used to recover
an accurate representation of a single MUAP from a weighted combination of surface
electrode recordings as depicted in Fig. 3.7. Invasive recordings, X (s), pass through
system identied FIR lters,H (s), to an array of surface recordings, represented asY (s).
From the forward transfer models, an inverse model, H (s), is attainable. Statistical
processing techniques such as independent components analysis, blind deconvolution and
Bayesian inference can be integrated to undo the ltering eects and triangulate a muscles
and obtain its activation level.
H(s)
Y(s)
X(s) H
-1
(s)
X(s)
Figure 3.7: System identication and equalization for use in source localization.
The ability to locate the source of ground-truth MUAPs within a muscle along with
transfer function tting enables the optimization of system identication, validation of
50
volume conduction models and provides information regarding electrical properties of
biological tissue. Analysis of this data may lead to methods of source localization and the
decomposition of individual muscle activation levels during low isometric force production,
slow nger movements, and motion-to-force transitions leading to improvements in the
dexterous capabilities of robotic and prosthetic hands.
Acknowledgments
Co-authors Dr. Francisco J. Valero-Cuevas. Krisha Nayak for setting aside time for the
use of the MRI machine, Jason J. Kutch for the use of his equipment and lab during
data collection, and Dr. Gerald E. Loeb for the insertions and guidance during ne-wire
electrode placement.
This material is in part based upon the work supported by NSF Grant EFRI-COPN
0836042, NIH Grant R01-052345, NIH Grant R01-050520 to FVC, and NIH Supplement
R01-050520-W1 to AR.
51
Chapter 4
Power Spectral Density Analysis in Phase II Epilepsy
Patients with Implanted Subdural Electrodes
Abstract
The ability of the human hand to precisely control the direction and magnitude of ngertip
endpoint force during a dynamic manipulation task serves as indication of dexterous
ability. However, the classication of cortical rhythms into its constituent frequency bands
as they relate to dexterity demand has been relatively unexplored. In this study, cortical
activity was recorded from a grid of subdural electrocorticographic (ECoG) electrodes
implanted in two patients undergoing monitoring for intractable epileptic seizures prior
to epilepsy surgery. Subject 1 performed a series of three tasks with each successive task
increasing in dexterity demand. They rst held a rigid object at three points of contact
using their thumb, index and middle ngers with just enough force to prevent the object
from slipping. Secondly, the object was rotated back and forth in a twisting motion as if
tightening and loosening a bottle cap at a frequency of approximately 1 Hz. Lastly, they
performed the Strength-Dexterity (SD) test, a measurement of dexterity involving the
52
compression of an unstable spring between the index nger and thumb. Subject 2 also
performed the SD test, but not the other two tests. Power spectral density (PSD) was
calculated when the subjects were fully engaged in the task to eliminate any extraneous
eects. PSD was referenced to a quiescent period when the subjects were relaxed with
no movements or distractions. Preliminary results reveal task-dependent spectral shifts
in the motor and sensory cortices based on task diculty. A rm understanding of the
changes that occur in these cortical rhythms may pave the way to understanding how the
brain controls for dynamic force production and instability.
4.1 Introduction
Surgical implantation of subdural electrodes is the primary method of successfully local-
izing and excising epileptogenic zones (Wyler, Ojemann, Lettich & Ward Jr 1984, Stefan,
Quesney, AbouKhalil & Olivier 1991, Weinand, Wyler, Richey, Phillips & Somes 1992).
In some cases, large subdural electrode arrays placed over the temporal lobe, a common
location of seizure foci, overlap with the hand areas of the primary motor cortex per-
mitting the analysis of cortical rhythms involved in manipulation. Functional MRI has
been used to investigate dexterous manipulation (Mosier et al. 2011), however due to its
poor temporal resolution, the characterization of frequency activity during these complex
sensorimotor tasks remains largely unexplored. It is known that cortical oscillations in
specic frequency bands are present in motor tasks (Murthy & Fetz 1992). The phys-
iological relevant cortical oscillations are often classied into distinct frequency bands,
these are delta (0 - 4 Hz), theta (4 - 8 Hz), mu (8 - 12 Hz), sigma (12 - 15 Hz), beta (15
53
- 30 Hz), and gamma (> 30 Hz). Positive increases in power when in a frequency band
are a result of synchronous ring of underlying neurons and is referred to as event-related
synchronization (ERS). Conversely, a decrease in power is attributed to randomized r-
ing of underlying neurons and is referred to as event-related desynchronization (ERD)
(MacKay 2005, Pfurtscheller & Lopes da Silva 1999). Many studies use these time-
locked measurements to associate motor function with cortical rhythms. For example,
ERS in the theta range is associated with preparation of movement (Popivanov, Mineva &
Krekule 1999) and rhythmic movement (Turak, Louvel, Buser & Lamarche 2001), ERD in
the alpha range is associated with hand and nger movements (Pfurtscheller & Lopes da
Silva 1999) and tactile stimulation (Chatrian, Petersen & Lazarte 1959), and ERS in
the beta frequency range during fast nger tapping and ERS during slow nger tapping
and weak muscle activity (Toma, Mima, Matsuoka, Gerlo, Ohnishi, Koshy, Andres &
Hallett 2002).
Currently, there have been no ECoG studies which investigate cortical power during
dynamic dexterous manipulation. Here, dexterity is dened as the ability to dynamically
control the direction of the applied force at the ngertip (Valero-Cuevas et al. 2003).
In this study, we challenged participants to perform a variety of manipulation tasks
varying in dexterity demand to investigate cortical oscillations associated with simple
versus dicult motor tasks. We found that cortical area and bandwidth increased with
more dicult tasks and that the beta frequency range provided the most information
regarding task diculty. Work in this area may lead to advancement in understanding
the neural control strategies employed by the nervous system during object manipulation.
54
4.2 Methods
4.2.1 Ethics
We obtained ECoG recordings in two male patients (34 and 43 years old) who were being
continuously monitored in the ICU for epileptic seizures. The participants gave written
consent prior to data collection and the protocol was approved by the Institutional Review
Board at the University of Southern California.
4.2.2 Experimental Paradigm
In this study, the rst subject performed three tasks manipulation tasks. In the rst,
they was instructed to small object weighting approximately 300 g with a three ngered
grasp (thumb, index, and middle ngers) statically as shown in Fig. 4.1a for 30 seconds
with just enough force to keep it from slipping. In the second task, they rotated the
same object for 30 seconds in a twisting motion back and forth twisting motion as if
tightening and loosening a bottle cap at a rate of approximately 1 Hz (Fig. 4.1b). In the
third task, both subjects compressed a small slender spring between the index nger and
thumb as much as possible while trying to prevent the spring from buckling to the points
of maximal compression (Fig. 4.1c), a paradigm known as the Strength-Dexterity test
(Valero-Cuevas et al. 2003). In all tasks, normal forces were recorded from a uni-axial
load cell (Measurement Specialties, Hampton, VA) at the points where the ngertips
came into contact with the objects. Due to the placement of the subdural electrodes, the
subject performed the motor tasks with their left (non-dominant) hand.
55
Task 3: Unstable
Spring Compression
Task 1: Static
Three-!ngered Grasp
Task 2: Three-!ngered
Rotation
Increasing Dexterity Demand
~1Hz
(a) (b) (c)
Figure 4.1: Manipulation tasks performed during electrocorticographic recordings. Nor-
mal forces at the point of contact were recorded using a uni-axial load cell. From left
the right the tasks increase in dexterity demand from static hold to slow movements and
nally to unstable object manipulation. (2) Three-ngered static grasp of a 300 g object
using the thumb, index and middle ngers. (b) Three-ngered rotation using the 300 g
object which was oscillated back and forth in a twisting motion at a rate of approximately
1 Hz. (c) Strength-Dexterity test in which the subject compressed a slender spring as
much as possible using a precision pinch.
4.2.3 Electrocorticography (ECoG)
Neurophysiological monitoring was achieved using a 128-channel head box (EMU 128FS,
XLTEC). Subject 1 had an 8 8 array of subdural electrodes implanted in the right
hemisphere overlaying the sensorimotor cortex and portions of the prefrontal, premotor
and posterior parietal cortices as shown in Fig. 4.2a. Subject 2 was implanted with a
6 8 grid covering the right sensorimotor cortex as shown in Fig. 4.2b. The platinum
subdural electrodes had a with 1 cm center-to-center spacing with a contact diameter of
2.5 mm. Data for Subject 1 were sampled at 500 Hz and at 1 kHz for Subject 2 with
56
open lters. Prior to analysis data were ltered from 5 - 100 Hz and notch ltered at 60
Hz and harmonics up the Nyquist frequency.
1 2 3 4 5 6 7 8
9
17
25
33
41
10 11 12 13 14 15 16
18 19 20 21 22 23 24
26 27 28 29 30 31 32
34 35 36 37 38 39 40
42 43 44 45 46 47 48
1 2 3 4 5 6 7 8
9
17
25
33
41
10 11 12 13 14 15 16
18 19 20 21 22 23 24
26 27 28 29 30 31 32
34 35 36 37 38 39 40
42 43 44 45 46 47 48
(a)
(b) Subject 1 Subject 2
Figure 4.2: Approximate electrode grid layout for Subjects (a) 1 and (b) 2.
4.2.4 Spectral Analysis
The average power spectral density (PSD) was calculated during each task and further
separated into the delta, theta, mu, and beta frequency bands. All data were compared
to a quiescent time when the patient performed no movements. Three-second long epochs
of data were analyzed during which the subjects were well into the task.
4.3 Results
Figure 4.3 shows the changes in PSD in Subject 1 over the entire grid as compared to
a resting period. Cortical areas experiencing ERS are shown in dark red, indicating an
increase in power requirements for the specied task than during rest. ERD is shown in
57
blue and indicates that there was more power in rest than in the task being performed.
For Task 1, the only increase in power during is conned to the beta frequency, which is
consistent with literature regarding static force production (Murthy & Fetz 1992, Fetz &
Cheney 1980, Baker et al. 1997, Kilner et al. 1999). For Task 2, there were large increases
in power in the delta, theta and mu frequency ranges over the primary motor area,
which is also consistent with ndings in literature that have investigated slow rhythmic
movements (Turak et al. 2001, Pfurtscheller & Lopes da Silva 1999) and increases power
over the in the medial cortex in the mu and and beta frequency range. In the primary
motor area however, beta frequency range was abolished. During Task 3, there were
increases in power only in the medial cortex in the mu and beta range and increased
power in M1 in the beta frequency range. The spectral plots for Task 2 suggest that
both cortical area and bandwidth requirements increase with steady rotation movements.
Tasks 2 and 3 involve active moment which seems to be correlated with increases in power
in the medial cortex towards the supplementary motor area (SMA). For the tasks that
required a steady force production (i.e. Task 1 and 3), beta range power increased in the
motor cortex.
58
-1
0
1
Task 1 Task 2 Task 3
Power (mW) ∆
Delta Theta Mu Beta
ERS
ERD
Figure 4.3: ERS and ERD.
The changes in power as compared to rest in the beta frequency range for Subject 2
is shown in Fig. refECoGBetaPower.eps. This subject only performed the SD test (i.e.
Task 3 in Subject 1). During the SD test, beta power increased slightly in the most
posterior part of the sensorimotor area, however beta power decreased anterior to this
section while the other areas remained relatively unchanged from rest.
59
−3.5
−3
−2.5
−2
−1.5
−1
−0.5
0
0.5
x 10
−3
Power (W) ∆
1 2 3 4 5 6 7 8
9
17
25
33
41
10 11 12 13 14 15 16
18 19 20 21 22 23 24
26 27 28 29 30 31 32
34 35 36 37 38 39 40
42 43 44 45 46 47 48
Figure 4.4: Electrocorticographic power associated with dexterity.
4.4 Discussion
We were given a unique opportunity to collect cortical activity from arrays of subdural
ECoG electrodes implanted in two subjects who underwent monitoring for intractable
epilepsy prior to epilepsy surgery. Subject 1 performed tasks that varied in dexterity
requirement. Subject 2 only performed the most dexterous task. The aim of this study
was to determine whether spectral analysis of cortical activity in the motor and sensory
cortices could provide information about spatiotemporal relationships in the brain during
manipulation of unstable objects.
Subject 1 showed an increase in beta power over the sensorimotor cortex (Fig. 4.3)
which was in direct contrast to Subject 2, who showed that the hand area of the sensory
cortex exhibits a large decrease in power during spring compression (Fig. 4.4). This
dierence in power changes is likely due to the fact that Subject 1 was squeezing the
60
spring well within the stable domain. The compression force on the spring for which the
PSD was calculated was relatively low, suggesting that this subject was not incorporating
increased sensory feedback. Instead, they were squeezing the spring at low enough levels
that mimics stable force production. It has been shown that beta range oscillations in the
cortex are associated with the maintenance of stable forces (Murthy & Fetz 1992, Fetz
& Cheney 1980, Baker et al. 1997, Kilner et al. 1999). Subject 2, on the other hand,
was performing the SD test correctly and was squeezing the spring to nearly the point at
which the spring would
y out of the hand. Because this subject made dynamic updates
to the forces applied to the ends of the spring, beta rhythms in the sensorimotor cortex
attenuated. It has been shown in literature that dynamic movements cause beta rhythms
to disappear (Toma et al. 2002, Baker et al. 1997, Brown 2000, Feige et al. 2000).
These preliminary data suggest that dierences in task complexity are detectable us-
ing ECoG electrodes and are re
ected as changes in the power spectral density. Continued
data collected will assist in the development of cortical maps that relate cortical activity
during during the production of static and dynamic. We will also begin recording data
kinematic data to relate moment variability to changes in the PSD. Furthermore, elec-
tromyographic recordings will be added with the goal of detecting relationship between
cortical activity to muscle activity for simple and dicult tasks as well as measure con-
duction delays from cortex to endpoint movement and force. Future work in this area
may lead to advancement in brain-computer interfaces that allow for the control of higher
degrees of freedom and real-time classication of nger function.
61
Acknowledgements
Co-authors Emily L. Lawrence, Sarine Babikian, Dr. Charles Y. Liu, Dr. Christianne
N. Heck, and Dr. Francisco J. Valero-Cuevas. I would also like to thank Rosie Arreola
and all of the EEG technicians in the Department of Neurology at the Keck School of
Medicine for their assistance in setting up the equipment in the ICU for data collection.
A special thanks to Emily Lawrence for the design of the mobile lab. This material is
in part based upon work supported by NSF Grant EFRI-COPN 0836042, NIH Grant
R01-052345, NIH Grant R01-050520 to FVC, and NIH Supplement R01-050520-W1 to
AR.
62
Chapter 5
Introduction to Coherence
Abstract
Understanding the cortical control of voluntary movements is an important topic in neu-
roscience. One method of assessing this control is through coherence analysis. In this
chapter, the development, calculation and implications of coherence are discussed. We
begin with a description of cortical oscillations, current ideas on their generating mech-
anisms and how they are measured. Next, the motor associations of specic frequency
bands are discussed followed by a brief description of the connectivity between cortex
and muscles. Lastly, coherence is introduced as a tool for measuring the functional con-
nectivity between the cortex and peripheral muscles.
5.1 Introduction
In healthy humans, electroencephalography (EEG) and magnetoencephalography (MEG)
serve as the predominant methodology for non-invasive measurements of the temporal
dynamics of the active cortex. The ability to record high temporal activity enables
63
experimentalists and clinicians to subdivide the power spectra of cortical oscillations
into discrete frequency bands and begin to associate changes the in rhythms activity
with motor actions. Positive increases in the power spectra of underlying neurons in a
frequency range is referred to as event-related synchronization (ERS). Conversely, de-
creases in power at a frequency is referred to as event-related desynchonization (ERD)
(MacKay 2005, Pfurtscheller & Lopes da Silva 1999, Pfurtscheller & Neuper 1994, Leo-
cani, Toro, Manganotti, Zhuang & Hallett 1997). For example, ERS of sigma oscillations
of approximately 14 Hz are associated with concentration and suppression of a motor
response (Rougeul-Buser & Buser 1997, Nashmi, Mendona & MacKay 1994), whereas
ERD of beta rhythms coincide with slow nger tapping and weak muscular activity
(Toma et al. 2002).
Measuring the power in the brain gives a general idea of the role of oscillations in motor
control, however, in order to truly understand the cortical control of the hand, a relation-
ship between the rhythms of the cortex and of the musculature must be established. It has
long been known that the primary motor cortex (M1) projects directly onto the motor neu-
rons in the dorsal column of the spinal cord (Bernhard et al. 1953). As a result, descending
commands from M1 have been shown to exhibit synchronous oscillations with spinal motor
neurons which can be detected in the electromyogram (EMG) (Conway et al. 1995, Baker
et al. 1997, Brown 2000). Corticomuscular coherence (CMC) assess the strength of the
temporal correlation between EEG and EMG. CMC has been heavily investigated in the
beta frequency range (15 - 30 Hz) for static force production (Murthy & Fetz 1992, Baker
et al. 1997, Baker 2007, Conway et al. 1995, Kilner et al. 2000, Kilner et al. 2004, Kil-
ner et al. 1999, Kristeva et al. 2007, Chen et al. 2013), yet recent work has shown that
64
gamma CMC is present during slow oscillatory force production (Omlor et al. 2007, Om-
lor, Patino, Mendez-Balbuena, Schulte-Mnting & Kristeva 2011, Patino et al. 2008),
strong muscular contractions (Mima et al. 1999, Brown 2000, Hari & Salenius 1999)
and movement (Brown 2000, Hari & Salenius 1999).
In several studies, precision pinch tasks have been used to detect corticospinal projec-
tions from the primary motor cortex to contralateral muscles(Baker, Pinches & Lemon
2003, Lemon et al. 1995, Lemon & Mantel 1989, Muir & Lemon 1983). However it is
known that, in addition to M1, the corticospinal tract (CST) is comprised of neuron
originating from the supplementary motor area (SMA), the dorsal and ventral premotor
cortices, and the cingulate cortex (Dum & Strick 1991, Dum & Strick 2005, He, Dum
& Strick 1995), yet the functional coupling in the prefrontal cortex has been relatively
unexplored. Only one study, to our knowledge, has demonstrated high beta-range CMC
in the SMA for a ne precision pinch task (Chen et al. 2013).
Excluding gross movements of the arms, CMC has been investigated during static
grasps, slow and fast nger movements, and slow oscillatory force production. Needless
to say, there is still much research to be done, specically in the domain of dynamic
dexterous manipulation, such as when typing, writing, or playing an instrument. Further
coherence studies will undoubtedly serve as an invaluable tool for understanding the
neural control of the hand.
65
5.2 Oscillations in the Cortex
The universe is lled with oscillatory systems: planetary orbits, electromagnetic radiation
such as light, the pendulum of a clock, and even sound is vibration of the air. Therefore,
it's not surprising that the nervous system also possess oscillatory characteristics. The
rst human electroencephalographic recordings of cortical oscillations in the 8 - 12 Hz
frequency range were recorded by Hans Berger in 1929 (Berger 1929). Since then, nu-
merous clinical and research studies have investigated the functional signicance of these
oscillations.
Studies have shown that individual cortical neurons possess the inherent ability to
re over wide frequency ranges (Llins 1988, Hutcheon & Yarom 2000), however, models
describing the generating mechanisms of these oscillations have yet to be developed. It
has been suggested that the rhythm of cortical oscillations is the result of pyramidal
neurons discharging during post-inhibitory rebound excitation following the discharge of
large populations of inhibitory neurons joined through gap junctions (Jasper & Stefanis
1965, Pauluis, Baker & Olivier 1999). Another theory hypothesizes that intrinsic cell
properties and cortico-thalomocortical networks serve as the main underlying generator
of cortical oscillations (Steriade 1997).
The parallel arrangement of long dendritic branches of pyramidal neurons act as neural
dipoles creating time-varying eld potentials that can be recorded through the scalp us-
ing electroencephalography (EEG) and magnetoencephalography (MEG) (MacKay 2005).
This accumulation of electrical activity from large populations of neurons is comprised of
several frequencies summed together and is best viewed in the frequency domain. Changes
66
in the power spectral density (PSD) of EEG and MEG recordings can be characterized
into two categories: event-related synchronization (ERS) and event-related desynchro-
nization (ERD). During ERS, the underlying neuron population res synchronously which
is seen as an increase in the PSD at the ring frequency. Conversely, during ERD, the
neuron population are ring in a more randomized pattern causing the PSD to decrease
at that particular frequency (MacKay 2005). In both of cases, the change in the PSD
must be measured relative to a reference period, typically the time prior to a stimulus
event (Pfurtscheller & Lopes da Silva 1999, Da Silva & Pfurtscheller 1999).
Spectral analysis of EEG and MEG signals over the sensorimotor cortex have enabled
the association of specic frequency bands with motor tasks. Physiologically relevant
oscillations have been observed up to 80 Hz and have been classied into ve distinct
bands: theta (4 - 8 Hz), alpha (8 - 12 Hz), sigma (12 - 15 Hz), beta (15 - 30 Hz), and
gamma (>30 Hz). Table 5.1 provides an overview of the motor associations for each
frequency band as it relates to changes in sensorimotor PSD.
Inspection of Table 5.1 clearly indicates that oscillations in the cortex are an important
factor in motor control. Yet the underpinnings surrounding the reasons why the brain
operates in a specic frequency band for a motor task and switching to another range for a
dierent task is currently unknown. Without a concrete explanation of how and why the
brain oscillates in the way that it does, research is limited to identifying consistencies in
PSD changes observed prior to and after a stimulus. Additional problems arise due to the
denition of what constitutes a certain frequencies band. For example the `beta' frequency
range is 13 - 24 Hz according to Gross et al. (Gross, Pollok, Dirks, Timmermann, Butz
& Schnitzler 2005) whereas Conway et al. extends this range by another 11 Hz to include
67
Name
Frequency
Motor Association
Range (Hz)
theta 4 8
ERS: preparation of movement (Popivanov et al. 1999), rhythmic movement (Turak
et al. 2001)
alpha 8 12
ERS: relationship to nger tremor (Jasper & Andrews 1938)
ERD: hand and nger movements (Pfurtscheller & Lopes da Silva 1999); tactile
stimulation (Chatrian et al. 1959); task complexity and attention (Boiten, Sergeant &
Geuze 1992, Dujardin, Derambure, Defebvre, Bourriez, Jacquesson & Guieu 1993)
sigma 12 15
ERS: attentive preparation/concentration, suppression of motor response
(Rougeul-Buser & Buser 1997, Nashmi et al. 1994)
ERS: fast nger tapping (Toma et al. 2002)
beta 15 30
ERS: fast nger tapping (Toma et al. 2002)
ERD: slow nger tapping, weak muscle activity (Toma et al. 2002), slow and brisk nger
movements (Stanck Jr & Pfurtscheller 1996)
gamma 30 50
ERS: during (rigorous) muscle activity and attention, onset of movement (Popivanov et
al., 1999)
Table 5.1: Event-related synchronization (ERS) and event-related desynchronization (ERD) in the cortical motor areas in asso-
ciation with specic motor tasks.
68
13-35 Hz (Conway et al. 1995). Those researchers working in the lower gamma frequency
range may feel that they are being infringed on. Overall, the investigation of power
spectral changes in EEG and MEG provides us with a view of one side of the coin.
The nervous system is a complex interaction between brain and body and as such, it is
necessary to investigate the transmission of cortical information to the periphery.
5.3 Correlation
To understand coherence and its derivation, we begin with the mathematical denition
of Pearson's correlation coecient between two variables x and y, which is given as:

