University of Southern California Dissertations and Theses

Identifying injury risk, improving performance, and facilitating learning using an integrated biomechanics informatics system (IBIS)

(USC Thesis Other)

Identifying injury risk, improving performance, and facilitating learning using an integrated biomechanics informatics system (IBIS)

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Identifying injury risk, improving performance, and facilitating
learning using an integrated biomechanics informatics system
(IBIS)
by
Harper E. Stewart
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BIOMEDICAL ENGINEERING)
August 2024
Copyright 2024 Harper E. Stewart

I dedicate this thesis to the sport of track and field for
teaching me the value of athletic and academic growth,
creating a community that feels like extended family,
and bringing forever friends into my life.
ii

Acknowledgements
The author acknowledges the partial financial support of the ARCS Foundation, Pac12 StudentAthlete Health & Well-Being Grant 3-03 Pac-12- Oregon - Hahn - 17 - 02, USA Track and Field,
and USC WiSE.
I would like to thank my advisor Dr. Jill L. McNitt-Gray. It’s such a privilege to have this
opportunity and I am honored to be one of your students. Your commitment and dedication to
athletes inspires me to learn more every day.
I would like to thank my committee members Dr. James Finley, Dr. Brent Liu, Dr. Rand
Wilcox, and Dr. Michael Khoo. Dr. Wilcox patiently answered many of my emails and the robust
statistics throughout my dissertation are because of his advice and expertise, thank you so much
for all of the time you put into these papers. Special thanks also to Dr. Henryk Flashner who
served as a member of my qualifying exam committee and constantly pushes me to be my fastest.
I appreciate my committee’s advice and support throughout this process.
Thank you to my lab mate, Casey Wiens, for spending the first three years helping me through
this process and bringing me into the best friend group I could have asked for. Thank you to all the
members of the USC Biomechanics Research Lab including Justin Gaither, Abbey Stepnitz, Peter
Cordi, Marisa Papp, Juwon Lee, Myles Shelton, Nathan Phillips, Michael Pozzi, Rachel.Amir
Chatman, Douglas Alford, Lauren Teubner, Ashley Soto, Adrien Canery, Amy Zhang. Thank you
to Kyle Davis, Amelia Goudzwaard, Grace Piper, and Rodolfo Amezcua-Cerda for the many hours
put in helping me do the heavy lifting to set up the data collections for sprint starts.
I would like to thank my collaborators including Anshu Goyal and Joseph Liu in the IPI Lab
and Bowen Song in the USC QED Lab. I would like to thank our collaborators at the University
iii

of Oregon, Dr. Michael Hahn and Dr. Kathryn Farina, and our collaborators at the University of
Colorado Boulder, Dr. Alena Grabowski, Dr. Rodger Kram, and Dr. Ryan Alcantara.
I appreciate all of the athletes who participated in the studies so that we could do this work.
It’s a privilege to get to work with the USA Track and Field jumpers and the USC Track and Field
athletes and coaches. Thank you to Robert Chapman and Tyler Noble giving us the opportunity to
support our Team USA athletes.
I would like to thank all of my friends for their support helping me through the PhD, especially
my friends in BME who I started this journey with. Thank you to my friend Natalie Khalil for
pulling me through hardest few months finishing up this process. Thank you to my VGSA community who brought me so many friends that I will cherish for the rest of my life. Thank you to all
of my friends in the biomechanics community from ASB, ISBS, ISB and beyond. Thank you for
all of your kind words and support of #biomechtalk.
Finally, I would like to thank my Mom, Dad, Morgan, and Mitchell for being the best family
I could ask for. They keep me humble by calling me a nerd frequently. When I am taking myself
too seriously I think about how, to my Dad, running force-time curves just look like wigs.
iv

Table of Contents
Dedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
Chapter 1: Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Motivations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Chapter 2: IBIS and Database Infrastructure . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2.1 Database, R, & IBIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.2 Foot Contact Detection Algorithm . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3.1 Database, R, & IBIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3.2 Foot Contact Detection Algorithm . . . . . . . . . . . . . . . . . . . . . . 11
2.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4.1 Database, R, & IBIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4.2 Foot Contact Detection Algorithm . . . . . . . . . . . . . . . . . . . . . . 16
Chapter 3: Can ground reaction force variables pre-identify the probability of a musculoskeletal injury in collegiate distance running? . . . . . . . . . . . . . . . . . . . . . . 20
3.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.3.1 Participants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.3.2 Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.3.3 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.3.4 Ground Reaction Force Variables . . . . . . . . . . . . . . . . . . . . . . 24
3.3.5 With and Without Impact Peak Subgroups . . . . . . . . . . . . . . . . . . 26
3.3.6 Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
v

3.4.1 GRF Variable Boundaries for Non-Injured Collegiate Runners . . . . . . . 28
3.4.2 GRF Variables Outside Normative Boundaries for Injured and Non-Injured
Runners . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Chapter 4: Comparison of impulse regulation and lower extremity control between full
jump and popoff takeoffs in the long jump . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.3.1 Participants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.3.2 Tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.3.3 Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.3.4 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.3.5 Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.4.1 Whole-Body: Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.4.2 Whole-Body Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.4.3 Multi-Joint Control of the Leg . . . . . . . . . . . . . . . . . . . . . . . . 46
4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Chapter 5: Multi-joint control strategies sprinters use to regulate impulse generation during
the first foot contact out of the blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.3.1 Participants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.3.2 Experimental Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.3.3 Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.3.4 Data Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.3.5 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.3.6 Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.4.1 Whole Body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.4.2 Joint Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Chapter 6: Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
vi