xy
=
cov (x;y)

x

x
(5.1)
where cov is the covariance and
x
and
y
are the standard deviations of x and y,
respectively. The correlation coecient is counted between -1 and +1. A correlation of
+1 indicates a perfect relationship (e.g. the signal area basically the same, but may have
dierent amplitudes), a -1 would indicate that the signals are anti-correlated (e.g. out of
phase by 180

), and a 0 suggests that there is no relationship between the two (e.g. both
signals are white noise).
Correlation is a bivariate analysis, meaning that this can only be performed with two
signals, unlike Fourier transform or evaluating the mean, which can be performed on a
single signal. This raises problems which are demonstrated using the following examples.
Consider the signals x
1
(t) = cos (215t) and x
2
(t) = 0:5cos (215t) shown in Fig.
5.1a and their Fourier domain transformations shown in Fig. 5.1b. These signals are
69
pure cosine waves with dierent amplitudes and no dierence in phase. This results in a
perfect linear correlation of +1. Notice in this example that, regardless of the amplitudes
of the oscillations, the correlation coecient is unaected.
(a) (b)
Time (s)
0 0.05 0.1 0.15 0.2
Amplitude (a.u.)
-1
-0.5
0
0.5
1
ρ = 1
Frequency (Hz)
0 10 20 30 40 50
Magnitude
0
0.5
1
Frequency Spectra
Figure 5.1: Two cosine signals and frequency representation with a perfect linear cor-
relation. (a) The blue trace is a pure cosine wave with a frequency of 15 Hz and an
amplitude of 1. The red trace is a 15 Hz cosine wave with an amplitude of 0.5. The
correlation between the two signals is = 1. (b) Frequency domain representation of the
cosine signals in (a). The peak frequency is at 15 Hz for both traces and their magnitude
directly relate to the amplitude of their respective cosine waves.
In the next example, signalsx
3
(t) =cos (215t=5),x
4
(t) = 0:5cos (215t=2)
and x
5
(t) = 0:5cos (215t) have been added and are shown in Fig. 5.2a and their
frequency domain transformations are shown in Fig. 5.2b. Here, it can be seen that
although the frequency of oscillations remains the same for all signal, phase shifts of =5,
=2 and result in correlation coecients of
x1x3
= 0:81,
x1x4
= 0 and
x1x5
=1.
70
The frequency spectra for the additional signalsx
3
(t),x
4
(t) andx
5
(t) are the same. The
critical problem with simple correlation begins to arise. The correlation between these
signals is now dependent of the phase shift. The magnitude of the correlation should
not depend on the phase shift as this can often happen in real data recordings due to
transmission delays.
Time (s)
0 0.05 0.1 0.15 0.2
Amplitude (a.u.)
-1
-0.5
0
0.5
1
ρ = 0.80958, ρ = 0, ρ = -1
Frequency (Hz)
0 10 20 30 40 50
Magnitude
0
0.5
1
Frequency Spectra
(a) (b)
Figure 5.2: The eects of phase shifting a signal on correlation. (a) The primary signal
(blue trace) is a cosine with an amplitude of 1 and frequency of 15 Hz. The next three
signals share the same frequency but are shifted by =5 (red trace), =2 (yellow trace)
and (purple trace), resulting in correlation coecients of = 0:81, = 0 and =1.
(b) Frequency domain representation of the cosine signals in (a). The main trace is
represented in blue with magnitude 1 and the three shifted waves overlap each other and
have magnitude 0.5 at 15 Hz.
71
One nal example demonstrates the problems with correlation coecients. Consider
two signals, each with two distinct frequencies, x
1
(t) =cos (215t) + 0:01cos (2250t)
and y
1
(t) = 0:5cos (215t=2) + 0:03cos (2250t). These are shown in Fig. 5.3a.
Both signals contain the same two frequencies, however they dier in amplitude. The
frequency spectra in Fig. 5.3b shows large peaks for the signals at 15 Hz and almost
unseen peaks at 250 Hz. The correlation coecient for this signal pair is
x1y1
= 6 10
4
(i.e. nearly zero). Now, if we multiply the 250 Hz component in both signals by 100,
eectively replacingx
1
withx
2
(t) = 0:5cos (215t=2)+cos (2250t) and replacingy
1
withy
2
(t) = 0:5cos (215t=2)+3cos (2250t), we obtain the signals in Fig.5.3c and
frequency transform in Fig. 5.3d. The resulting correlation coecient is now
x2y2
= 0:7,
a much dierent value than before.
72
Time (s)
0 0.05 0.1 0.15 0.2
Amplitude (a.u.)
-1
-0.5
0
0.5
1
ρ = 0.00059686
Frequency (Hz)
0 100 200 300
Magnitude
0
0.5
1
Frequency Spectra
Time (s)
0 0.05 0.1 0.15 0.2
Amplitude (a.u.)
-3
-2
-1
0
1
2
3
ρ = 0.69752
Frequency (Hz)
0 100 200 300
Magnitude
0
1
2
3
Frequency Spectra
(a) (b)
(c) (d)
Figure 5.3: Eect of frequency component magnitude on correlation. (a) The primary
wave (shown in blue) consists of two frequencies: a 15 Hz component with unit amplitude
and a 250 Hz component with amplitude 0.01. The second signal consists of the same
frequency components, however the 15 Hz component has an amplitude of 0.5 and a 250
Hz component of 0.03. The correlation between the signals is nearly 0 (b) Frequency
domain representation of signals in (a). The 15 Hz components for both signals are
much larger than the 250 Hz components. (c) The two signals have similar frequency
components as in (a) however, the 250 Hz components have been amplied by 100. The
correlation is now = 0:7. (d) Frequency spectra of the signals in (c). The 250 Hz
components now dominate the 15 Hz components.
73
These examples demonstrate that initially, when one frequency component was shared
between the two signals, the phase dominated the magnitude of the correlation coecient
while amplitude had no eect. However, as the complexity of the signal increases by
adding more tones, both phase and amplitude of the signal components had drastic
eects on the correlation coecient. In real world electrophysiological signals, we are
not aorded the luxuries of pure sine and cosine waves, therefore it seems a shame to
describe the relationship between two complex signals using a single coecient, such as
correlation. Instead, it is necessary to describe the relationship between two signals on a
per frequency basis. This is the basis of coherence.
5.4 Calculation of Coherence
Brie
y, coherence is a measure of the temporal correlation between two signals (Nunez,
Srinivasan, Westdorp, Wijesinghe, Tucker, Silberstein & Cadusch 1997). The concept of
coherence can be described with the following example. Imagine that we were observing
a wave traveling down the length of a rope as one end is moved up and down and arrives
at the other end some time later. We can clearly see that the two ends are physically
connected and the peak of the wave at one end always appears at a xed time later at the
other end. However, imagine now that we were unable to see the length of the rope, but
we could oscillate one end at a desired frequency and amplitude and observe the response
at the other end. After a few trials, it could quickly be concluded that if an oscillation
at a particular frequency occurring at the end we control always occurs at a constant
time later at the other end, there likely exists a relationship between the two systems.
74
Continued experimentation would only strengthen the argument that the systems are
functionally connected and by extending this relationship to include all frequencies, we
could develop a correlation spectrum. This is the basic underlying principle of coherence.
For the time-series variable x (t), let the auto spectral density be dened as
P
xx
(f) =
1
L
L
X
i=1
X
i
(f)X

i
(f) (5.2)
where X (f) is the Fourier transform of x (t) of the current segment (i = 1:::L) and
denotes the complex conjugate. Similarly, for y (t),
P
yy
(f) =
1
L
L
X
i=1
Y
i
(f)Y

i
(f): (5.3)
The auto spectral density of a signal represents the power of the signal. The cross
spectrum of the signals x (t) and y (t) is represented as
P
xy
(f) =
1
L
L
X
i=1
X
i
(f)Y

i
(f) (5.4)
Coherence is calculated by normalizing the square of the cross-spectral density be-
tween two signals by the product of their individual auto spectral densities (Rosenberg,
Amjad, Breeze, Brillinger & Halliday 1989, Farmer, Bremner, Halliday, Rosenberg &
Stephens 1993, Baker et al. 1997, Nunez et al. 1997) as indicated in Eq. 5.5. This gives
the equation for coherence as
C
xy
(f) =
jP
xy
(f)j
2
P
xx
(f)P
yy
(f)
(5.5)
75
Notice the similarity in the equation of coherence with that of Eq. 5.1 for correlation.
The dierence is that coherence is bounded between 0 and 1 (rather than -1 and +1) and
is a function of frequency. Hence for each frequency of interest, a correlation coecient
is assigned. For a coecient of 1 at a particular frequency, there exists a perfect linear
relationship between the two signals, while a 0 indicates linear independence (Farmer
et al. 1993).
As with any study, it is necessary to compare across conditions and subjects. There-
fore an appropriate transform which accounts for dierences in trial length and number
of samples. A Z-transformation provides an adequate technique for handling these un-
avoidable dierences. The Z-transformation for coherence is given as
Z (f) =
tanh
1
(C
xy
(f))
1
2T2
q
1
2T2
(5.6)
where C
xy
is the coherence between two signals, T is the number of sections used in
the spectral estimation, and tanh
1
is the hyperbolic arctangent.
In physiological measurements, coherence between cortical activity and muscular ac-
tivity is given the name corticomuscular coherence (CMC). Before getting into the inter-
pretation of CMC in literature, it is important to know that coherence is a measurement
that relies on spectral power estimation (i.e. P
xx
, P
xy
and P
xy
). The topic of power
spectral estimation is discussed next.
76
5.5 Multitaper Power Spectral Density Estimation
Thousands of books and articles have been written on the subject of estimating the power
content in a time series signal, and justiably so, as this is a topic of immense importance.
While numerous methods exist for calculating PSD of a signal, this section focuses on
the multitaper method (Thomson 1982, Percival & Walden 1993).
Perhaps the most widely known (and used) spectral estimation technique is the pe-
riodogram. In this method, the power spectral density of a time-series signal, x (t), is
measured within a within a xed time window of length t. This is often referred to as a
rectangular window or, due to its resemblance to that of a railroad car on a train, a box-
car. The boxcar window is shifted (with or without overlap) N times over the length of
the signalx (t). The estimate of the PSD is found by averaging the N power calculation.
Overlapping the windowed section of data window helps to reduce variance in the PSD
estimation. The periodogram is biased and does not provide a an accurate estimation
of the PSD. Furthermore, the sharp edges of the boxcar window causes spectral leakage.
Spectral leakage occurs when the true power of a frequency leaks into nearby frequencies.
The frequency representation of the boxcar shows shows side lobes that allow power from
nearby frequencies to leak into the power estimation. To overcome this problem, windows
that taper o at the ends have been developed, these include the Hann, Hamming, and
Gaussian windowing functions to name only a few. The frequency domain representations
of these window functions are designed to have smaller side lobes to reduce the in
uence
of nearby frequencies in the spectral estimation.
77
In each of these functions, a single estimate of the power spectral density in a calcu-
lated by multiplying the time-series data by the tapers. However, multiple measurements
of the spectral content within a time window will help produce an unbiased estimate of
the PSD. This method is represented mathematically in Eqs. 5.7 and 5.8 (Pesaran 2008).
In these equations, S
MT
is the multitaper spectral estimate of the individual segment
estimatione x
k
(f) with K tapers of length N and w
t
(k) is the window function operated
on time segment x
t
.
S
MT
(f) =
1
K
K
X
k=1
je x
k
(f)j
2
(5.7)
e x
k
(f) =
N
X
t=1
w
t
(k)x
t
e
2ift
(5.8)
A special class of orthogonal window functions, known as the discrete prolate spheroidal
sequences (DPSS) or Slepian tapers (Slepian & Pollack 1961, Thomson 1982), provides
an optimal method of maximizing the spectral concentration properties within a spec-
ied bandwidth (Pesaran 2008). The rst three Slepian tapers are shown in Fig. 5.4.
With each successive taper sequence, the function has one more zero crossing and works
to accurately estimate higher frequencies within the time segment. The goal of these
sequences is to maximize the frequency estimation in the bandwidth [W;W ], where W
is the half-bandwidth parameter. For a taper length of N, K = 2NW 1 sequences are
created.
78
Time (s)
0 0.1 0.2 0.3 0.4 0.5
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
Sequence 1
Sequence 2
Sequence 3
Figure 5.4: First three Slepian multitapers.
The selection of the half-bandwidth parameter and length is dependent on the re-
quirement of the experimenter and determines how smooth the spectral estimate will be.
Narrow bandwidths will produce noisy spectral estimations, while a large half-bandwidth
parameter will prohibit the detection of specic frequencies in the signal of interest. In
physiological signals, it can be assumed that oscillations specic frequency are shared
with nearby frequencies. Increasing the number of beyond K will only provide worse
spectral estimation (Pesaran 2008).
79
5.6 Data Preprocessing
Numerous factors impact the calculation of coherence which can lead to erroneous co-
herence spectra. Therefore, it is important to take the necessary care in the prepro-
cessing of electrophysiological data and consider methodological in
uences such as EEG
reference, EMG rectication, epoch normalization, and frequency spectra estimation to
name a few. The choice of reference for electroencephalographic data is a topic of much
debate. Re-referencing schemes commonly used are Laplacian (Zhou 2012, Nunez, Silber-
stein, Cadusch, Wijesinghe, Westdorp & Srinivasan 1994, McFarland, McCane, David &
Wolpaw 1997, Peters, Pfurtscheller & Flyvbjerg 2001), common average reference (CAR)
(Bertrand, Perrin & Pernier 1985), linked earlobe reference, and Hjorth transformation
(Hjorth 1975, Hjorth 1980). The choice of reference can be used to detect the eects of the
underlying electrodes while removing the in
uence of distance sources. Normalization of
data epochs in EEG is recommended for coherence analysis in order to remove sections of
data that favor larger amplitudes (Halliday & Rosenberg 2000). Electrophygiolocal data
is typically non-stationary, therefore smaller sections are also preferred so as to assume
quasi-stationarity in shorter time windows. Relevant EEG has been reported up to 100 Hz
(Ohara et al. 2001, Marsden, Werhahn, Ashby, Rothwell, Noachtar & Brown 2000, Crone,
Miglioretti, Gordon, Sieracki, Wilson, Uematsu & Lesser 1998), and thus EEG data are
typically low-pass ltered below 200 Hz.
Rectication is a standard procedure for EMG data prior to coherence analysis. This
enables the detection of grouped ring rates of rates (Mima & Hallett 1999, Myers,
Lowery, O'malley, Vaughan, Heneghan, Gibson, Harley & Sreenivasan 2003). The literate
80
debates on the necessity of rectifying EMG and some argue for (Yao, Salenius, Yue, Brown
& Liu 2007, Myers et al. 2003) and against it (McClelland, Cvetkovic & Mills 2012). As
with EEG, normalization of EMG is recommended for the same purposes mentioned
earlier.
5.7 Corticomuscular Coherence Review
Specically, the analysis of coherence between cortical activity (e.g. EEG, MEG, ECoG,
and LFP) and muscular activity (e.g. sEMG and ne-wire) is termed corticomuscular
coherence (CMC). Unlike, fMRI, coherence cannot be calculated during a resting pe-
riod since at least some amount of muscular activity is required. Table 5.2 lists several
examples of CMC ndings across several cortical frequency bands.
As is evident in Table 5.2, the beta frequency range is heavily studied. The main
nding in these studies is that beta CMC is found mostly during isometric force contrac-
tions (Murthy & Fetz 1992, Baker et al. 1997, Baker 2007, Conway et al. 1995, Kilner
et al. 2000, Kilner et al. 2004, Kilner et al. 1999, Kristeva et al. 2007) and gamma CMC
engages cortico-spinal coupling during dynamic force production (Omlor et al. 2007, Om-
lor et al. 2011, Patino et al. 2008, Chakarov et al. 2009). It can be concluded that beta
and gamma CMC are an essential part of object manipulation, however further stud-
ies are required to truly understand how the cortex and periphery utilize synchronous
frequencies in the control of movement.
81
Name
Frequency
Motor Association
Range (Hz)
theta 4 8 N/A
alpha 8 12 brief nger movements (Feige et al. 2000, Ohara et al. 2001)
sigma 12 15 during startle re
ex (Grosse & Brown 2003)
beta 15 30
static force production (Murthy & Fetz 1992, Baker et al. 1997, Baker 2007, Conway
et al. 1995, Kilner et al. 2000, Kilner et al. 2004, Kilner et al. 1999, Kristeva et al. 2007),
attention (Kristeva-Feige, Fritsch, Timmer & Lcking 2002, Safri, Murayama, Igasaki &
Hayashida 2006, Safri, Murayama, Hayashida & Igasaki 2007), performance (Witte,
Patino, Andrykiewicz, HeppReymond & Kristeva 2007, Kristeva et al. 2007)
gamma 30 50
strong muscular contractions (Mima et al. 1999), slow oscillatory force production
(Omlor et al. 2007, Omlor et al. 2011, Patino et al. 2008, Chakarov, Naranjo,
Schulte-Mnting, Omlor, Huethe & Kristeva 2009)
Table 5.2: Corticomuscular coherence studies across the dierent frequency bands.
82
5.8 Separation of Power and Coherence
Lastly, an important distinction must be between power measured in the cortex and
coherence between these oscillations and muscular activity. Measures of cortical power
provide experimentalists with a view of one side of a multi-dimensional coin. In order to
understand how the brain controls the body, it is necessary to observe the transmission of
oscillations throughout the nervous system. To state it boldly and succinctly: power does
not equal coherence. Despite the fact that coherence relies on spectral power estimation
its calculation, it has been shown in numerous studies that reductions and/or increases
in cortical power in a specic frequency range, do not necessarily aect coherence magni-
tude. For example, intravenous injections of diazepam, a known enhancer of beta cortical
activity, were given to healthy subjects. Cortical activity in the beta frequency range in-
creased signicantly, however, corticomuscular coherence remained relatively unaected
(Riddle, Baker & Baker 2004, Baker & Baker 2003).
As an important reminder, coherence is a measure of the temporal correlation be-
tween two signals. The strength of coherence depends on the synchronous activity of
two measurement, as determined by the phase relationship of the two signals. A prime
example of this was demonstrated by Nunez et al. (Nunez et al. 1997). In this example,
two signals comprised of the same three frequency components (5, 12 and 20 Hz) with
unequal amplitudes (5, 12 and 1), as shown in Fig. 5.5a, were generated with each hav-
ing dierent consistencies in their phase relationships. In each signal the phase would
randomly change every second. In the 5 Hz component, the phase would vary between
83
18

, the 12 Hz signal would change vary by180

and the 20 Hz signal had a consistent
phase throughout the entire duration.
The power in each of the frequency components was directly related to the amplitude,
hence the power in 5 Hz component was the lowest of the three components, the 20 Hz
component had the greatest power, and the 20 Hz component had the least power. The
signal with the most chaotic change in phase shift from one second to the next was the
12 Hz component. As a result, even though it had the highest power, the coherence,
as shown in 5.5b is non-existent. The signal with relatively small changes in the phase
consistency, (i.e. the 5 Hz component), retained a strong measure of coherence between
the two signals. For the 20 Hz signal which had the lowest power (Fig. 5.5a), exhibited
the highest coherence (Fig. 5.5b) due to the unchanging phase relationship with each
passing second.
84
(a)
(b)
Amplitude Coherence
Figure 5.5: Eect on the randomization of signal phase on coherence. (a) Two signals
which share frequency components at 5, 12 and 20 Hz are created, each with phase
relationships that vary after each second. (a) The amplitude of the 5, 12 and 20 Hz
components are 5, 12 and 1, respectively. The phase of the 5 Hz component is randomly
varied between18

, the 12 Hz component varies by180

and the 20 Hz component
contains an unchanging phase with each passing second. (b) Since the phase of the 5 Hz
components was bounded within a small range, the coherence at this frequency remains
relatively strong. Because the phase of the 20 Hz component varies drastically from
second to second, the coherence is extremely low. Lastly, the 20 Hz signal, although it
had the smallest amplitude, had the strongest coherence due to the consistency in the
phase throughout the duration of the signals. Taken from Nunez et al. (1997).
85
In summary, this chapter provided an overview of the range of physiologically relevant
cortical oscillations and discussed their possible generating mechanisms. The problems
with correlation as a measure of the relationship between two complex physiological sys-
tems was discussed. Coherence was introduce was a measure that provides a quantitative
analysis of the likelihood that two signals are either physically or functionally related.
Over repeated measures of the consistency of the phase relationship, the conclusion that
there is indeed a relationship is only strengthened. A review of the studies which have
used coherence were presented. It was shown that the beta and gamma frequency ranges
are of the greatest interest and high beta corticomuscular coherence is present during
sustained muscular contractions while gamma CMC is present during more dynamic mo-
tor tasks. Lastly, it was shown that changes in the power of a signal do not provide an
indication of how the coherence will be aected, but instead is dependent on the temporal
correlation between the two signals.
86
Chapter 6
Synchronous Corticomuscular Oscillations During Dynamic
Unstable Manipulation
Abstract
The control of ngertip forces is essential for dexterous manipulation. However, the
functional role of the primary motor cortex and other cortical areas remains relatively
unexplored for dynamic manipulation. In this study, we investigated corticomuscular
coherence in the beta (15 - 30 Hz) and low gamma (30 - 45 Hz) frequency ranges to
quantify the functional connectivity of the cortex to hand musculature during the pro-
duction of precision pinch forces on a rigid object (a wooden dowel) and a compliant and
unstable object (a slender spring prone to buckling). For both objects, 15 right-handed
participants produced and held force levels set to 40% and 80% of the greatest com-
pression force they could sustain with the spring. This produced four unique conditions:
spring-low (SL), dowel-low (DL), spring-high (SH), and dowel-high (DH). We used EEG
to calculate corticomuscular coherence (CMC) with the rst dorsal interosseous (FDI)
and abductor policies brevis (APB) muscles during steady hold periods. In the DL and
87
DH conditions, signicant CMC appeared over the sensorimotor cortex, however, their
magnitudes were not signicantly dierent. The highest beta CMC was found in the
sensorimotor cortex during the SL task, however, in the SH condition, beta range co-
herence in M1 was abolished. A linear mixed-eects model was used to investigate the
eect of task condition on beta range coherence in M1. The results showed that there
were no signicant dierences in eect for low force compressions for either object on
beta coherence. However, there was a clear dierence in eect for the DH and SH condi-
tions (p< 0:001). Our investigations of gamma range CMC showed signicant increases
over the supplementary motor area (SMA) during the SH task. We speculate that the
presence of gamma coherence in the SMA during the most dexterously demanding task
supports the notion that higher frequencies may be an indication of rapid sensorimotor
integration and strategic motor planning. Overall, these ndings suggest that for preci-
sion force control, cortical drive to contralateral hand muscles is modulated by dexterity
demand and there exist context-sensitive cortical circuits involved in the control of stable
and unstable manipulation.
6.1 Introduction
Beta range (15 - 30 Hz) oscillations in the sensorimotor cortex of humans and non-human
primates have been frequently observed (Sanes & Donoghue 1993, Donoghue, Sanes, Hat-
sopoulos & Gal 1998, Witham, Wang & Baker 2010, Murthy & Fetz 1992, Murthy &
Fetz 1996a, Murthy & Fetz 1996b, Conway et al. 1995, Lebedev & Wise 2000, Mima
88
et al. 1999, Stanck Jr & Pfurtscheller 1996). Similar oscillations recorded in the elec-
tromyogram (EMG) have been shown to be phase-locked to these cortical rhythms,
specically around 20 Hz, during the maintenance of static forces (Baker et al. 1997, Con-
way et al. 1995, Murthy & Fetz 1992, Salenius, Portin, Kajola, Salmelin & Hari 1997).
The functional connectivity between cortical and muscular activities can be quanti-
ed using corticomuscular coherence (CMC), which measures the strength of the phase
relationship between two spatially separated systems (Conway et al. 1995, Rosenberg
et al. 1989, Farmer et al. 1993). While the neural mechanisms generating these syn-
chronous oscillations are poorly understood, research suggests that CMC magnitude re-