List of Tables
3.1 Median and scaled median absolute deviation (MADN) for 10 ground reaction
force (GRF) variables of Non-Injured runners. Note. The group boundary values
are presented for Non-Injured runners (left), Non-Injured runners Without transient
vertical GRF impact peaks (middle), and Non-Injured runners With Impact Peaks
(right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2 Statistical comparison using Cliff’s method of the number of ground reaction force
(GRF) variables outside of Non-Injured normative boundaries between both legs
of Non-Injured runners versus the injured legs of Injured runners. . . . . . . . . . . 30
4.1 Median and normalized median absolute deviation (MADN) are reported for the
net joint moment (NJM) impulse magnitude and percent distributions to the overall
moment impulse for 14 elite to world-class long jumpers without prosthetics during impact, post-impact, and foot contact. There were no significant differences in
NJM impulse magnitude or percent distribution between full jumps and popoffs at
the ankle, knee, or hip. During impact, long jumpers controlled the leg primarily
using the hip, while in post-impact, the multi-joint control was dominated by the
knee and ankle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.1 Median (MADN) change in the horizontal center of mass velocity for 15 highlytrained to world-class sprinters in the first foot contact out of blocks. . . . . . . . . 61
5.2 Net joint moment (NJM) impulses for the ankle, knee, and hip, during impact, and
post-impact, and foot contact for 15 highly-trained to world-class sprinters in the
first foot contact out of blocks. Median (Med) and normalized median absolute
deviation (MADN) for the magnitude and percent of overall lower-extremity NJM
impulse are reported. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
vii

List of Figures
1.1 IBIS System and Database Infrastructure facilitates the ability to answer biomechanics research questions that require large sample size, longitudinal analysis,
and immediate feedback. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1 Biomechanics research workflow using the Integrated Biomechanics Informatics
System (IBIS). Gray box indicates the overall IBIS system which consists of a
database and web server. The blue arrows indicate how data can be uploaded from
data collection (orange box) or pre-processed (purple box) and then integrated with
the system. Once the biomechanics data is within the system, it can be accessed
by the users. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Example of video data integrated with Ground Reaction Force (RF) data from force
plates to communicate to stakeholders how RF applied at the foot (green RF vector,
anterior-posterior and vertical components) acts on the body and how components
of this RF applied over time contribute to momentum regulation during ground
contact. Synchronization of the RFs with the video requires determination of initial
contact of the foot with the force plate. . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3 Series of video images prior to and during ground contact during steady-state running. The middle image reflects the instant of initial ground contact between the
foot and the ground. The purpose of the algorithm is to help automate and reliably
detect a video frame containing initial foot contact from a series of video frames. . 10
2.4 The foot contact detection algorithm requires the user to manually select the region
of interest (purple box) where the foot is expected to initiate contact with the force
plates. Changes in the pixels within this region of interest, identified here in the
first frame of the video, are then used to detect when and where the foot contact
with the plate occurs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
viii

2.5 The blue trace shows the comparison difference values for each from of a video.
These pixel changes are compared with the first frame of the video. The purple X
indicates the peak maximum found by the select peaks function. Finding this peak
maximum narrows the selection region down from thousands to tens of frames,
saving time for users. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.6 Example of the utility of the R Shiny analysis tool which was used to create a
graphical user interface (GUI) that shows the progression of jump distance over
time for nine elite long jumpers. The gray vertical lines designate the Olympic cycles and enable the user to see a clear progression during each cycle for individual
athletes. Other features in this interactive example are enable the user to filter by
date, wind speed, and athlete’s personal best (max) jump. . . . . . . . . . . . . . . 15
2.7 This example graphical user interface (GUI) shows a dashboard of the web-based
R Shiny application for statistical analysis. This application allows research to
choose which variables are displayed on each axis and to assess the presence of
covariates using a color gradient. Specific athletes and variables can be selected
using the built-in filters and color identifiers. The dataset, once selected, will be
queried and displayed on the GUI plot. . . . . . . . . . . . . . . . . . . . . . . . . 17
2.8 Example demonstration of the process using the R Shiny analysis tool to link a
trial to more information about that performance. (1) User is shown a simple plot.
(2) User can select specific trails and additional relative data populates the table
below. Here the user is also able to select to view the video that corresponds to a
specific data point. (3) The video appears in a new window ready to view. . . . . . 17
2.9 Exemplar video image used to assist the user in visual verification of initial foot
contact determination use the Graphical User Interface (GUI). . . . . . . . . . . . 18
3.1 Example depictions and definitions of the 10-ground reaction force (GRF) variables used to characterize timing, force, and change in velocity (∆v). GRF curves
(black traces) were used to calculate the given variable. The dashed gray vertical
line separates the braking and propulsion phases. The green horizontal lines show
the length of braking and contact times. The blue horizontal lines show the average GRFs in the vertical and anterior-posterior directions during each phase. The
blue area under the GRF curve represents the impulse in the vertical and anteriorposterior directions during each phase. The gray areas represent the vertical impulse due to body weight (BW), which is subtracted to calculate the change in
vertical velocity in each phase. Abbreviations: Avg, average; Prop, propulsion;
Brake, braking; v, vertical; ap, anterior-posterior. . . . . . . . . . . . . . . . . . . 25
ix