ects both eerent commands and aerent sensory feedback (Fisher et al. 2002, Riddle
& Baker 2005, Baker, Chiu & Fetz 2006).
It has been shown that factors such as applied force level, fatigue, muscle cooling,
movement preparation, and learning largely in
uence beta CMC magnitude (Witte et al.
2007, Mendez-Balbuena, Huethe, Schulte-Mnting, Leonhart, Manjarrez & Kristeva 2011,
Tecchio, Porcaro, Zappasodi, Pesenti, Ercolani & Rossini 2006, Riddle & Baker 2005).
Other studies which show the disappearance of beta CMC during movement provide
further evidence for the association of static force production with beta rhythms (Baker
et al. 1997, Brown 2000, Feige et al. 2000, Kilner et al. 2000, Kilner et al. 2004, Kilner
et al. 1999). Low force (< 5:0 N) precision pinch paradigms have been used to study
cortico-spinal interactions (Baker et al. 2003, Lemon et al. 1995, Lemon & Mantel 1989,
Muir & Lemon 1983, Lemon, Baker, Davis, Kirkwood, Maier & Yang 1998). Other studies
have found that CMC is modulated by digit displacement (Riddle & Baker 2006) and
object compliance (Kilner et al. 2000).
89
The beta range CMC literature tends to focus heavily on static force production,
however the manipulation of objects involves far more dynamical interactions, thereby
mandating the exploration of CMC in other frequency ranges. Gamma (30 - 45 Hz)
CMC has been shown to be associated with moderate (Brown et al. 1998) and maximal
muscular contractions (Brown et al. 1998, Mima et al. 1999). Gamma CMC has also been
reported in preparation of wrist movement prior to a go cue (Schoelen, Poort, Oostenveld
& Fries 2011, Schoelen, Oostenveld & Fries 2005). More recently, dynamic movement
investigations have shown that peak coherence shifts from beta into the gamma frequency
range (30 - 45 Hz) during slow predictable (Omlor et al. 2007, Patino et al. 2008, Brown
et al. 1998) and unpredictable (Omlor et al. 2011) oscillatory force tracking tasks.
Anatomical studies have shown that, in addition to M1, axons of the corticospinal
tract originate in the dorsal and ventral premotor cortex and SMA (Dum & Strick 1991,
Dum & Strick 2005, Kuypers 1960, Shinoda et al. 1981). In one study, it was shown that
during a ne uni-manual manipulation task, beta CMC extended into the SMA (Chen
et al. 2013). This evidence suggests that context-sensitive cortical circuits are engaged
based on dexterity demand.
From these studies, it is clear that cortico-muscular coupling in the beta range regu-
lates static force while gamma coherence presides during dynamic movements and forces.
However, dexterous precision pinch manipulation involves both static and dynamic ele-
ments, therefore, it is necessary to investigate the entire coherence spectrum when per-
forming a task that requires the continual adjustment of ngertip forces. In this study, we
expand on these previous ndings to investigate the role of synchronous cortico-muscular
oscillations during a dexterously demanding paradigm in which the precise control of
90
endpoint nger force is paramount to task performance. Dexterity is dened here as
the ability to precisely control the magnitude and direction of ngertip endpoint force
(Valero-Cuevas et al. 2003). Based on this denition, the Strength-Dexterity (SD) test
was developed to quantify dexterous ability by asking participants to squeeze a slender
spring with unstable characteristics as much as possible before buckling (Valero-Cuevas
et al. 2003). This paradigm has been used in fMRI studies to reveal context-sensitive
cortical areas involved in the control of unstable springs (Mosier et al. 2011, Holmstrom
et al. 2011, Talati et al. 2005).
Based on the promising results of these ndings, we developed a visuomotor force
tracking version of the SD test to assess the modulation of coherence during matched
force compression of an unstable compressible spring as compared to the compression of
a rigid dowel. We found that beta CMC was present in the primary motor area during
low and high force compressions with the wooden dowel and at the low force compression
with the spring. Furthermore, we found that, despite matched force levels, beta range
CMC during the high compression levels with the spring was signicantly reduced in
comparison to the high force dowel task in the primary motor cortex. In this dexterously
demanding task, we observed a spatial shift of CMC into the SMA and a spectral shift
into the gamma frequency range. These ndings suggest that the extinction of beta CMC
in the sensorimotor cortex in favor of gamma in the SMA for the most unstable and
dynamic task may indicate concentration, rapid sensorimotor integration, and strategic
motor planning.
91
6.2 Methods
6.2.1 Subjects
15 healthy participants (30:3 4:6 years, 6 females) who were self reported as right-
handed took part in this study. There were no known neurological conditions in any of
the participants, nor did they report any prior hand injuries or surgeries. Subjects were
consented prior to the experiment and the protocol was approved by the Institutional
Review Board (IRB) at the University of Southern California.
6.2.2 Experimental Paradigm
6.2.2.1 Task 1: Strength-Dexterity (SD) Test
We obtained a measure of each participant's dexterous performance by employing a
paradigm known as the Strength-Dexterity (SD) test (Valero-Cuevas et al. 2003). The
SD test provides a quantitative measure of hand dexterity by challenging participants to
use a precision pinch to compress a slender spring prone to buckling to the point of max-
imal stability. Performance during this task is based on two key components: strength
and dexterity. Here, strength is dened as the ability to produce a sucient amount
of force to compress the spring to solid length (i.e. where all of the spring coils are
touching); anddexterity is dened as the ability to dynamically regulate endpoint force
direction and magnitude to stabilize the spring throughout compression (Valero-Cuevas
et al. 2003).
Subjects rested their arm on a table and compressed the spring using a posture that
was comfortable for them. They were given four attempts (90 seconds each) to try to
92
compress the spring as much as possible before it slipped out of their hand. During the
task, we asked participants to ensure that their 3rd-5th ngers did not assist in the task.
The average maximal compression force reached prior to spring buckling was taken as a
normalized measure of their dexterous performance. We rounded this value to the nearest
tenth and dened this as the subject-specic F
max
. From this, we calculated 40% and
80% of the subject-specic F
max
to be used in the second phase of the study.
6.2.2.2 Task 2: Visuomotor Force Tracking
In the visuomotor force tracking portion of the experiment, subjects were seated com-
fortably in front of a computer monitor with their arm resting on a table with the spring
in their hand. When the trial began, subjects were given real-time visual feedback of
their applied compression force and instructed to squeeze the spring to match on-screen
force levels displayed as a horizontal red line. Breaks were given in between trials when
necessary to reduce fatigue eects. In each trial, ten randomized target forces of 40%
and 80% of their F
max
were presented in 30 second intervals with ve seconds of rest in
between. During the resting periods, the subjects would hold the object with just enough
force to prevent it from dropping. Each target level was presented ve times per trial and
was performed over three trials for a total presentation of 450 seconds per target level.
Subjects then repeated the procedure with a rigid wooden dowel for the same duration
and matched force levels. This two-by-two factorial design yielded four manipulation
conditions: spring-low (SL) , spring-high (SH), dowel-low (DL), and dowel-high (DH).
93
6.2.3 Compliant and Rigid Object Characteristics
Subjects manipulated two objects: a compliant spring and a rigid wooden dowel.
Spring: The spring used in this study is shown in Figure 6.1. The force required to
bring the spring to solid length was approximately 3.7 - 3.8 N. We specically chose a
spring with a low strength requirement (< 15 % maximal precision pinch force) to focus
primarily on cortical drive involved in dexterity demand rather than applied force. The
spring alone measured 4.2 cm long, weighed approximately 2 grams and had a spring
constant of 0.86 N/cm. Custom designed 3-D printed acrylonitrile butadiene styrene
(ABS) plastic end caps were glued to both ends of the spring to create
at surfaces on
which to attach a force transducer. Additional ABS end caps were attached on top of the
sensor for two purposes: (1) to provide a place for subjects to grasp the object and (2)
to serve as a thermal barrier to prevent body heat from adding a bias to the temperature
sensitive transducer. With the addition of the end caps and sensor, the eective length
of the uncompressed compliant object was 5.7 cm as shown in Figure 6.1b.
94
(a) (b)
Figure 6.1: Spring used in the quantication of hand dexterity. (a) Typical precision
pinch hand posture used in the Strength-Dexterity test. Endcaps at either end with a
load cell attached to the index nger side of the spring. (b) Close-up of spring and force
sensor next to a ruler.
Dowel: Subjects also manipulated a rigid wooden dowel measuring 3.8 cm in length
and a diameter of 0.12 cm as shown in Fig. 6.2. With the load cell and ABS end caps,
the eective length of the rigid object was 5.2 cm (Fig. 6.2b). This object was chosen for
its similarity in length, shape, and weight to the spring. The resting length of the spring
was 0.5 cm greater than that of the dowel. A compression force of 0.4 N would be need
to be applied to the spring to match the resting length of the dowel.
95
(a) (b)
Figure 6.2: (a) Wooden dowel with uni-axial load cell attached. (b) Dowel with load cell
and end caps next to a ruler.
6.2.4 Recordings
6.2.4.1 Force
Throughout the SD test and visuomotor force tracking task, normal compression forces
were measured by axing a uni-axial load cell (Measurement Specialties, Hampton, VA)
with double-sided tape to the index nger side of the compliant spring. The circular
load cell measured 0.41 cm in height and 1.27 cm in diameter and aligned perfectly with
the diameter of the objects. Signals from the sensor were dierentially amplied with a
custom designed circuit set to operate in the range of 0 to 5 N. Data were captured using
a USB Data Acquisition (DAQ) system (National Instruments, Austin, TX) sampling at
a rate of 2048 Hz. Prior to data collection the sensor voltage was converted to Newtons by
removing the DC oset and calibrating the load-cell using a four-point linear regression
96
with xed weights. The oset and gain of the load cell were corrected periodically to
ensure accurate force recordings.
6.2.4.2 Electromyography (EMG)
Bipolar surface EMG were collected using a Delsys Bagnoli desktop system (Delsys, Nat-
ick, MA) from the rst dorsal interosseous (FDI) and the abductor pollicis brevis (APB).
Data were amplied from 1000 - 10000 and sampled at a rate of 2048 Hz. The reference
electrode for the recordings was placed on the olecranon of the right arm. Recording
locations were identied by palpating the muscle during force production in the direction
of mechanical action for that particular muscle.
(a) (b)
FDI
Ch. 1
APB
Ch. 2
Netter, F. Atlas of Human Anatomy, 3rd Edition.
Figure 6.3: Intrinsic muscles of the hand that were recorded. (a) First dorsal interosseous.
(b) Abductor policis brevis.
97
6.2.4.3 Electroencephalography (EEG)
64 channels of EEG were recorded at a sampling rate of 2048 Hz (ANT Neuro, Enschede,
The Netherlands). The recording sites remained xed within a
exible cap according
to the international 10-20 system for scalp electrode placement. We ensured repeatable
recordings of cortical areas across subject by taking skull measurements to place electrode
Cz at the cross section of the midway point between the nasion and inion and the midway
point between the left and right tragi of the ear. Electrode impedances were kept below
10 k
with respect to the reference electrode CPz.
98
(d)
(b) (a)
(c)
AFz
CPz
Figure 6.4: eegosports EEG cap. (a) Front view. (b) Left view. (c) Top view. (d) 2-D
layout of all channels. The ground electrode, AFz, is shown in red and the reference
electrode, CPz, is shown in green.
Electrophysiological data were band pass ltered from 10 to 500 Hz and notch ltered
at 60 Hz and its harmonics up to 500 Hz using a 4th order Butterworth lter implemented
99
in MATLAB (Mathworks, Natick, MA) and FieldTrip, a software package for EEG and
EMG analysis (Delorme & Makeig 2004, Oostenveld, Fries, Maris & Schoelen 2011).
Subsequently, EMG were rectied to extract group activity of motor units (Halliday,
Rosenberg, Amjad, Breeze, Conway & Farmer 1995, Mima & Hallett 1999). Data formats
collected from two separate systems were synchronized by conguring the NI-DAQ to send
a trigger pulse to the EEG and EMG systems via a split BNC cable. Custom scripts were
created to read in trigger events and synchronize all data.
6.2.5 Trial Selection
The start of a trial is dened as the time when the on-screen target transitioned from a
resting value to either 40% or 80% of F
max
and its end is dened as the time when the
target value returns back to rest. Trial windows were 30 seconds in duration with a ve
second inter-stimulus interval. Force data during these steady hold phases were visually
examined to determine if the task was performed correctly. Our requirement was that
the hold phase must be within a15% tolerance of the target force value and be held
within this range for at least ve seconds; force proles not meeting these criteria were
excluded from analysis. Synchronized EEG and EMG from each condition that satised
the steady-state criteria were pooled across conditions and subjects and further divided
into ve second long epochs and normalized prior to coherence analysis (Amjad, Halliday,
Rosenberg & Conway 1997). Partitioning the data in this fashion has the overall eect of
weighing each trial equally to overcome problems of non-stationary electrophysiological
recordings in lengthy trials by assuming the data to be quasi-stationary over smaller
time windows. Additionally, this method eectively eliminates coherence bias that favors
100
sections with high EMG amplitude (Amjad et al. 1997, Laine, Yavuz & Farina 2014, Laine,
Negro & Farina 2013, James, Halliday, Stephens & Farmer 2008, Schoelen et al. 2011).
6.2.6 Coherence Analysis
Synchronous oscillations between cortical activity and EMG indicate functional connec-
tivity which can be assessed through coherence analysis (Mima & Hallett 1999, Nunez
et al. 1997). Coherence describes the relationship between two signals through the
strength of their consistent phase lag as a function of frequency. The result is a co-
herence spectrum bounded between 0 and 1 for each frequency of interest. A value of 1
indicates a perfect linear relationship, while a 0 indicates independence. For a given time
series, x (t), let the auto spectra be represented as
P
xx
(f) =
1
L
L
X
i=1
X
i
(f)X

i
(f) (6.1)
where X
i
(f) represents the Fourier transform of the signal of segment i of L, and
indicates the complex conjugate. A similar spectrum exists for the signaly (t), represented
as P
yy
(f). The cross spectrum between the signals x (t) and y (t) is dened by
P
xy
(f) =
1
L
L
X
i=1
X
i
(f)Y

i
(f) (6.2)
Corticomuscular coherence (CMC) is calculated by normalizing the square of the cross-
spectral density between an EMG and an EEG signal by the product of their individual
auto spectral densities (Baker et al. 1997, Nunez et al. 1997) as indicated in Eq. 6.3.
101
C
xy
(f) =
jP
xy
(f)j
2
P
xx
(f)P
yy
(f)
(6.3)
CMC was computed for each EEG-EMG electrode pair using FieldTrip, an open-
source toolbox in MATLAB for the analysis of EEG and MEG data (Oostenveld et al.
2011). We used discrete prolate spheroidal sequences (DPSS) or Slepian tapers (Slepian
& Pollack 1961) for the calculation of the auto and cross spectra. The multitaper method
provides several measures of the spectral estimation by multiplying the data series by a se-
ries of orthogonal tapers prior to calculating the Fourier transform (Pesaran 2008). Three
tapers were used in our analysis, providing a spectral bandwidth of5 Hz (Schoelen
et al. 2011, Maris, Schoelen & Fries 2007).
6.2.7 Selection of EEG Electrodes
To restrict our statistical analysis to a subset of the electrodes, we selected only those
electrodes which showed coherence over the 99% condence limit in the DL task. Our
condence limit for electrode selection set according to Eq. 6.4 (Rosenberg et al. 1989),
where is our desired condence level and N is the total number of tapers used in the
analysis (Schoelen et al. 2011, Maris et al. 2007).
CL () = 1 (1)
1
N1
(6.4)
We then obtained a dierence statistic of standard Z-scores to compare coherence
spectra calculated across conditions and subjects to account for dierences in the number
102
of segments (Schoelen et al. 2011, Kilner et al. 1999, Baker et al. 1997, Laine et al. 2013,
Laine et al. 2014). The transformation is given in Eq. 6.5,
Z (f)
n
=
tanh
1
(C
xyn
(f))
1
2Tn2
q
1
2Tn2
(6.5)
In Eq. 6.5,C
xyn
is the coherence,T
n
is the number of tapers used in condition n and
tanh
1
is the inverse hyperbolic tangent function.
6.2.8 Linear Mixed-Eects Model
A linear mixed-eect (LME) model provides a method of describing a relationship for a
measurable quantity as a function of the sum of weighted independent variables (Winter
2013b, Winter 2013a). We investigated the eects of task condition on mean beta coher-
ence, using a LME model with the following format:
CMC