3.2 Non-Injured boundaries for contact time shown on exemplar ground reaction force
(GRF) curves. The solid vertical line is the median contact time, and the dashed
vertical lines are the Non-Injured boundaries, which are ± 1 scaled median absolute
deviation (MADN). The left graph represents the Non-Injured runner boundaries
with two sample GRF curves, the middle graph is an exemplar GRF curve with no
transient vertical GRF impact peak, and the right graph is an exemplar GRF with
an impact peak. Abbreviations: BW, body weight. . . . . . . . . . . . . . . . . . . 27
3.3 Top Row: Exemplar Non-Injured runner with 1 ground reaction force (GRF) variable outside the Non-Injured boundaries (blue). Bottom Rows: Nine Injured runners with varying GRF variables outside the Non-Injured boundaries. The green
region indicates the Non-Injured boundaries defined as the median ± scaled median
absolute deviation (MADN) of all 24 Non-Injured runners. The circles represent
the robust distance measurement (RDM) from the median of each runner for each
GRF variable. RDM of zero indicates that the runner’s individual GRF variable
matched the median value from the Non-Injured runners. Positive RDM indicates
greater forces, longer contact and braking times, and greater changes in velocity
than the median Non-Injured value. Negative RDM values indicate smaller forces,
shorter braking and contact times, and smaller changes in velocity than the median Non-Injured value. Abbreviations: Avg, average; Prop, propulsion; Brake,
braking; ∆v, change in velocity; v, vertical; ap, anterior-posterior. . . . . . . . . . . 34
3.4 Eight runners with different lower extremity injuries with varying GRF variables
outside the Non-Injured boundaries. The orange region indicates the Non-Injured
boundaries defined as the median ± scaled median absolute deviation (MADN) of
the 12 Non-Injured runners Without impact peaks. The purple region indicates the
Non-Injured boundaries defined by the 7 runners With impact peaks. The circles
represent the robust distance measurement (RDM) from the median of each runner
for each GRF variable. RDM of zero indicates that the runner’s individual GRF
variable matched the median value from the Non-Injured runners. Positive RDM
indicates greater forces, longer contact and braking times, and greater changes
in velocity than the median Non-Injured value. Negative RDM values indicate
smaller forces, shorter braking and contact times, and smaller changes in velocity
than the median Non-Injured value. Abbreviations: Avg, average; Prop, propulsion; Brake, braking; ∆v, change in velocity; v, vertical; ap, anterior-posterior. . . . 35
x

3.5 The number of Non-Injured runners’ legs (open bars) and Injured (Inj) runners’
injured legs (grey bars) with ground reaction force (GRF) variables outside the
normative boundaries from Non-Injured (Non-Inj) runners. Solid vertical lines indicate the median (Med) for each group, and dashed vertical lines indicate each
group’s scaled median absolute deviation (MADN). Top: Green lines indicate the
distribution of the 48 Non-Injured runners’ legs, and gray lines indicate the distribution of the 9 Injured runners’ injured legs. Middle: Orange lines indicate the
distribution of the 24 Non-Injured runners’ legs, and gray lines indicate the distribution of the 5 Injured runners’ injured legs Without transient vertical GRF impact
peaks. Bottom: Purple lines indicate the distribution of the 14 Non-Injured runners’ legs, and gray lines indicate the distribution of the 3 Injured runners’ injured
legs With transient vertical GRF impact peaks. . . . . . . . . . . . . . . . . . . . . 36
4.1 Ground reaction force (GRF) time-curves and filmstrips for an exemplar jumper’s
full jump and popoff. Bold gray vertical line divide impact and post-impact. Green
arrows on the filmstrips show the GRF vector at that particular event. Abbreviations: Fv: Vertical GRF; Fh: Horizontal GRF. . . . . . . . . . . . . . . . . . . . . 41
4.2 In full jumps and popoffs, 15 elite to world-class long jumpers generated similar changes in center of mass (CM) horizontal and vertical velocity during takeoff
(Top). There was a statistically significant correlation between the horizontal velocity lost and the vertical velocity gained during takeoff for both full jumps and
popoffs (Top). The same data was plotted with a color gradient to show incoming horizontal velocity, demonstrating that similar changes in CM vertical velocity
can be achieved with lower incoming horizontal velocities (Bottom). Despite coming in with lower incoming horizontal velocities, there were female long jumpers
who generated comparable changes in CM velocity during the takeoff to their male
counterparts in both full jumps and popoffs (Bottom). The two data points in the
top right corner are from a unilateral trans-tibial amputee jumper with a prosthesis
on their takeoff leg. This jumper was not included in the correlations or regression
equations for the 14 jumpers without prosthetics. The unilateral amputee jumper
was able to lose less horizontal CM velocity while generated similar changes in
vertical CM velocity to the jumpers without prosthetics (Bottom). The solid gray
line corresponds to the full jump regression, and the black dotted line corresponds
to the popoff regression. Note. *: Statistically significant. . . . . . . . . . . . . . . 43
xi

4.3 Arrows connect each long jumper’s full jump to their popoff attempt. Red arrows
indicate that jumpers generated more vertical CM velocity and lost more horizontal
CM velocity on their full jump takeoff than their popoff (8 jumpers), while blue
arrows indicated that jumpers generated less vertical CM velocity and lost less
horizontal CM velocity on their full jump than their popoff (5 jumpers, Left). Gray
arrows indicate that a different pattern was found. Numbers above the full jump
indicate each jumper’s participant number. Jumper 7 is a unilateral trans-tibial
amputee with a prosthetic on their takeoff leg. Starting with a larger leg angle at
initial contact seemed to contribute to the greater loss in horizontal CM velocity
and greater gain in vertical CM velocity during takeoff for 12 of 15 jumpers during
full jumps and popoffs, as shown by the matching arrow colors on each plot (Left,
Right). If a jumper had a larger leg angle at initial contact, they typically generated
a greater loss in horizontal CM velocity and greater gain in vertical CM velocity
during takeoff for both full jumps and popoffs (Right). Red arrows indicate that
7 jumpers lost more horizontal CM velocity and started with a larger leg angle at
initial contact, while 4 jumpers did the opposite (Right). . . . . . . . . . . . . . . 45
4.4 Net joint moment (NJM) impulses for full jumps and popoffs organized by individual for 14 elite to world-class long jumpers without prosthetics at the ankle, knee,
and hip during foot contact (Top), impact (Bottom Left), and post-impact (Bottom
Right). Vertical gray lines separate the data for each jumper with their full jump
attempt (left) and then their popoff attempt (right). . . . . . . . . . . . . . . . . . 46
4.5 Net joint moment (NJM) impulses for full jumps (Left) and popoffs (Right) for
a group of 14 elite to world-class long jumpers without prosthetics at the ankle,
knee, and hip during impact (Top), post-impact (Middle), and foot contact (Bottom). Data is ordered left to right by smallest to largest leg angle at initial contact.
No significant difference was found in the NJM impulse magnitudes between full
jumps and popoffs at the ankle, knee, or hip during impact, post-impact, and foot
contact during long jump takeoffs. . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.6 Net joint moment (NJM) time curves for an exemplar long jumper’s popoff attempt
during takeoff. (Top) Filmstrip of the joint kinetics at key instances in time with
associated free body diagrams (FBDs) and NJMs at the ankle, knee, and hip. The
blue lines on the FBDs are the net joint forces, and the circles correspond to the
NJMs. The circle diameter indicates the magnitude of the NJM, while the color
indicates direction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
xii

5.1 Exemplar feedback video where resultant ground reaction force (GRF) vectors,
force-time curves, and kinematics were synchronized for a sprinter’s first foot contact out of the blocks. This feedback was provided to coaches and athletes within 3
minutes of sprint start completion. (Left) Shows an early instant in the video where
the GRF vector is directed posteriorly early in foot contact. (Right) Demonstrates
a later instant in the video of the foot contact when the GRF vector is directed anteriorly. The blue vertical line indicates the instance in time on the force-time curve
synchronized with the athlete’s corresponding kinematic position. The following
measures were provided to the coaches and athletes; change in horizontal velocity
(∆vh), change in vertical velocity (∆vv), and contact time (s). . . . . . . . . . . . 58
5.2 Foot position relative to the center of mass (CM) at initial contact (IC) was significantly correlated to both contact time (Left) and average horizontal force (Right)
for the first foot contact out of the blocks for sprinters. Larger positive values
indicate that the foot was positioned further behind the CM. Blue lines indicate
the group median for each variable for 15 highly-trained to world-class sprinters.
The purple highlighted dot represents Athlete A’s exemplar trial, where the foot
was positioned further behind the center of mass (CM). The yellow highlighted
dot represents Athlete B’s exemplar trial, where their foot is closer to the CM. The
black line indicates the linear regression. Note: *; Statistically significant. . . . . . 62
5.3 Exemplar athlete trials, which illustrate how foot position relative to the center of
mass (CM) at initial contact (IC) influenced contact time and average horizontal
force. Positioning the foot further behind the CM at initial contact led to faster
contact times and a greater average horizontal force for Athlete A. Athlete kinematics at IC are on top and associated exemplar force-time curves are shown on
the bottom. Black dots along the red skeleton indicate the CM of the segment, and
the large yellow dot indicates the total body center of mass. The gray vertical lines
divide braking and propulsion. Abbrev: ∆vh; Change in horizontal velocity, ∆vv;
Change in vertical velocity, BW; body weight. . . . . . . . . . . . . . . . . . . . . 63
xiii

5.4 Average horizontal force generated during the first foot contact out of blocks was
improved when athletes reduced shank yield in impact (Top Left) and reached their
peak thigh angular velocity faster (Top Right). Greater average horizontal forces
in the first foot contact out of blocks in a sprint start were significantly correlated
to smaller magnitudes of peak shank angular velocities (less shank shield) during
impact (Top Left, 0.52*) and faster times to peak thigh angular velocity in postimpact (Top Right, -0.75*). Shank and thigh coordination generally improved
when the foot was positioned further behind the CM (Bottom). This is shown by
the blue dots in the bottom right corner of the plot with smaller peak shank angular
velocity magnitudes in impact and faster times to peak thigh angular velocity in
post-impact. Blue lines indicate the group median for each variable for 15 highlytrained to world-class sprinters. The purple highlighted dot represents Athlete A’s
exemplar trial, where the foot was positioned further behind the center of mass
(CM). The yellow highlighted dot represents Athlete B’s exemplar trial, where
their foot is closer to the CM. Note: *; Statistically significant. . . . . . . . . . . . 64
5.5 Foot position further behind the center of mass (CM) at initial contact (IC) was
significantly correlated to greater net joint moment (NJM) impulses at the knee and
hip during impact for the first foot contact out of blocks in a sprint start (Bottom
Left). There was no statistically significant correlation between the foot position
relative to the CM at IC for the ankle during impact or for any lower-extremity
joint during post-impact. Net joint moment impulses are shown for the ankle,
knee, and hip of each sprint trial during foot contact (bottom right), impact (left top
and bottom), and post-impact (top right). Trials are organized so a foot positioned
further behind the center of mass (CM) at initial contact is on the right. The bottom
left is a zoomed-in version of the NJM impulse at each lower-extremity joint during
impact. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.6 Foot position relative to the center of mass (CM) at initial contact (IC) influenced
the multi-joint control of the lower extremity for exemplar Athlete C during the first
foot contact of a sprint start. When the foot was further behind the CM, the shank
yield was smaller during impact, the range of thigh excursion was smaller, and the
ankle moment was larger throughout foot contact. The images show the free body
diagrams for the following events for each trial: initial contact, horizontal force
zero-crossing, peak vertical force, peak horizontal force, and final contact. Blue
lines show the direction and magnitude of the net joint forces. The diameter of the
circles indicates the magnitude of the net joint moment, and the color of the circles
indicates the direction at that instant in time. . . . . . . . . . . . . . . . . . . . . . 68
xiv

Abstract
In the field of biomechanics, we rely on principles of physics and biology in order to study the
control and dynamics of human movement. For years, there have been studies done in labs across
the world using techniques for collecting kinetic and kinematic data. The disparate methods for
employing the fundamental equations of motion have limited standardization for managing data
in biomechanics. These challenges are similar to those faced in the medical imaging community
standardizing CT scans or MRIs. The development of the Integrated Biomechanics Informatics
System (IBIS) aims to standardize data management practices in biomechanics to provide timely
and accurate feedback to coaches, athletes, clinicians, and patients. Through the continued work
on the IBIS system, we are making headway on the ability to understand the control and dynamics
of human movement in realistic contexts.
One application of the IBIS system is to use bioinformatics to identify injury risk, improve
performance, and facilitate learning for athletes that participate regularly in track and field. The
organized data sets in the IBIS system were leveraged to fill essential knowledge gaps for coaches
and athletes in steady-state running, long and triple jump, and sprinting. For elite endurance runners in the Pac-12 conference, we characterized force-time curves at training pace to determine if
attributes were different between injured and non-injured runners. In the long jump, we characterized whole-body and multi-joint control during the takeoff of full jumps and popoffs. This was
used to determine to what degree popoffs are representative of the multi-joint control needed in a
full jump. Finally, for sprinting we integrated force and video data using force-time curves and
force vector overlay representations. These were used to facilitate learning and inform coaching
decisions and feedback during an interactive training session on the track. Adopting best practices
in imaging informatics enabled us to leverage the IBIS system to study these research questions.
xv