0
+
1
Condition +
2
(1jParticipant) + (6.6)
whereCMC

is the average beta range coherence, Condition is the xed-eect term,
Participant is the random-eect term, the
n
terms are the coecients for the independent
variables, and is the error. The random-eects term was inserted to account for subject
variability since there were several measurements taken for each condition. Two models
were generated to estimate FDI-EEG and APB-EEG beta coherence.
103
6.3 Results
6.3.1 SD Test Performance
A typical force prole during the SD test for a single trial in a representative subject is
shown in Figure 6.5. Sudden drops in the force trace indicate where the spring buckled
and slipped out of their hand. Each subject was given four 90 second attempts to reach
their maximal compression force. For this subject, their average maximal compression
force prior to spring bucking was calculated to be 2.6 N, shown as the red dotted line.
This resulted in 0:4F
max
= 1:04 N and 0:8F
max
= 2:08 N, however rounding these
values to the nearest tenth gave target low and high forces of 1.0 N and 2.1 N, which are
shown as the purple and green dotted lines, respectively. These values were presented to
the subject for the visuomotor force tracking task.
104
Time (s)
0 10 20 30 40 50 60 70 80
Force (N)
0
0.5
1
1.5
2
2.5
3
Fmax = 2.6 N
40% Fmax = 1.0 N
80% Fmax = 2.1 N
SD Test Trial
Figure 6.5: Sample force prole during the Strength-Dexterity test. The average maximal
compression force for this subject was 2.6 N (red dotted line). As the spring is compressed,
it becomes unstable and dicult to control, resulting in the subject dropping the spring.
These drops are clearly shown as sudden decreases in the force prole. 40% and 80% of
F
max
were calculated to be 1.0 N and 2.1 N, respectively, and shown as the purple and
green dotted lines.
Table 6.1 shows the maximal SD compression force measured for each subject. The
last two columns show the calculated 40% and 80% target values that were used for the
visuomotor force tracking portion of the study. The average compression force was 2.2 N
with a range of 1.8 - 2.8 N and a median of 2.2 N.
Figure 6.6a demonstrates the precision pinch paradigm with the spring. Figure 6.6b
shows a single trial of the visuomotor task for a single representative subject consisting
of ve randomized presentations of each force level. Subjects were asked to match the
target force, which is represented as the black dashed line. The trials are classied into
spring high (SH) and spring low (SL) conditions. The low and high forces, for this
105
Subject Identier Gender Age (years) F
max
(N) 0:4F
max
(N) 0:8F
max
(N)
P001S002 M 36 2.8 1.1 2.2
P002S002 M 40 2.0 0.8 1.6
P003S001 M 27 2.2 0.9 1.8
P004S001 M 36 1.8 0.7 1.4
P005S001 F 26 2.2 0.9 1.8
P006S001 M 26 2.6 1.0 2.1
P007S001 F 33 2.2 0.9 1.8
P008S001 F 32 2.0 0.8 1.6
P009S004 M 32 2.7 1.1 2.2
P010S003 M 32 2.5 1.0 2.0
P011S002 F 27 2.2 0.9 1.8
P012S001 F 26 2.1 0.8 1.7
P013S001 M 31 1.8 0.7 1.4
P014S001 F 26 1.8 0.7 1.4
P015S001 M 25 2.4 1.0 2.0
Table 6.1: F
max
values for each subject. 15 right-handed subjects participated in this
study, six of which were female. Mean age was 30:3 4:6 years. Mean F
max
was 2.2 N,
median was 2.2 N and range was 1:8 2:8 N.
subject, corresponded to 1.0 and 2.1 N, respectively. The red dashed lines indicate the
performance criteria threshold values set to15% of the target value. Only those sections
which were within threshold for a minimum of ve seconds were used for the analysis,
which is shown as the grey shaded areas. In Fig. 6.6c, the same paradigm as in Figure
6.6a is repeated with the spring replaced with a stable wooden dowel. Figure 6.6d shows
the force tracking prole for the dowel. Low and high force levels are the same as those
in Fig. 6.6b. A qualitative comparison between the force variability in the DH and SH
conditions illustrates the diculty in maintaining a steady force with the spring despite
the force levels being identical. Furthermore, it can be seen that the force compression
of the spring dropped outside of the threshold level in the second to last hold phase
(high target). Force compressions with the rigid dowel are clearly distinguished from the
106
compliant spring by the overshoot following the onset of the target force, which did not
appear in the spring condition.
107
DL DL DL DL DL DH DH DH DH DH
Time (s)
0 50 100 150 200 250 300 350
Force (N)
0
1
2
SL SL SL SL SL SH SH SH SH SH
Time (s)
0 50 100 150 200 250 300 350
Force (N)
0
1
2
(a) (b)
(c) (d)
load cell
spring
FDI
load cell
dowel
FDI
Figure 6.6: Visuomotor paradigm and force proles (a) Spring-low precision pinch task. Subjects squeezed a small spring between
their index nger and thumb. Normal index nger forces and bipolar surface EMG from several muscles of the hand were recorded
(only FDI shown here). (b) Typical force trace for a representative subject during the visuomotor task. Black dashed line is
the target force, which, for this subject are 1.0 N and 2.1 N for the low and high target forces, respectively. Red lines indicate
tolerance limits of15 N. The grey area represents valid hold data. The criteria were that the force had to be within the tolerance
limits and be held within that range for a minimum of ve seconds. In the last SH condition, it can be seen that the force fell
out of range and thus this data was not included in the analysis. (c) The spring object is replaced by a wooden dowel. (d) Force
proles for the visuomotor force tracking task with the same force level performance criteria as in (b).
108
6.3.2 Muscle Activation
The activation pattern of the intrinsic hand muscles (FDI and APB) for each task is
shown in Fig. 6.7. The activation for a particular muscle was obtained by dividing the
average value of the rectied EMG during a trial by the mean EMG for that muscle over
all trials and conditions. For the FDI muscle, there was steady increase in the activation
with force applied for both objects. The FDI becomes more involved in the task during
the SH condition than the DH condition despite the force levels being equal. For the APB,
the activity was not signicantly dierent for compression of the dowel, however there is
a trend towards decreasing activation with applied force. This inverse relationship has
been described in literature as the trade-o synergy (Sirin & Patla 1987, Maier & Hepp-
Reymond 1995, Maier & Hepp-Reymond 1994, Huesler, Maier & Hepp-Reymond 2000).
Tasks with the spring show an overall increase in muscle activity as compared to the
dowel.
Figure 6.8 shows a scatter plot indicating the muscle activations across all subjects and
tasks to show the relative contribution of the intrinsic hand muscles for each condition.
The rst principal component (PC) is shown for each condition as a line through the
scatter points which aligns with the direction of the most variance. Figure 6.8a compares
the resting conditions for the spring and dowel. The rst PCs for each task are aligned
during the resting condition. In Figure 6.8b, the slope of the PCs have decreased for each
condition. In the high conditions, depicted in Fig. 6.8c, the PC for the dowel task shows
that there is little APB activation required for this task, however, the PC for the spring
109
task has a slope of approximately one, suggesting equal contribution for both muscles
despite the normal force being the same.
DR DL DH SR SL SH
0
0.5
1
FDI
DR DL DH SR SL SH
0
0.5
1
APB
Figure 6.7: Muscle coordination patterns for the FDI and APB across all conditions.
110
FDI Activation
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
APB Activation
0
0.2
0.4
0.6
0.8
1
DowelRest
SpringRest
FDI Activation
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
APB Activation
0
0.2
0.4
0.6
0.8
1
DowelLow
SpringLow
FDI Activation
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
APB Activation
0
0.2
0.4
0.6
0.8
1
DowelHigh
SpringHigh
(a)
(b)
(c)
Figure 6.8: Scatter plots showing the muscle activation for the FDI and APB for matched
force levels of the objects. The rst principal component for each condition is shown to
capture the direction of the maximum variance. (a) Rest condition. During the period
between the 40% and 80% F
m
ax target force compressions, the activation the muscles
were calculated. The PCs corresponding to the spring and dowel objects are aligned in
this task demonstrating that the FDI and APB muscle are activated similarly during
rest. (b) Low force condition. As in the resting condition, the rst PCs for each object
are aligned, but with a slightly lower slope than in (a). (c) High force condition. During
high compression with the dowel, the activation of the APB muscle is minimal and is
dominated by FDI activity. Compression at the high force of the spring object shows a
slop approximately equal to one, suggesting that there is an equal contribution from both
muscles in order to maintain a constant force on the spring.
111
6.3.3 FDI-EEG Coherence
Figure 6.10a shows a head map of the grand average of the spatial Z-transformed cortico-
muscular coherence between the FDI and all EEG channels during the DL condition. The
map shows only those electrodes which exceeded a 99% Bonferroni corrected threshold
based on the number of EEG electrodes. Four signicant electrodes over the primary
motor cortex (C3, C1, CP3, and CP1) remained after thresholding. Figure 6.10b shows
the average Z-transformed coherence of the four electrodes in Fig. 6.10a. The peak co-
herence of the average appeared in the beta range (grey area of Fig. 6.10b) at 18.15 Hz
with a value of 6.7. Figure 6.10 shows the coherence spectra for the electrodes that were
above signicance for each of the four conditions. The same general trend was present
in these four electrodes for all conditions. The highest beta coherence exists for the SL
condition, while the DL and DH conditions have comparable coherence spectra, and the
SH consistently has the lowest beta CMC.
112
(a)
C3 C1
CP3 CP1
(b)
Figure 6.9: Results for the dowel-low task. (a) Grand average Z-transformed coherence
head map for the FDI muscle to all EEG electrodes for the DL task. A 99% Bonfer-
onni corrected Z-score threshold was applied to the head map to account for multiple
comparisons based on the number of EEG channels. The four electrodes with signicant
coherence above the threshold were C1, C3, CP1, and CP3 with respective Z-transformed
coherence values of 4.81, 5.26, 4.66, and 4.54. (b) Average coherence spectra for the four
electrodes shown in (a). The beta frequency range (15 - 30 Hz) is shown as the grey
shaded area. Peak coherence for the average was 6.7 at a frequency of 18.15 Hz.
113
C3
Frequency (Hz)
20 40 60 80 100
Coherence
(Z-transformed)
0
2
4
6
8
10
12
C1
Frequency (Hz)
20 40 60 80 100
Coherence
(Z-transformed)
0
2
4
6
8
10
12
CP1
Frequency (Hz)
20 40 60 80 100
Coherence
(Z-transformed)
0
2
4
6
8
10
12
DL
SL
DH
SH
CP3
Frequency (Hz)
20 40 60 80 100
Coherence
(Z-transformed)
0
2
4
6
8
10
12
(a) (b)
(c) (d)
Figure 6.10: Coherence spectra for the four EEG electrodes with average beta range
coherence values over the threshold limit. In each plot, all four conditions. The respective
condition and trace color are as follows: DL - blue, SH - red, DH - yellow, and SH
- purple. Red dashed line in each gure corresponds to the 99% Bonferroni corrected
threshold value and the grey shaded areas indicate the beta frequency range (15 - 30 Hz).
Individual coherence spectra for each condition for electrode (a) C3, (b) C1, (c) CP3, and
(d) CP1.
114
6.3.4 LME Model
Figure 6.11 shows the eects of each condition on FDI-EEG (Fig. 6.11a) and APB-EEG
(Fig. 6.11b) beta range coherence. Using an F-test, we compared the coecients for
the eects of the SL and DL conditions on beta CMC and found that there was no
statistical dierence in the eects of the low conditions in either FDI-EEG (p = 0:24502)
or APB-EEG (p = 0:2357) coherence. A comparison between the coecients for the DH
and SH tasks, however, showed a clear signicance in eect of beta CMC in FDI-EEG
(p = 6:859 10
8
) and APB-EMG (p = 1:6889 10
5
).
115
Condition
DL SL DH SH
Average Beta Coherence
(Z-transformed)
1.2
1.3
1.4
1.5
1.6
n.s. ***
FDI
Condition
DL SL DH SH
Average Beta Coherence
(Z-transformed)
1.2
1.3
1.4
1.5
1.6
n.s. ***
APB
*** p<0.001
(a) (b)
Figure 6.11: Results of the linear mixed-eects model. The model was constructed to
predict mean beta range coherence using Condition as the xed-eect and Participant
as the random eect. In each bar graph, the mean beta range CMC is shown on the
vertical axis and condition is on the horizontal axis. Standard error bars are included
for each condition and the indicators above the bars represent the statistical dierence
in the linear mixed eects coecients as determined using an F-test. n.s. indicates that
there was no signicant dierence in eect between two conditions and indicates
that the p-value was less than 0.001. Linear mixed-eect models for the prediction of (a)
FDI-EEG beta coherence and (b) APB-EEG beta coherence.
6.3.5 Power
We investigated the individual EMG beta power for the intrinsic hand muscles and the
EEG electrodes of interest. Figure 6.12 shows the average beta power in the the FDI and
APB muscles. There is a large increase in beta FDI (Fig. 6.12a) and APB (Fig. 6.12b)
power during the SH task compared to all other tasks. There is a highly signicant
116
dierence in the eect of the SH vs. the DH task on EMG power (p<0.001) in both
intrinsic hand muscles. The ring rates of the motor unit action potentials as well as
their shape during compression of the spring at the high force levels could in
uence
spectral content in the beta frequency range leading to the signicant increase in the
beta range power for the FDI and APB.
(a)
(b)
Condition
DL SL DH SH
Average Beta Power
× 10
-6
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
FDI
Condition
DL SL DH SH
Average Beta Power
× 10
-5
0
0.2
0.4
0.6
0.8
1
1.2
APB
DL SL DH SH
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
RMSE of Force
Figure 6.12: Average beta power for the FDI and APB muscles for the DL, SL, DH and
SH conditions. (a) FDI power for all four conditions. The largest beta power for the FDI
was apparent during the SH task. The power during the SH was signicantly higher than
for all other conditions. (b)
117
Additionally, we explored beta power for each condition in the four scalp electrodes.
Figure 6.13 show a representative example for electrode C3 since all other electrodes had
a similar trend. The statistics for this electrode showed that there was no signicant
dierence between the DH and SH conditions for C3 beta power, and in fact there there
were no dierences in eect for any condition.
Condition
DL SL DH SH
Average Beta Power
0
0.002
0.004
0.006
0.008
0.01
0.012
C3
Figure 6.13: Average beta power for the EEG electrode C3 for the DL, SL, DH and SH
conditions.
Based on literature ndings, power does not necessarily have an eect on coherence.
This has been demonstrated in studies which have used drugs to modify cortical power and
118
coherence. In one study diazepam was used to increase beta power in the motor cortex,
however there was no change in beta corticomuscular coherence (Baker & Baker 2003).
The second study, which used carbamazepine (CBZ) on epilepsy patients, showed that
after taking CBZ there was an increase in coherence, however no signicant change in
EEG power (Riddle et al. 2004). This demonstrates a disconnect between spectral power
and coherence.
6.3.6 Root Mean Square Error of Force
Lastly, we investigated the performance of the force variability in the tasks for all four
conditions. Figure 6.14 shows the root mean square error (RMSE) of compression force to
target force for all four conditions. As would be expected, force variability increased dur-
ing higher force production. Furthermore, since there was no dierence in force variability
during high force compression for either object, hence this rules out force variability as
the leading cause of the change in corticomuscular and inter muscular coherence.
119
DL SL DH SH
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
RMSE of Force
Figure 6.14: Root mean square error of compression force to target force.
6.4 Discussion
The role of the sensorimotor cortex during dexterous manipulation in humans is a topic of
intense interest. In this study, we present for the rst time, a measure of corticomuscular
coherence during unstable and unpredictable dynamic manipulation. To our knowledge,
the functional signicance of cortico-muscular coupling under these conditions has never
been unexplored. fMRI studies under these dexterously demanding conditions have shown
a continuous and context-sensitive involvement of the brain, depending on the functional
parameters of the task (Mosier et al. 2011). We have extended this approach to include
cortical coupling to hand muscles and emphasize tasks where the level of instability can
be regulated. Our results indicate that in the DL and DH conditions, signicant corti-
comuscular coherence appeared in M1, however, their magnitudes were not signicantly
120
dierent. The highest magnitude of beta CMC was recorded over M1 during the SL con-
dition with coherence extension into the supplementary motor area (SMA). Prior studies
have shown that object compliance aects coherence magnitude (Kilner et al. 2000, Riddle
& Baker 2006). During ne manipulation, SMA coherence was also shown to be increased
in the beta frequency range (Chen et al. 2013). However, no study has investigated CMC
under unstable conditions.
Investigators have reported that the index nger and thumb are independently con-
trolled (Schieber & Santello 2004). Based on this idea, the model in Fig. 6.15 demon-
strates the possible underlying principles that may enable stable versus dexterous manip-
ulation. When performing a task that requires minimal sensorimotor integration, such as
the DL, DH and SL tasks, we speculate that the cortical representations of the FDI and
APB operate in synchrony to perform a stabilizing task involving a predominantly co-
contraction strategy of the muscles. This idea is depicted in Fig. 6.15a. The underlying
neural oscillators controlling the FDI and APB are coupled during stable manipulation.
In addition, there would be strong coupling in the FDI-APB coherence, representative of
a common neural drive to both muscles. However, in the unstable SH condition, the in-
dex nger and thumb must operate more independently thereby disrupting the coupling
between the FDI and APB cortical representations. Since the common drive between
the FDI and APB would no longer exist, the coherence between FDI and APB would
decrease.
121
FDI APB
Cortex
Periphery
FDI APB
Cortex
Periphery
Stable Domain Unstable Domain
(a) (b)
Figure 6.15: Model of cortical drive to hand muscles during stable and unstable tasks.
In the stable domain, ares of the cortex representing the FDI and APB are driven by
underlying neural oscillators.
Figure 6.16 shows the FDI-APB coherence across the low and high force production
conditions. These muscles were chosen due to their critical involvement in the mainte-
nance of the static force production. As expected, the peak Z-transformed coherence in
all four conditions appeared in the beta range with the largest value existing in the SL
condition at a frequency of 23.3 Hz and a coherence value of 10.5. Coherence spectra
during compression tasks with the dowel were similar with the higher force level being
slightly greater than that of the low force compression with the dowel. In the SH con-
dition, the beta coherence was much lower than any of the other conditions and had a
122
peak value of 2.3 at 24.4 Hz. These EMG-EMG coherence spectra supports the idea that
the cortical drive to the FDI and APB are more independently controlled in the presence
of instability. To investigate the disruption in neural oscillatory coupling between the
FDI and APB areas of the cortex, higher density cortical electrodes would be required.
Nonetheless, the decrease in FDI-EEG, APB-EEG and FDI-APB coherence during the
SH task help to support our model
123
Frequency (Hz)
10 20 30 40 50 60 70 80 90 100
Coherence
(Z-transformed)
0
2
4
6
8
10
12
DL
DH
SL
SH
Max = 5.7485
f = 24.8186
Max = 6.5896
f = 24.631
Max = 10.4958
f = 23.3182
Max = 2.3271
f = 24.4435
Figure 6.16: EMG to EMG coherence between the rst dorsal interosseous and the
abductor pollicis brevis. The SL condition is shown as the yellow trace with a peak
coherence of 10.5 at 23.3 Hz. The DL (blue trace) and DH (red trace) have similar peak
coherence values within the beta range at 5.7 and 6.6 at 24.8 and 24.6 Hz, respectively.
The SH condition (purple trace) has the lowest overall beta range coherence with a max
value of 2.3 at 24.4 Hz
124
We performed an additional analysis to investigate the changes in gamma frequency
CMC in the task. Figure 6.17 shows non-thresholded head maps of the grand averages
of the spatial Z-transformed corticomuscular coherence between the FDI and all EEG
channels during the DL (Fig. 6.17a) and SL (Fig. 6.17b) conditions. For both low force
tasks, a signicant peak in beta coherence appears over the left primary motor cortex.
The measured peak coherence values were 6.42 and 8.23 for the DL and SL condition,
respectively. In the SL condition (Fig. 6.17b), coherence extends medially into electrodes
Cz and FCz, over the supplementary motor area.
0
2
4
6
8
DL SL
(a) (b)
Coherence
(Z-transformed)
Figure 6.17: Grand average FDI-EEG beta coherence head maps during the low force
conditions. (a) DL condition. Peak CMC appears over contralateral M1. (b) SL condi-
tion. Peak coherence exists over contralateral M1 with greater magnitude than in the DL
condition. Coherence extends medially into the supplementary motor area (i.e. electrodes
Cz and FCz).
Based on the spatial shift into the SMA in the SL task, we extended our investigation
to include changes in gamma range CMC over electrode C3 (sensorimotor) and Cz (SMA).
Figure 6.18 shows the change in average gamma coherence between objects at matched
125
force levels was calculated for each subject. A Wilcoxson rank sum test was used to show
that the average gamma coherence signicantly increased only in the SMA electrode (Cz)
during the SH task as compared to the DH task (p = 0:007).
Given the unavoidable delays in the nervous system, it is poorly understood whether
and how the cerebral cortex can contribute to the time-critical control of ngertip forces
during dexterous dynamic manipulation. Corticomuscular coherence gives insight into
how the brain communicates with the body. Given that our interaction with objects
involves a multitude of complex force productions which dynamically change with task
demands, it is necessary to not limit our analysis to the beta range coherence which
is associated with static force. Furthermore, since strategic planning is involved, the
investigation of CMC in cortical regions other than the M1 and S1 are imperative. In
this study, we did not have any premonitions about what frequency range to investigate
nor the cortical areas to include since this task varied drastically from previous literature.
The goal of the tasks in this study was to maintain a constant level of force. From the
perspective of the force recordings, there did not exist a noticeable dierence in the force
production during the SH and DH tasks. However, the nervous system was faced with
the added challenge of interacting with an unstable object. As a result, we determined
that the cortex communicates with the musculature in a far dierent manner when the
need to exert a constant force on an unstable spring is required. This suggests that the
neural strategies involved are not classied into a uni-dimensional goal directed objective,
such as constant force production. Instead the cortex chooses to engage higher frequency
ranges involving context-sensitive cortical circuits which are dependent on the temporal,
dynamic and dexterity demands of the task. Lastly, it is likely that the cortex requires the
126
..
DH SH
1
1.5
2
2.5
Electrode C3
p = 0.6783
DH SH
0.8
1
1.2
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
Electrode Cz
p = 0.0070162
DL SL
0.8
1
1.2
1.4
1.6
1.8
2
2.2
2.4
Electrode C3
p = 0.70892
DL SL
0.8
1
1.2
1.4
1.6
1.8
2
2.2
Electrode Cz
p = 0.6783
DR SR
0.5
1
1.5
2
Electrode C3
p = 0.6783
DR SR
0.6
0.8
1
1.2
1.4
1.6
1.8
2
2.2
2.4
Electrode Cz
p = 0.77155
Figure 6.18: Change in gamma coherence across subjects and the rest, low and high conditions in electrodes C3 (over sensorimotor)
and Cz (over SMA). Using a Wilcoxon rank sum test, it was determined that there was no statistical dierence in the change
in gamma coherence in any of the C3 electrodes for matched force conditions. However in the SMA, there existed a signicant
increase in the average gamma coherence for the spring-high task as compared to the dowel-high task.
127
use of higher frequency communication to account for the instabilities and uncertainties
seen in everyday object manipulation.
Acknowledgements
Co-authors Christopher M. Laine, Jason J. Kutch and Francisco J. Valero-Cuevas. Data
collected for this study was done in the Applied Mathematics and Physiology Lab under
the direction of Jason J. Kutch. This material is in part based upon work supported
by NSF Grant EFRI-COPN 0836042, NIH Grant R01-052345, NIH Grant R01-050520 to
FVC, and Supplement R01-050520-W1 to AR.
128
Chapter 7
Conclusions and Future Work
The human hand is an amazingly complex and fascinating apparatus. There appar-
ently seems to be no limit to its capabilities. Our hands have enabled humans to build
skyscrapers that pierce the clouds, paint breathtaking sceneries, compose soothing music;
it is strong and robust enough to grip a sledgehammer with immense strength to drive
a spike into the ground, yet is nimble enough to thread a needle. However, the neural
mechanisms that give us the ability to precisely control nger movements and forces is
by no means a trivial one.
In this dissertation, one piece of the puzzle in the neural control of movement was
presented. By investigating the temporal relationship two electrophysiological measure-
ments, EEG and EMG, it is possible to gain insight into how cortical activity is trans-
formed into motor precision ngertip control. Oscillations are an abundant phenomenon
in the universe and are present in the nervous system as well. Therefore, the role of these
oscillations under real-world conditions of dexterous manipulation are essential to our
understanding our own bodies.
129
Continued research in this area and similar elds will undoubtedly lead to the advance-
ment of brain-computer interfaces and robotic hands. It is my hope that, as technological
advances are made, and it becomes possible to probe and track neural activity from cortex
to muscles to endpoint force along each step of the way, we will come closer to the devel-
opment of a truly dexterous prosthetic hand that is indistinguishable from the biological
hand.
130
Bibliography
Alexander, R. M. (1997), `A minimum energy cost hypothesis for human arm trajectories',
Biological cybernetics 76(2), 97{105.
Amjad, A., Halliday, D., Rosenberg, J. & Conway, B. (1997), `An extended dierence
of coherence test for comparing and combining several independent coherence esti-
mates: theory and application to the study of motor units and physiological tremor',
Journal of neuroscience methods 73(1), 69{79.
Andreasson, J., Ekstrom, M., Fard, A., Castao, J. & Johnson, T. (2002), `Remote system
for patient monitoring using bluetooth', 1, 304{307.
Ashe, J. (1997), `Force and the motor cortex', Behavioural Brain Research87(2), 255{269.
Baker, M. R. & Baker, S. N. (2003), `The eect of diazepam on motor cortical oscil-
lations and corticomuscular coherence studied in man', The Journal of physiology
546(3), 931{942.
Baker, S. N. (2007), `Oscillatory interactions between sensorimotor cortex and the pe-
riphery', Current opinion in neurobiology 17(6), 649{655.
Baker, S. N., Chiu, M. & Fetz, E. E. (2006), `Aerent encoding of central oscillations in
the monkey arm', Journal of Neurophysiology 95(6), 3904{3910.
Baker, S. N., Pinches, E. M. & Lemon, R. N. (2003), `Synchronization in monkey motor
cortex during a precision grip task. ii. eect of oscillatory activity on corticospinal
output', Journal of Neurophysiology 89(4), 1941{1953.
Baker, S., Olivier, E. & Lemon, R. (1997), `Coherent oscillations in monkey motor cortex
and hand muscle emg show taskdependent modulation', The Journal of Physiology
501(1), 225{241.
Bear, M., Connors, B. W., Paradiso, M., Bear, M., Connors, B. & Neuroscience, M.
(1996), `Exploring the brain', Neuroscience: Williams and Wilkins .
Bennett, K. & Lemon, R. (1996), `Corticomotoneuronal contribution to the fractionation
of muscle activity during precision grip in the monkey', Journal of Neurophysiology
75(5), 1826{1842.
Berger, H. (1929), `ber das elektrenkephalogramm des menschen', European Archives of
Psychiatry and Clinical Neuroscience 87(1), 527{570.
131
Bernhard, C., Bohm, E. & Petersen, I. (1953), `Investigations on the organization of the
corticospinal system in monkeys', Acta Physiologica Scandinavica29(S106), 79{105.
Bernstein, N. A. (1967), `The co-ordination and regulation of movements'.
Bertrand, O., Perrin, F. & Pernier, J. (1985), `A theoretical justication of the aver-
age reference in topographic evoked potential studies', Electroencephalography and
Clinical Neurophysiology/Evoked Potentials Section 62(6), 462{464.
Blok, J. H., Stegeman, D. F. & van Oosterom, A. (2002), `Three-layer volume conductor
model and software package for applications in surface electromyography', Annals of
Biomedical Engineering 30(4), 566{577.
Boiten, F., Sergeant, J. & Geuze, R. (1992), `Event-related desynchronization: the ef-
fects of energetic and computational demands', Electroencephalography and clinical
neurophysiology 82(4), 302{309.
Brown, P. (2000), `Cortical drives to human muscle: the piper and related rhythms',
Progress in neurobiology 60(1), 97{108.
Brown, P., Salenius, S., Rothwell, J. C. & Hari, R. (1998), `Cortical correlate of the piper
rhythm in humans', J Neurophysiol 80(6), 2911{2917.
Burgar, C. G., Valero-Cuevas, F. J. & Hentz, V. R. (1997), `Fine-wire electromyographic
recording during force generation: Application to index nger kinesiologic studies1',
American journal of physical medicine and rehabilitation 76(6), 494{501.
Chakarov, V., Naranjo, J. R., Schulte-Mnting, J., Omlor, W., Huethe, F. & Kristeva, R.
(2009), `Beta-range eeg-emg coherence with isometric compensation for increasing
modulated low-level forces', Journal of neurophysiology 102(2), 1115{1120.
Chatrian, G. E., Petersen, M. C. & Lazarte, J. A. (1959), `The blocking of the rolandic
wicket rhythm and some central changes related to movement', Electroencephalogra-
phy and clinical neurophysiology 11(3), 497{510.
Chen, S., Entakli, J., Bonnard, M., Berton, E. & De Graaf, J. B. (2013), `Functional
corticospinal projections from human supplementary motor area revealed by corti-
comuscular coherence during precise grip force control', Plos One 8(3).
Cianchetti, F. A. & Valero-Cuevas, F. J. (2010), `Anticipatory control of motion-to-
force transitions with the ngertips adapts optimally to task diculty', J Neu-
rophysiol 103(1), 108{16. Cianchetti, Flor A Valero-Cuevas, Francisco J AR-
050520/AR/NIAMS NIH HHS/ AR-052345/AR/NIAMS NIH HHS/ J Neurophysiol.
2010 Jan;103(1):108-16. Epub 2009 Nov 4.
Conway, B., Halliday, D., Farmer, S., Shahani, U., Maas, P., Weir, A. & Rosenberg, J.
(1995), `Synchronization between motor cortex and spinal motoneuronal pool during
the performance of a maintained motor task in man', The Journal of physiology
489(Pt 3), 917{924.
132
Cosmanescu, A., Miller, B., Magno, T., Ahmed, A. & Kremenic, I. (2006), `Design and
implementation of a wireless (bluetooth) four channel bio-instrumentation amplier
and digital data acquisition device with user-selectable gain, frequency, and driven
reference', pp. 2053{2056.
Crone, N. E., Miglioretti, D. L., Gordon, B., Sieracki, J. M., Wilson, M. T., Uematsu,
S. & Lesser, R. P. (1998), `Functional mapping of human sensorimotor cortex with
electrocorticographic spectral analysis. i. alpha and beta event-related desynchro-
nization', Brain 121(12), 2271{2299.
Da Silva, F. L. & Pfurtscheller, G. (1999), `Basic concepts on eeg synchronization and
desynchronization'.
Davare, M., Lemon, R. & Olivier, E. (2008), `Selective modulation of interactions be-
tween ventral premotor cortex and primary motor cortex during precision grasping
in humans', The Journal of physiology 586(11), 2735{2742.
Day, S. J. (1997), The properties of electromyogram and force in experimental and com-
puter simulations of isometric muscle contraction: data from an acute cat prepara-
tion, University of Calgary.
Dayanidhi, S., Hedberg, ., Valero-Cuevas, F. J. & Forssberg, H. (2013), `Developmental
improvements in dynamic control of ngertip forces last throughout childhood and
into adolescence', Journal of neurophysiology 110(7), 1583{1592.
Deiber, M.-P., Honda, M., Ibaez, V., Sadato, N. & Hallett, M. (1999), `Mesial motor areas
in self-initiated versus externally triggered movements examined with fmri: eect of
movement type and rate', Journal of neurophysiology 81(6), 3065{3077.
Delorme, A. & Makeig, S. (2004), `Eeglab: an open source toolbox for analysis of single-
trial eeg dynamics including independent component analysis', Journal of neuro-
science methods 134(1), 9{21.
Desmurget, M. & Grafton, S. (2000), `Forward modeling allows feedback control for fast
reaching movements', Trends in cognitive sciences 4(11), 423{431.
Donoghue, J. P., Sanes, J. N., Hatsopoulos, N. G. & Gal, G. (1998), `Neural discharge
and local eld potential oscillations in primate motor cortex during voluntary move-
ments', Journal of Neurophysiology 79, 159{173.
Dujardin, K., Derambure, P., Defebvre, L., Bourriez, J., Jacquesson, J. & Guieu,
J. (1993), `Evaluation of event-related desynchronization (erd) during a recogni-
tion task: eect of attention', Electroencephalography and clinical neurophysiology
86(5), 353{356.
Dum, R. P. & Strick, P. L. (1991), `The origin of corticospinal projections from the
premotor areas in the frontal lobe', J Neurosci 11(3), 667{689.
133
Dum, R. P. & Strick, P. L. (2005), `Frontal lobe inputs to the digit representations of the
motor areas on the lateral surface of the hemisphere', The Journal of neuroscience
25(6), 1375{1386.
Duncan, P. W., Propst, M. & Nelson, S. G. (1983), `Reliability of the fugl-meyer assess-
ment of sensorimotor recovery following cerebrovascular accident', Physical therapy
63(10), 1606{1610.
Ehrsson, H. H., Fagergren, A., Jonsson, T., Westling, G., Johansson, R. S. & Forssberg,
H. (2000), `Cortical activity in precision-versus power-grip tasks: an fmri study',
Journal of neurophysiology 83(1), 528{536.
Faisal, A., Stout, D., Apel, J. & Bradley, B. (2010), `The manipulative complexity of
lower paleolithic stone toolmaking', PloS one 5(11), e13718.
Farina, D., Arendt-Nielsen, L., Merletti, R. & Graven-Nielsen, T. (2002), `Assessment
of single motor unit conduction velocity during sustained contractions of the tib-
ialis anterior muscle with advanced spike triggered averaging', J Neurosci Methods
115(1), 1{12.
Farina, D., Mesin, L., Martina, S. & Merletti, R. (2004), `A surface emg generation
model with multilayer cylindrical description of the volume conductor', Biomedical
Engineering, IEEE Transactions on 51(3), 415{426.
Farina, Dario; Merletti, R. (2001), `A novel approach for precise simulation of emg signal
detected by surface electrodes'.
Farmer, S., Bremner, F., Halliday, D., Rosenberg, J. & Stephens, J. (1993), `The frequency
content of common synaptic inputs to motoneurones studied during voluntary iso-
metric contraction in man', The Journal of physiology 470(1), 127{155.
Feige, B., Aertsen, A. & Kristeva-Feige, R. (2000), `Dynamic synchronization between
multiple cortical motor areas and muscle activity in phasic voluntary movements',
Journal of Neurophysiology 84(5), 2622{2629.
Fetz, E. E. & Cheney, P. D. (1980), `Postspike facilitation of forelimb muscle activity by
primate corticomotoneuronal cells', J Neurophysiol 44(4), 751{772.
Fisher, R. J., Galea, M. P., Brown, P. & Lemon, R. N. (2002), `Digital nerve anaesthe-
sia decreases emg-emg coherence in a human precision grip task', Exp Brain Res
145(2), 207{14.
Flash, T. & Hogan, N. (1985), `The coordination of arm movements: an experimentally
conrmed mathematical model', The journal of Neuroscience 5(7), 1688{1703.
Fuglevand, A. J., Winter, D. A., Patla, A. E. & Stashuk, D. (1992), `Detection of motor
unit action potentials with surface electrodes: in
uence of electrode size and spacing',
Biological Cybernetics 67(2), 143{153.
134
Galla, C., de Graaf, J. B., Bonnard, M. & Pailhous, J. (2005), `High level of dexterity:
dierential contributions of frontal and parietal areas', Neuroreport 16(12), 1271{
1274.
Gootzen, T., Stegeman, D. & Van Oosterom, A. (1991), `Finite limb dimensions and nite
muscle length in a model for the generation of electromyographic signals', Electroen-
cephalography and Clinical Neurophysiology/Evoked Potentials Section 81(2), 152{
162.
Gross, J., Pollok, B., Dirks, M., Timmermann, L., Butz, M. & Schnitzler, A. (2005),
`Task-dependent oscillations during unimanual and bimanual movements in the hu-
man primary motor cortex and sma studied with magnetoencephalography', Neu-
roimage 26(1), 91{98.
URL: http://www.sciencedirect.com/science/article/pii/S1053811905000480
Grosse, P. & Brown, P. (2003), `Acoustic startle evokes bilaterally synchronous oscillatory
emg activity in the healthy human', Journal of neurophysiology 90(3), 1654{1661.
Haller, S., Chapuis, D., Gassert, R., Burdet, E. & Klarhfer, M. (2009), `Supplementary
motor area and anterior intraparietal area integrate negraded timing and force
control during precision grip', European Journal of Neuroscience30(12), 2401{2406.
Halliday, D. M., Conway, B. A., Farmer, S. F. & Rosenberg, J. R. (1998), `Using electroen-
cephalography to study functional coupling between cortical activity and electromyo-
grams during voluntary contractions in humans', Neuroscience letters 241(1), 5{8.
Halliday, D. & Rosenberg, J. (2000), `On the application, estimation and interpretation of
coherence and pooled coherence', Journal of neuroscience methods100(1), 173{174.
Halliday, D., Rosenberg, J., Amjad, A., Breeze, P., Conway, B. & Farmer, S. (1995),
`A framework for the analysis of mixed time series/point process datatheory and
application to the study of physiological tremor, single motor unit discharges and
electromyograms', Progress in biophysics and molecular biology 64(2), 237{278.
Hari, R. & Salenius, S. (1999), `Rhythmical corticomotor communication', Neuroreport
10(2), R1{10.
He, S.-Q., Dum, R. P. & Strick, P. (1995), `Topographic organization of corticospinal pro-
jections from the frontal lobe: motor areas on the medial surface of the hemisphere',
Journal of Neuroscience 15(5), 3284{3306.
Hjorth, B. (1975), `An on-line transformation of eeg scalp potentials into orthogonal
source derivations', Electroencephalography and clinical neurophysiology 39(5), 526{
530.
Hjorth, B. (1980), `Source derivation simplies topographical eeg interpretation', Amer-
ican Journal of EEG Technology 20(3), 121{132.
135
Holmstrom, L., de Manzano, O., Vollmer, B., Forsman, L., Valero-Cuevas, F. J., Ullen, F.
& Forssberg, H. (2011), `Dissociation of brain areas associated with force production
and stabilization during manipulation of unstable objects', Exp Brain Res 215(3-
4), 359{67.
Huesler, E. J., Maier, M. A. & Hepp-Reymond, M.-C. (2000), `Emg activation patterns
during force production in precision grip. iii. synchronisation of single motor units',
Experimental brain research 134(4), 441{455.
Humphrey, D. R., Schmidt, E. & Thompson, W. (1970), `Predicting measures of motor
performance from multiple cortical spike trains', Science 170(3959), 758{762.
Hutcheon, B. & Yarom, Y. (2000), `Resonance, oscillation and the intrinsic frequency
preferences of neurons', Trends in neurosciences 23(5), 216{222.
James, L. M., Halliday, D. M., Stephens, J. A. & Farmer, S. F. (2008), `On the devel-
opment of human corticospinal oscillations: agerelated changes in eegemg coherence
and cumulant', European Journal of Neuroscience 27(12), 3369{3379.
Jancke, L., Specht, K., Mirzazade, S., Loose, R., Himmelbach, M., Lutz, K. & Shah,
N. J. (1998), `A parametric analysis of the rate eect in the sensorimotor cortex:
a functional magnetic resonance imaging analysis in human subjects', Neuroscience
letters 252(1), 37{40.
Jasper, H. & Andrews, H. (1938), `Brain potentials and voluntary muscle activity in man',