Chapter 1
Introduction
1.1 Motivations
In biomechanics, its important to disseminate knowledge to key stakeholders including other
researchers, coaches, athletes, and clinicians. This includes gathering insight by integrating and
assimilating data from force plates, high speed video, inertial measurement units (IMU), electromyography (EMG), and other sources. However, communicating effectively requires organizing, synthesizing, and analyzing the information to fill gaps in our existing working knowledge.
These challenges for biomechanists in doing research limit the ability to answer important research
questions.
Biomechanics research labs have different processes for storing and organizing data. This may
include storing data on many hard drives or in different places. Without successful coordination of
FAIR (Findable, Accessible, Interoperable, Reusable)[1] data management practices across institutions, it is difficult to work across universities, achieve the sample size for improving statistical
power, analyzing participants over time, and providing immediate feedback for performance or
clinical improvement.
However, by combining best practices from other disciplines like imaging informatics it may
be possible to remedy these problems. For example in the field of medical imaging, despite differences that may arise with different machines for CT and MRI images, steps have been taken
to standardize how images are stored and communicated to benefit patients. Similar efforts have
1

been taken to develop an open data commons for spinal cord injury that is well accepted by the
research community.[2] These practices can be adopted in biomechanics. In the healthcare field,
electronic patient records record and store patient information, in biomechanics we can develop
similar processes to store athlete information by applying FAIR data management practices .[1]
With the development of the Integrated Biomechanics Informatics System (IBIS), the ability
to identify risk, improve performance, and facilitate learning for stakeholders is improving. By
organizing and storing the data using FAIR principles, it is possible to begin answering more questions that require larger sample size, longitudinal data, or integrated forms of data.
In this body of work, the IBIS system facilitated the development of studies to answer the
following questions (Figure 1.1):
• Do elite runners without injury experience different ground reaction forces during steadystate running than non-injured runners?
• Is the whole-body and multi-joint control of the takeoff leg the same when practicing full
jumps versus popoff in the long jump?
• How could the integration of video and ground reaction force data be used in a timely and
insightful way by coaches to help sprinters improve their first foot contact out of blocks
during a track training session?
The purpose of the first cross-institutional study was to characterize force-time curves of elite
endurance runners in the Pac-12 to determine if characteristics differ between injured and noninjured runners (Chapter 3). Ground reaction forces were collected during repetitive foot contacts
at training pace in steady-state running. The purpose of the second study was to determine how well
popoffs replicate the whole-body and multi-joint control of a full jump for long takeoffs (Chapter
4). Coaches commonly use popoffs as a training drill in practice where athletes complete the early
phases of the jump without the landing. The analysis provided insight into practice planning during a training progression. This was done using high speed video and ground reaction force data
to generate force-time curves and joint kinetics of the takeoff leg for analysis. Finally, the third
2

Figure 1.1: IBIS System and Database Infrastructure facilitates the ability to answer biomechanics
research questions that require large sample size, longitudinal analysis, and immediate feedback.
study facilitated feedback about the first foot contact out of blocks in a sprint start with coaches
integrating force and high speed video (Chapter 5). Synchronized force-time curves and force vector overlays were generated for coaches to provide athletes with feedback between attempts. The
proposed series of studies leveraged the infrastructure of the IBIS system to facilitate knowledge
discovery.
3

Chapter 2
IBIS and Database Infrastructure
2.1 Abstract
The field of biomechanics integrates a variety of data types to provide meaningful feedback to
athletes, coaches, and healthcare professionals to reduce injury and improve performance. These
different sources of data include waveform, video, discrete, and performance metrics. An integrated biomechanics informatics system (IBIS)[3, 4] has been developed to apply imaging informatics principles to the field of biomechanics. The IBIS system helps create a space for stakeholders to make informed decisions using multimodal biomechanics data. This current work details the
development of two new decision support tools within the IBIS system that enhance the ability of
users to effectively analyze and interpret data. These two tools include an R Shiny statistical analysis tool and an algorithm for detecting foot contact with the ground. Organizing data into SQL
databases on the IBIS system provides the backend for the R Shiny analysis tools. This connection
provides statistical analysis tools to review data over time and across individuals. Its integration
also incorporates interactive data visualization techniques to put biomechanical data in context. A
semi-automated foot contact detection algorithm was created to reduce the need for researchers
to manually search through thousands of video frames to synchronize video data accurately with
other information such as force data. Overall, these two improvements to the system will support
researchers working to provide meaningful insight about human performance and injury prevention. These additions to the IBIS system provide quicker access to data, develop standardized
4

pipelines, and create additional time for biomechanists to focus on interpreting results rather than
data management.
2.2 Introduction
As more technology is introduced into clinical and athletic settings, the pathway for data integration in biomechanics needs to be robust enough to manage the increasing quantity and diversity
of data so that it is useful to decision-makers.[5] Decision-makers reside at varied levels within
an organization and their role influences how data will be used to inform decisions over time (i.e.,
technique coach, strength and conditioning coach, medical personnel). Because of the unique technical background of each user, it is critical to create data integration pathways that are easy to use
and facilitate the dissemination of knowledge at all levels of the organization. This is similar to the
clinical environment where decision-makers need to be able to analyze patient data and access it
efficiently while maintaining privacy and security. In biomechanics, the same needs must be met
for accessing an athlete’s data to make training and injury prevention decisions.[6] Assimilation
of information and report generation as end products also need to consider intended audience and
preferred approaches for communication.[7] Analysis of the data is significantly enhanced by integrating multiple data sources, as this adds context and data layers to inform decision-making.
Specific to our current project, sport scientists are interested in integration of multiple sources
of information for multiple athletes to track improvements in performance over time.[7] Before
data can be used to inform decisions, biomechanics data collected as part of sports performance
research requires pre-processing and post-processing steps. Integration of these data from video
recordings, image sequences and signals undergo a comprehensive pre-processing workflow that
includes cleaning and data preparation for analysis whereas post-processing involves the data visualization and report generation (Figure 2.1).
The IBIS system is a novel web-based platform that integrates data from various standalone
sources and allows users to upload, share and download data with any collaborating institution,
5

Figure 2.1: Biomechanics research workflow using the Integrated Biomechanics Informatics System (IBIS). Gray box indicates the overall IBIS system which consists of a database and web
server. The blue arrows indicate how data can be uploaded from data collection (orange box) or
pre-processed (purple box) and then integrated with the system. Once the biomechanics data is
within the system, it can be accessed by the users.
6

while also providing an interface to quickly locate and simultaneously display various forms of
multimedia data for review and analysis.[4] The IBIS (Integrated Biomechanics Informatics System) was developed using medical imaging informatics guiding principles to act as a secure and
centralized location for storing, sharing, and analyzing biomechanics data.[3] The goal of this work
is to further develop the decision support and analysis tools within the system to continue streamlining the integrated biomechanics workflow and provide novel tools for knowledge discovery. In
this paper, we present an R Shiny application for statistical analysis and knowledge discovery as
well as a foot contact detection algorithm, which underscore the value of having an integrated
system for improving performance and identifying risk of injury.
2.2.1 Database, R, & IBIS
Adding databases with multimodal data to the IBIS system improves the biomechanics workflow. Databases designed using FAIR Principles (Findable, Accessible, Interoperable, Reusable)
are the foundation for other tools in sport science,[1] and they support the goal of making better
decisions. These decisions may include evaluating the success of an intervention and assessing
if individuals have progressed over time. While one can answer these questions without using a
database, it becomes progressively difficult as the data’s quantity and diversity increase. Due to
database storage design, data are in a format that can be easily shared (e.g., collaboration across
organizations), accessed across multiple software languages and programs (e.g., R, Python, MatLab), and queried by key information (e.g., gather all data between 2016 and 2020). A key to this
is designing the data’s storage and organization on two fundamental principles: non-redundancy
and data uniqueness.[8] The benefits to storing the data once include; minimizing the size of the
database and ensuring that updates (e.g., athlete’s age, height, weight, coach) are reflected throughout the database. Data from different disciplines such as biomechanical, physiological, and medical
can be queried together to ask specific questions such as “what is the difference in an athlete’s median knee net joint moment during running before and after experiencing a tibial stress fracture?”
The IBIS system continues to improve upon the existing biomechanics workflow by including
7

software to help visualize data in a meaningful way within the web-based platform and removes
many barriers to use. Open-source software (e.g., R and Python) are great tools for data visualization and statistical analyses. They also work seamlessly with databases, which contain organized
data. However, these software programs require a system that already has these programs installed.
It also requires that the user must have knowledge of the software and language. This presents an
obvious barrier for clinicians, coaches, athletes, or other researchers who lack the technical experience or time to make learning these novel skills a priority. Creating a user-interface that leverages
the software’s data visualization and statistical analyses prowess can lower those barriers. This can
be accomplished through the web-based R Shiny application. Shiny is an interactive application
that can be incorporated in generating customized reports and dashboards. Once the application
has been developed, the user is not required to have knowledge of R programming, yet they must
be able to run R locally on their computer or have the application hosted elsewhere. The installation requirement is removed when R is integrated with a secure web-based platform (i.e., IBIS).
Including databases and open-source software on the IBIS system speeds up previously difficult processes in the biomechanics workflow and allows researchers to communicate more effectively with coaches, athletes, and clinicians. Hosting a user interface application on a web-based
platform provides researchers and collaborating groups the freedom to review and analyze data
remotely from any location with internet access. All users are then able to make use of the customizable system, which informs important decisions for injury mitigation and performance. The
advanced interactive application that is hosted on the same web-based platform further streamlines
the biomechanics research process as data upload, sharing, review, and analysis can be performed
in one secure and centralized location.
2.2.2 Foot Contact Detection Algorithm
Another advancement to the IBIS system includes the development of a foot contact detection
algorithm. Many valuable biomechanical measures require the accurate detection of foot contact such as contact time, flight time, center of mass velocity, as well as the need to synchronize
8

video data with other biomechanical measures from force plates or sensors. These measures enable researchers to support coaches and athletes with informed decision making when it comes
to modifications in training or technique. This advancement of a foot contact detection algorithm
will decrease the lengthy process that is required to synchronize video and other biomechanical
data using image and signal processing methods in addition to calculating discrete variables identified previously. We will demonstrate the value of this tool using an example where force and
video data is synchronized to provide coaches, athletes, and other stakeholders with a connection
between their performance and the forces being measured (Figure 2.2).
Foot contact identification requires extensive time spent by trained researchers to identify the
initial contact with the force plate (Figure 2.3). An algorithm has been developed to reduce the
thousands of frames researchers must go through to get to the instant where the foot first contacts
the force plate or ground.
2.3 Methods
2.3.1 Database, R, & IBIS
By using fundamental principles from imaging informatics, a usable database was set up on the
IBIS system, which could be quickly accessed. A relational database model[9] and Structured
Query Language (SQL) was used to access the data. While proper organizational strategies for
relational databases are beyond this paper’s scope, one of the fundamental concepts used in each
table within the database is the relationship between tables. Primary keys - column(s) that contain
unique identifiers for each row - and foreign keys - column(s) that include the identifiers of the
related table’s primary keys - allow for multiple table connections and enforce the two fundamental principles (uniqueness and non-redundancy). The data model of the IBIS system along with
the development of the database model stems from the approach adopted in Medical Imaging Informatics and ePR (Electronic Patient Record) models where the data is organized at the highest
level by patient, and then studies for that patient, and the data from those studies1. By drawing
9