Journal of neurophysiology .
Jasper, H. & Stefanis, C. (1965), `Intracellular oscillatory rhythms in pyramidal tract
neurones in the cat', Electroencephalography and clinical neurophysiology18(6), 541{
553.
Jebsen, R. H., Taylor, N., Trieschmann, R., Trotter, M. J. & Howard, L. A. (1969), `An
objective and standardized test of hand function', Archives of physical medicine and
rehabilitation 50(6), 311.
Johansson, R. S. & Flanagan, J. R. (2009), `Coding and use of tactile signals from the
ngertips in object manipulation tasks', Nature Reviews Neuroscience 10(5), 345{
359.
Johansson, R. & Westling, G. (1984), `Roles of glabrous skin receptors and sensorimotor
memory in automatic control of precision grip when lifting rougher or more slippery
objects', Experimental brain research 56(3), 550{564.
Jonsson, B., Omfeldt, M. & Rundgren, A. (1967), `Discomfort from the use of wire
electrodes for electromyography', Electromyography 8(1), 5{17.
Kawato, M. (1999), `Internal models for motor control and trajectory planning', Current
opinion in neurobiology 9(6), 718{727.
136
Kilner, J., Baker, S., Salenius, S., Jousmki, V., Hari, R. & Lemon, R. (1999), `Task-
dependent modulation of 15-30 hz coherence between rectied emgs from human
hand and forearm muscles', The Journal of Physiology 516(2), 559{570.
Kilner, J. M., Baker, S. N., Salenius, S., Hari, R. & Lemon, R. N. (2000), `Human cortical
muscle coherence is directly related to specic motor parameters', The Journal of
neuroscience 20(23), 8838{8845.
Kilner, J. M., Fisher, R. J. & Lemon, R. N. (2004), `Coupling of oscillatory activity
between muscles is strikingly reduced in a deaerented subject compared with normal
controls', Journal of neurophysiology 92(2), 790{796.
Kristeva-Feige, R., Fritsch, C., Timmer, J. & Lcking, C.-H. (2002), `Eects of atten-
tion and precision of exerted force on beta range eeg-emg synchronization during a
maintained motor contraction task', Clinical Neurophysiology 113(1), 124{131.
Kristeva, R., Patino, L. & Omlor, W. (2007), `Beta-range cortical motor spectral power
and corticomuscular coherence as a mechanism for eective corticospinal interaction
during steady-state motor output', Neuroimage 36(3), 785{792.
KuhtzBuschbeck, J. P., Ehrsson, H. H. & Forssberg, H. (2001), `Human brain activity in
the control of ne static precision grip forces: an fmri study', European Journal of
Neuroscience 14(2), 382{390.
Kuypers, H. (1960), `Central cortical projections to motor and somato-sensory cell
groups', Brain 83(1), 161{184.
Laine, C. M., Negro, F. & Farina, D. (2013), `Neural correlates of task-related changes
in physiological tremor', Journal of neurophysiology 110(1), 170{176.
Laine, C. M., Yavuz, . U. & Farina, D. (2014), `Taskrelated changes in sensorimotor inte-
gration in
uence the common synaptic input to motor neurones', Acta Physiologica
211(1), 229{239.
Lebedev, M. & Wise, S. (2000), `Oscillations in the premotor cortex: single-unit activity
from awake, behaving monkeys', Experimental brain research 130(2), 195{215.
Lemon, R., Baker, S., Davis, J., Kirkwood, P., Maier, M. & Yang, H. (1998), The impor-
tance of the cortico-motoneuronal system for control of grasp, in `Novartis Found
Symp', Vol. 218, pp. 202{215.
Lemon, R. & Mantel, G. (1989), `The in
uence of changes in discharge frequency of
corticospinal neurones on hand muscles in the monkey', The Journal of physiology
413(1), 351{378.
Lemon, R. N. (2008), `Descending pathways in motor control', Annu. Rev. Neurosci.
31, 195{218.
Lemon, R. N., Johansson, R. & Westling, G. (1995), `Corticospinal control during reach,
grasp, and precision lift in man', the Journal of Neuroscience 15(9), 6145{6156.
137
Leocani, L., Toro, C., Manganotti, P., Zhuang, P. & Hallett, M. (1997), `Event-related
coherence and event-related desynchronization/synchronization in the 10 hz and 20
hz eeg during self-paced movements', Electroencephalography and Clinical Neuro-
physiology/Evoked Potentials Section 104(3), 199{206.
Llins, R. R. (1988), `The intrinsic electrophysiological properties of mammalian neurons:
insights into central nervous system function', Science 242(4886), 1654{1664.
Loeb, G., Brown, I. & Cheng, E. (1999), `A hierarchical foundation for models of senso-
rimotor control', Experimental brain research 126(1), 1{18.
Lowery, M. M., Stoykov, N. S., Dewald, J. & Kuiken, T. (2004), `Volume conduction in an
anatomically based surface emg model', Biomedical Engineering, IEEE Transactions
on 51(12), 2138{2147.
Lowery, M. M., Stoykov, N. S., Ta
ove, A. & Kuiken, T. A. (2002), `A multiple-layer
nite-element model of the surface emg signal', IEEE Transactions on Biomedical
Engineering 49(5).
MacKay, W. A. (2005), `Wheels of motion: oscillatory potentials in the motor cortex',
Motor Cortex in Voluntary Movements: A Distributed System for Distributed Func-
tions pp. 181{211.
Maier, M. A. & Hepp-Reymond, M.-C. (1995), `Emg activation patterns during force
production in precision grip', Experimental Brain Research 103(1), 108{122.
Maier, M. & Hepp-Reymond, M. (1994), `Emg activation patterns during force produc-
tion in precision grip. ii. muscular synergies in the spatial and temporal domain',
Experimental brain research 103(1), 123{136.
Maris, E., Schoelen, J.-M. & Fries, P. (2007), `Nonparametric statistical testing of co-
herence dierences', Journal of neuroscience methods 163(1), 161{175.
Marsden, J., Werhahn, K., Ashby, P., Rothwell, J., Noachtar, S. & Brown, P. (2000),
`Organization of cortical activities related to movement in humans', The Journal of
Neuroscience 20(6), 2307{2314.
Martin, J. H. (2003), Neuroanatomy: text and atlas 3rd Ed., McGraw Hill.
Mathiowetz, V., Volland, G., Kashman, N. & Weber, K. (1985), `Adult norms for the
box and block test of manual dexterity', American Journal of Occupational Therapy
39(6), 386{391.
Mathiowetz, V., Weber, K., Kashman, N. & Volland, G. (1985), `Adult norms for the
nine hole peg test of nger dexterity', OTJR: Occupation, Participation and Health
5(1), 24{38.
McClelland, V. M., Cvetkovic, Z. & Mills, K. R. (2012), `Rectication of the emg is an
unnecessary and inappropriate step in the calculation of corticomuscular coherence',
Journal of neuroscience methods 205(1), 190{201.
138
McDonnell, M. N., Ridding, M. C., Flavel, S. C. & Miles, T. S. (2005), `Eect of hu-
man grip strategy on force control in precision tasks', Experimental brain research
161(3), 368{373.
McFarland, D. J., McCane, L. M., David, S. V. & Wolpaw, J. R. (1997), `Spatial lter
selection for eeg-based communication', Electroencephalography and Clinical Neuro-
physiology 103(3), 386{394.
Mendez-Balbuena, I., Huethe, F., Schulte-Mnting, J., Leonhart, R., Manjarrez, E. &
Kristeva, R. (2011), `Corticomuscular coherence re
ects interindividual dierences
in the state of the corticomuscular network during low-level static and dynamic
forces', Cerebral Cortex p. bhr147.
Mima, T. & Hallett, M. (1999), `Corticomuscular coherence: a review', Journal of Clinical
Neurophysiology 16(6), 501.
Mima, T., Simpkins, N., Oluwatimilehin, T. & Hallett, M. (1999), `Force level modulates
human cortical oscillatory activities', Neuroscience letters 275(2), 77{80.
Mosier, K., Lau, C., Wang, Y., Venkadesan, M. & Valero-Cuevas, F. J. (2011), `Con-
trolling instabilities in manipulation requires specic cortical-striatal-cerebellar net-
works', J Neurophysiol 105(3), 1295{305.
Muir, R. & Lemon, R. (1983), `Corticospinal neurons with a special role in precision grip',
Brain research 261(2), 312{316.
Murthy, V. N. & Fetz, E. E. (1992), `Coherent 25-to 35-hz oscillations in the sensorimotor
cortex of awake behaving monkeys', Proceedings of the National Academy of Sciences
89(12), 5670{5674.
Murthy, V. N. & Fetz, E. E. (1996a), `Oscillatory activity in sensorimotor cortex of awake
monkeys: synchronization of local eld potentials and relation to behavior', Journal
of Neurophysiology 76(6), 3949{3967.
Murthy, V. N. & Fetz, E. E. (1996b), `Synchronization of neurons during local eld
potential oscillations in sensorimotor cortex of awake monkeys', Journal of Neuro-
physiology 76(6), 3968{3982.
Myers, L., Lowery, M., O'malley, M., Vaughan, C., Heneghan, C., Gibson, A. S. C.,
Harley, Y. & Sreenivasan, R. (2003), `Rectication and non-linear pre-processing
of emg signals for cortico-muscular analysis', Journal of neuroscience methods
124(2), 157{165.
URL: http://www.sciencedirect.com/science/article/pii/S0165027003000049
Nashmi, R., Mendona, A. J. & MacKay, W. A. (1994), `Eeg rhythms of the sensorimotor
region during hand movements', Electroencephalography and clinical neurophysiology
91(6), 456{467.
139
Nunez, P. L., Srinivasan, R., Westdorp, A. F., Wijesinghe, R. S., Tucker, D. M., Silber-
stein, R. B. & Cadusch, P. J. (1997), `Eeg coherency: I: statistics, reference elec-
trode, volume conduction, laplacians, cortical imaging, and interpretation at multiple
scales', Electroencephalography and clinical neurophysiology 103(5), 499{515.
Nunez, P., Silberstein, R., Cadusch, P., Wijesinghe, R., Westdorp, A. & Srinivasan, R.
(1994), `A theoretical and experimental study of high resolution eeg based on surface
laplacians and cortical imaging', Electroencephalography and clinical neurophysiology
90(1), 40{57.
Oaube, J. R. (1991), `Aaem minimonograph 11: needle examination in clinical elec-
tromyography', Muscle and nerve 14(8), 685{700.
Ohara, S., Mima, T., Baba, K., Ikeda, A., Kunieda, T., Matsumoto, R., Yamamoto,
J., Matsuhashi, M., Nagamine, T. & Hirasawa, K. (2001), `Increased synchroniza-
tion of cortical oscillatory activities between human supplementary motor and pri-
mary sensorimotor areas during voluntary movements', The Journal of Neuroscience
21(23), 9377{9386.
Omlor, W., Patino, L., Hepp-Reymond, M.-C. & Kristeva, R. (2007), `Gamma-range
corticomuscular coherence during dynamic force output', Neuroimage 34(3), 1191{
1198.
Omlor, W., Patino, L., Mendez-Balbuena, I., Schulte-Mnting, J. & Kristeva, R. (2011),
`Corticospinal beta-range coherence is highly dependent on the pre-stationary motor
state', The Journal of Neuroscience 31(22), 8037{8045.
Oostenveld, R., Fries, P., Maris, E. & Schoelen, J.-M. (2011), `Fieldtrip: open source
software for advanced analysis of meg, eeg, and invasive electrophysiological data',
Computational intelligence and neuroscience 2011, 1.
Patino, L., Omlor, W., Chakarov, V., Hepp-Reymond, M.-C. & Kristeva, R. (2008), `Ab-
sence of gamma-range corticomuscular coherence during dynamic force in a deaer-
ented patient', Journal of neurophysiology 99(4), 1906{1916.
Pauluis, Q., Baker, S. N. & Olivier, E. (1999), `Emergent oscillations in a realistic network:
the role of inhibition and the eect of the spatiotemporal distribution of the input',
Journal of computational neuroscience 6(1), 27{48.
Percival, D. & Walden, A. (1993), `Spectral analysis for physical applications: Multitaper
and conventional univariate techniques (cambridge', Press, New York .
Pesaran, B. (2008), `Spectral analysis for neural signals', Short Course III p. 1.
Peters, B. O., Pfurtscheller, G. & Flyvbjerg, H. (2001), `Automatic dierentiation of mul-
tichannel eeg signals', Biomedical Engineering, IEEE Transactions on 48(1), 111{
116.
140
Pfurtscheller, G. & Lopes da Silva, F. H. (1999), `Event-related eeg/meg synchronization
and desynchronization: basic principles', Clinical neurophysiology 110(11), 1842{
1857.
Pfurtscheller, G. & Neuper, C. (1994), `Event-related synchronization of mu rhythm in
the eeg over the cortical hand area in man', Neuroscience letters 174(1), 93{96.
Plonsey, R. (1974), `The active ber in a volume conductor', Biomedical Engineering,
IEEE Transactions on 21(5), 371{381.
Popivanov, D., Mineva, A. & Krekule, I. (1999), `Eeg patterns in theta and gamma
frequency range and their probable relation to human voluntary movement organi-
zation', Neuroscience letters 267(1), 5{8.
Proulx, J., Cliord, R., Sorensen, S., Lee, D. & Archibald, J. (2006), `Development and
evaluation of a bluetooth ekg monitoring sensor', pp. 507{511.
Rajaratnam, B. (2014), `Ch goh j, kumar vp (2014) a comparison of emg signals from
surface and ne-wire electrodes during shoulder abduction', Int J Phys Med Rehabil
2(206), 2.
Rash, G. S. & Quesada, P. (2003), `Electromyography fundamentals', Retrieved February
4.
Reyes, A. & Valero-Cuevas, F. J. (2013), `Inexpensive, wearable, wireless, multi-channel,
and multi-sensor data acquisition system', pp. 216{219.
Riddle, C. N., Baker, M. R. & Baker, S. N. (2004), `The eect of carbamazepine on
human corticomuscular coherence', Neuroimage 22(1), 333{340.
Riddle, C. N. & Baker, S. N. (2005), `Manipulation of peripheral neural feedback loops
alters human corticomuscular coherence', The Journal of physiology566(2), 625{639.
Riddle, C. N. & Baker, S. N. (2006), `Digit displacement, not object compliance, under-
lies task dependent modulations in human corticomuscular coherence', Neuroimage
33(2), 618{627.
Roeleveld, K., Blok, J., Stegeman, D. & Van Oosterom, A. (1997), `Volume conduction
models for surface emg; confrontation with measurements', Journal of Electromyog-
raphy and Kinesiology 7(4), 221{232.
Roland, P., Larsen, B., Lassen, N. & Skinhoj, E. (1980), `Supplementary motor area and
other cortical areas in organization of voluntary movements in man', J Neurophysiol
43(1), 118{136.
Rosenberg, J., Amjad, A., Breeze, P., Brillinger, D. & Halliday, D. (1989), `The fourier
approach to the identication of functional coupling between neuronal spike trains',
Progress in biophysics and molecular biology 53(1), 1{31.
141
Roth, B. J., Gielen, F. L. & Wikswo, J. P. (1988), `Spatial and temporal frequency-
dependent conductivities in volume-conduction calculations for skeletal muscle',
Mathematical Biosciences 88(2), 159{189.
Rougeul-Buser, A. & Buser, P. (1997), `Rhythms in the alpha band in cats and their
behavioural correlates', International Journal of Psychophysiology26(1-3), 191{203.
Safri, N. M., Murayama, N., Hayashida, Y. & Igasaki, T. (2007), `Eects of concur-
rent visual tasks on cortico-muscular synchronization in humans', Brain research
1155, 81{92.
Safri, N. M., Murayama, N., Igasaki, T. & Hayashida, Y. (2006), `Eects of visual stim-
ulation on cortico-spinal coherence during isometric hand contraction in humans',
International journal of psychophysiology 61(2), 288{293.
Salenius, S., Portin, K., Kajola, M., Salmelin, R. & Hari, R. (1997), `Cortical control of
human motoneuron ring during isometric contraction', Journal of Neurophysiology
77(6), 3401{3405.
Sanes, J. N. & Donoghue, J. P. (1993), `Oscillations in local eld potentials of the primate
motor cortex during voluntary movement', Proceedings of the National Academy of
Sciences 90(10), 4470{4474.
Schieber, M. H. & Santello, M. (2004), `Hand function: peripheral and central constraints
on performance', Journal of Applied Physiology 96(6), 2293{2300.
Schoelen, J. M., Oostenveld, R. & Fries, P. (2005), `Neuronal coherence as a mechanism
of eective corticospinal interaction', Science 308(5718), 111{113.
Schoelen, J.-M., Poort, J., Oostenveld, R. & Fries, P. (2011), `Selective movement prepa-
ration is subserved by selective increases in corticomuscular gamma-band coherence',
The Journal of Neuroscience 31(18), 6750{6758.
Semciw, A. I., Neate, R. & Pizzari, T. (2014), `A comparison of surface and ne wire emg
recordings of gluteus medius during selected maximum isometric voluntary contrac-
tions of the hip', Journal of Electromyography and Kinesiology 24(6), 835{840.
Shinoda, Y., Yokota, J.-I. & Futami, T. (1981), `Divergent projection of individual cor-
ticospinal axons to motoneurons of multiple muscles in the monkey', Neuroscience
letters 23(1), 7{12.
Sirin, A. & Patla, A. (1987), `Myoelectric changes in the triceps surae muscles under sus-
tained contractions', European journal of applied physiology and occupational physi-
ology 56(2), 238{244.
Slepian, D. & Pollack, H. O. (1961), `Prolate spheroidal wavefunctions: Fourier analysis
and uncertainty i', Bell System Tech Journal 40, 43{63.
Slock, D. T. & Kailath, T. (1991), `Numerically stable fast transversal lters for recursive
least squares adaptive ltering', Signal Processing, IEEE Transactions on39(1), 92{
114.
142
Smith, A. M. (1981), `The coactivation of antagonist muscles', Canadian journal of phys-
iology and pharmacology 59(7), 733{747.
Stanck Jr, A. & Pfurtscheller, G. (1996), `Event-related desynchronisation of central
beta-rhythms during brisk and slow self-paced nger movements of dominant and
nondominant hand', Cognitive Brain Research 4(3), 171{183.
Stefan, H., Quesney, L., AbouKhalil, B. & Olivier, A. (1991), `Electrocorticography in
temporal lobe epilepsy surgery', Acta neurologica scandinavica 83(2), 65{72.
Steriade, M. (1997), `Synchronized activities of coupled oscillators in the cerebral cortex
and thalamus at dierent levels of vigilance', Cerebral Cortex 7(6), 583{604.
URL: http://cercor.oxfordjournals.org/content/7/6/583.full.pdf
Talati, A., Valero-Cuevas, F. J. & Hirsch, J. (2005), `Visual and tactile guidance of
dexterous manipulation tasks: An fmri study 1, 2', Perceptual and motor skills
101(1), 317{334.
Tecchio, F., Porcaro, C., Zappasodi, F., Pesenti, A., Ercolani, M. & Rossini, P. M. (2006),
`Cortical short-term fatigue eects assessed via rhythmic brainmuscle coherence',
Experimental brain research 174(1), 144{151.
Thickbroom, G., Phillips, B., Morris, I., Byrnes, M. & Mastaglia, F. (1998), `Isometric
force-related activity in sensorimotor cortex measured with functional mri', Experi-
mental brain research 121(1), 59{64.
Thomson, D. J. (1982), `Spectrum estimation and harmonic analysis', Proceedings of the
IEEE 70(9), 1055{1096.
Toma, K., Mima, T., Matsuoka, T., Gerlo, C., Ohnishi, T., Koshy, B., Andres, F. &
Hallett, M. (2002), `Movement rate eect on activation and functional coupling of
motor cortical areas', Journal of neurophysiology 88(6), 3377{3385.
Turak, B., Louvel, J., Buser, P. & Lamarche, M. (2001), `Parieto-temporal rhythms in
the 69 hz band recorded in epileptic patients with depth electrodes in a self-paced
movement protocol', Clinical neurophysiology 112(11), 2069{2074.
Uno, Y., Kawato, M. & Suzuki, R. (1989), `Formation and control of optimal trajectory
in human multijoint arm movement', Biological cybernetics 61(2), 89{101.
Valero-Cuevas, F. J., Smaby, N., Venkadesan, M., Peterson, M. & Wright, T. (2003),
`The strength-dexterity test as a measure of dynamic pinch performance', Journal
of biomechanics 36(2), 265{270.
Weinand, M., Wyler, A., Richey, E., Phillips, B. & Somes, G. (1992), `Long-term ictal
monitoring with subdural strip electrodes: prognostic factors for selecting temporal
lobectomy candidates', Journal of neurosurgery 77(1), 20{28.
Westling, G. & Johansson, R. (1984), `Factors in
uencing the force control during preci-
sion grip', Experimental Brain Research 53(2), 277{284.
143
Westling, G. & Johansson, R. S. (1987), `Responses in glabrous skin mechanoreceptors
during precision grip in humans', Experimental Brain Research 66(1), 128{140.
Winstein, C. J., Grafton, S. T. & Pohl, P. S. (1997), `Motor task diculty and brain
activity: investigation of goal-directed reciprocal aiming using positron emission
tomography', Journal of Neurophysiology 77(3), 1581{1594.
Winter, B. (2013a), `Linear models and linear mixed eects models in r: Tutorial 11',
arXiv preprint arXiv:1308.5499 .
Winter, B. (2013b), `A very basic tutorial for performing linear mixed eects analyses',
arXiv preprint arXiv:1308.5499 .
Witham, C. L., Wang, M. & Baker, S. N. (2010), `Corticomuscular coherence between
motor cortex, somatosensory areas and forearm muscles in the monkey', Front Syst
Neurosci 4.
Witte, M., Patino, L., Andrykiewicz, A., HeppReymond, M. & Kristeva, R. (2007), `Mod-
ulation of human corticomuscular betarange coherence with lowlevel static forces',
European Journal of Neuroscience 26(12), 3564{3570.
Wyler, A., Ojemann, G., Lettich, E. & Ward Jr, A. (1984), `Subdural strip electrodes for
localizing epileptogenic foci', Journal of neurosurgery 60(6), 1195{1200.
Yao, B., Salenius, S., Yue, G. H., Brown, R. W. & Liu, J. Z. (2007), `Eects of surface
emg rectication on power and coherence analyses: an eeg and meg study', Journal
of neuroscience methods 159(2), 215{223.
Zhou, P.-b. (2012), Numerical analysis of electromagnetic elds, Springer Science and
Business Media.
144