Figure 2.2: Example of video data integrated with Ground Reaction Force (RF) data from force
plates to communicate to stakeholders how RF applied at the foot (green RF vector, anteriorposterior and vertical components) acts on the body and how components of this RF applied over
time contribute to momentum regulation during ground contact. Synchronization of the RFs with
the video requires determination of initial contact of the foot with the force plate.
Figure 2.3: Series of video images prior to and during ground contact during steady-state running.
The middle image reflects the instant of initial ground contact between the foot and the ground.
The purpose of the algorithm is to help automate and reliably detect a video frame containing
initial foot contact from a series of video frames.
10

inspiration from this model, we have organized the biomechanics research data in the same way,
with the athlete at the highest level.
Once the database was installed on the server, R was added to facilitate a user interface for
data exploration. In order to integrate R and its Shiny application capabilities into the IBIS system,
the necessary packages were installed on the server hosted in the USC Data Center. The IBIS
web-based user interface is written in HTML/PHP and hosted by the Apache web server. First, the
R Shiny packages were installed on the Linux operating system. Next, the R packages were configured and tested to make sure a local SQL database in the same folder as the R Shiny application
could be read by the application and the data displayed correctly. Finally, the last step was to install
the driver that allowed the R Shiny application to connect with the pre-existing MySQL database
directly on the server used by the IBIS system. The R Shiny application was hosted by the web
server as a web-based application and the user interface was customized in R. The application was
developed to display toggles and filters to select customized datasets for display on a plot. The
data displayed on the plot is directly queried from the MySQL database tables. The web-based
application was tested to ensure that it could query and display data over time, individual athletes,
and multiple athletes simultaneously.
Further development of the R integration with the IBIS server allowed users to query data
directly on the graph and then display links to associated video/image files for the queried data.
The functionality for the users to drag and highlight any set of data points and display that set of
data was developed on the system. The data sets would also include the links to any associated
multimedia files. Links were established so that the multimedia files can be opened in a new window for quick easy-access viewing. Videos and images would be opened with video and image
viewers, respectively.
2.3.2 Foot Contact Detection Algorithm
The goal of the foot contact detection algorithm is to reduce time spent synchronizing video
data with other biomechanical data so that data can be used within an organized database stored
11

on the server. Prior to the development of this algorithm, foot contact was determined as the first
frame or instance where the foot comes in contact with the ground by a biomechanics researcher.
Initial contact is found by advancing past the point where there is space between the foot and the
ground but before deformation of the shoe begins to occur. The algorithm as developed locally in
Python and will be incorporated in pre-processing workflows for uploading video and force data
to the IBIS system to reduce the workload required of researchers. From an image processing
perspective, as a foot contacts the ground or a force plate during running, there will be a large
change in pixels as the foot enters the region of view around the force plate (Figure 2.3). The
algorithm functions by reading the video frames into the program and the region of interest where
the foot will contact the ground or the force plate is selected (Figure 2.4).
Figure 2.4: The foot contact detection algorithm requires the user to manually select the region of
interest (purple box) where the foot is expected to initiate contact with the force plates. Changes
in the pixels within this region of interest, identified here in the first frame of the video, are then
used to detect when and where the foot contact with the plate occurs.
After narrowing the region of interest, the RGB values for each pixel are converted to grayscale.
The first picture within the video series is used as a baseline (Figure 2.4) to which all of the subsequent images are compared. Frames are converted to image maps storing the grayscale intensity of
the pixels comprising the image. The difference in grayscale values from the first original image
12

map with no foot contact and every frame following is calculated and stored. These difference values are plotted as a series showing the relative change as the foot comes into the frame comparing
the Frame Number (x-axis) versus the frame difference (y-axis) (Figure 2.5).
Figure 2.5: The blue trace shows the comparison difference values for each from of a video. These
pixel changes are compared with the first frame of the video. The purple X indicates the peak
maximum found by the select peaks function. Finding this peak maximum narrows the selection
region down from thousands to tens of frames, saving time for users.
Once the frame difference values are plotted, a function is used to find the maximums throughout the series to identify the peaks of interest. The user confirms with a GUI selection that this
peak contains a portion of the video where the foot enters the frame. This ensures that the foot
contact is detected, rather than non-athletes crossing the frame, shadows, or other events that may
result in large pixel changes. Then the number of frames is reduced from thousands or hundreds of
frames to a modified selection identified by the peak. The user is presented with a limited selection
of frames to choose from containing where the foot contacts the ground.
13

2.4 Results
2.4.1 Database, R, & IBIS
Creating a database that is stored on the IBIS system and encompasses biomechanical data helps
to provide insight into how the individual performed the task. The data can be categorical, such
as event, sex, competition, or discrete, such as jump distance, impulse, or continuous including
force-time, velocity-time curves. Imagine viewing the performance outcome (e.g., jump distance)
and viewing critical information of how the athlete jumped that far (e.g., horizontal and vertical
velocity at take-off) - now expand that over the four years of the Olympiad cycle (Figure 2.6).
For example, two athletes can have the same jump distance in the long jump; however, one
athlete leaves the ground with more vertical velocity while the other leaves with more horizontal
velocity. A coach can use this information to design future training programs. Due to the relational
structure of the database, users can include more context by querying based on categorical (e.g.,
athlete, event, sex, competition) and performance (e.g., jump distance, impulse) variables. Developing this database helps to create a foundation for clinicians and coaches to integrate various data
sources collected over an extended period. Doing so provides a great resource to identify if their
modifications had the desired effect and to track the individual’s progression. Incorporating the
database in a user-friendly environment inclusive of all technical expertise levels transforms the
database into an actionable juggernaut.
Once the data is organized within a database on the server, user-friendly tools were created
through an R Shiny application that is hosted on the same web server as the IBIS system. The application consists of a graphical user interface (GUI) that authenticated users can log into through
their web browser remotely with an internet connection. Once logged in, the application dashboard
contains a variety of tools to display and filter the data available. Two of the current features of the
application include interactively plotting and filtering the data, and linking data points in a graph
and the corresponding video. Interactively plotting and filtering the data allows users to toggle and
14

Figure 2.6: Example of the utility of the R Shiny analysis tool which was used to create a graphical
user interface (GUI) that shows the progression of jump distance over time for nine elite long
jumpers. The gray vertical lines designate the Olympic cycles and enable the user to see a clear
progression during each cycle for individual athletes. Other features in this interactive example are
enable the user to filter by date, wind speed, and athlete’s personal best (max) jump.
15

filter using the dashboard shown in Figure 2.7 to select and display the data they need on the plot
in the GUI.
The R Shiny application can be used to query and analyze datasets longitudinally or query any
combination of data variables needed. The color identifier can be used to set an additional variable
or represent multiple athletes allowing for cross-study analysis between multiple individuals. The
developed application creates a streamlined user interface hosted on the same location as the IBIS
server where users can quickly perform statistical analysis from any remote location.
Additional features added into the GUI include a drag and highlight functionality that allows
the users to click and drag the cursor to highlight any set of data points displayed on the plot and
this will query additional detailed data for the trials associated with this subset of data. This feature
(Figure 2.8, Step 1) allows users to narrow the focus of the query while allowing them to delve
deeper in the analysis of the specified results. Along with the subset of data queried, the link to the
video data associated with the highlighted trials will also be displayed immediately (Figure 2.8,
Step 2). The user is able to click the button to the video link and it will open the video in a separate
video viewer window (Figure 2.8, Step 3). The RShiny application was developed and integrated
into the IBIS system. The web-based statistical analysis application consists of a dashboard that
can be accessed remotely to query customized data sets, display them graphically, and links to any
associated video files.
2.4.2 Foot Contact Detection Algorithm
The foot contact detection algorithm developed reduces the time spent by researchers to align
video with other data. Raw videos from video collections contain thousands of frames, which
would previously require dedicated attention of a researcher to scrub through. The foot contact
detection algorithm allows the user to choose the region of interest with one selection and then
greatly reduces the potential frames for contact for the selection options.
The algorithm will be tested on multiple sets of data including foot contacts during steady-state
running, long jump, and triple jump. Foot contact with the ground was determined for a sample
16

Figure 2.7: This example graphical user interface (GUI) shows a dashboard of the web-based
R Shiny application for statistical analysis. This application allows research to choose which
variables are displayed on each axis and to assess the presence of covariates using a color gradient.
Specific athletes and variables can be selected using the built-in filters and color identifiers. The
dataset, once selected, will be queried and displayed on the GUI plot.
Figure 2.8: Example demonstration of the process using the R Shiny analysis tool to link a trial
to more information about that performance. (1) User is shown a simple plot. (2) User can select
specific trails and additional relative data populates the table below. Here the user is also able to
select to view the video that corresponds to a specific data point. (3) The video appears in a new
window ready to view.
17

set of data during steady state running which will be used as the gold standard to evaluate the
algorithm. Thirteen athletes with 40 videos will be included in the test set for a total of 520 videos.
If the frame of foot contact is included within the five frames selected and displayed in the GUI
(Figure 2.9), that will be classified as a success for the algorithm. The algorithm will then be tested
and supervised by a researcher to determine its effectiveness for long and triple jump contacts. This
reduction of necessary user input saves hours of valuable research time that can be spent advancing
biomechanical knowledge and providing feedback to clinicians, coaches and athletes.
Figure 2.9: Exemplar video image used to assist the user in visual verification of initial foot contact
determination use the Graphical User Interface (GUI).
The IBIS system was created to aid decision-makers and streamline the biomechanics sports
research workflow using FAIR guiding principles for digital assets (Findable, Accessible, Interoperable, Reusable).[1] Advances were made in centralizing data in one secure location that users
can access remotely through the web to upload, share, and review data from collections. The goal
of this study was to expand the utility of the IBIS system by adding novel decision support and
analysis tools to enhance the ability of researchers to organize, process, and analyze the collected
data for multiple uses.
18

Introduction of statistical analysis tools based in R Shiny greatly improves the efficiency of data
query and analysis because the data is compiled into centralized database using FAIR principles.
By providing a web-based application hosted on the IBIS platform, users can remotely access the
system. This provides a huge advantage in flexibility during the biomechanics research workflow
by removing the limitations of local data analysis, processing considerations, and additional sharing methods. All authenticated users have access to the streamlined statistical analysis dashboard
that allows them to quickly filter and select specific data sets to display on the graphical interface.
Additional features allow users to highlight specific subsets of data to review in more detail and
retrieve any associated video or image files for the data. These powerful functionalities remove the
need for any third party programming software installation and processing, greatly streamlining
the data review and analysis phase for researchers, coaches, and athletes.
Incorporating a foot contact detection algorithm using image-processing techniques into the
preprocessing workflow of the IBIS system greatly reduces the time spent by researchers to identify key events used to synchronize multiple sources of data temporally. The algorithm enables
limited user input required in order to efficiently and accurately identify foot contact so video data
can be coordinated with other biomechanical measures such as force data. This allows researchers
to increase time spent working on communicating valuable findings related to performance enhancements and injury mitigation. Together, these tools help to advance the IBIS system and
present a framework for the improvement of biomechanics research by making use of practices
commonly used in imaging informatics.
19