Abstract (if available)

Abstract This dissertation focuses on characterizing cortical involvement during low force dexterous manipulation. Cortical oscillations in the beta frequency range (15 - 30 Hz) are synchronous with contralateral muscular activity during static precision pinch, indicative of strong cortico-muscular coupling. However, it is poorly understood how the cortex modulates the control of fingertip forces during a time-critical dexterous task. The goal of this research was to examine the functional connectivity between cortex and muscle during a force tracking precision pinch task using a rigid wooden dowel and a compliant unstable spring at two force levels. At the low force level for both objects and at the high force level with the dowel, the difficulty in maintaining a steady compression was minimal. However, at the high force level with the unstable spring, the dexterity requirements to maintain a steady force compression were significantly more challenging and required heightened sensorimotor integration. Using this novel paradigm, we showed that increases in sensory feedback and dexterity demand disrupt consistent descending commands seen in stable grasps and are reflected as a reduction in beta corticomuscular coherence. Despite the fact that the force levels were kept constant for both objects, these findings suggest that for precision force control there exist functionally different cortical circuits that are highly dependent on the temporal demands of the task.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Neurophysiological bases of the strength‐dexterity paradigm with the use of H‐reflex

PDF

Demographic and clinical covariates of sensorimotor processing

PDF

Hierarchical tactile manipulation on a haptic manipulation platform

PDF

Facilitating myocontrol for children with cerebral palsy

PDF

Spinal-like regulator for control of multiple degree-of-freedom limbs

PDF

Neuromuscular dynamics in the context of motor redundancy

PDF

Behavioral, muscular and dynamical changes in low force dexterous manipulation during development and aging

PDF

Model-based studies of control strategies for noisy, redundant musculoskeletal systems

PDF

A percutaneously implantable wireless neurostimulator for treatment of stress urinary incontinence

PDF

insideOut: Estimating joint angles in tendon-driven robots using Artificial Neural Networks and non-collocated sensors

PDF

Investigating the role of muscle physiology and spinal circuitry in sensorimotor control

PDF

Finite element model to understand the role of fingerprints in generating vibrations

PDF

Inference of computational models of tendon networks via sparse experimentation

PDF

Design and use of a biomimetic tactile microvibration sensor with human-like sensitivity and its application in texture discrimination using Bayesian exploration

PDF

Behavioral and neurophysiological studies of hand dexterity in health and Parkinson's disease

PDF

Feasibility theory

PDF

A non-invasive and intuitive command source for upper limb prostheses

PDF

Control and dynamics of turning tasks with different rotation and translation requirements

PDF

Experimental and model-based analyses of physiological determinants of force variability

PDF

Coordination in multi-muscle systems: from control theory to cortex

Asset Metadata

Creator Reyes, Alexander (author)

Core Title Task-dependent modulation of corticomuscular coherence during dexterous manipulation

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Biomedical Engineering

Publication Date 09/16/2015

Defense Date 08/25/2015

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag coherence,electroencephalography,electromyography,OAI-PMH Harvest

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Valero-Cuevas, Francisco J. (committee chair), Kutch, Jason J. (committee member), Loeb, Gerald E. (committee member)

Creator Email areyes30@gmail.com,reyesale@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-181668

Unique identifier UC11275126

Identifier etd-ReyesAlexa-3910.pdf (filename),usctheses-c40-181668 (legacy record id)

Legacy Identifier etd-ReyesAlexa-3910.pdf

Dmrecord 181668

Document Type Dissertation

Format application/pdf (imt)

Rights Reyes, Alexander

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA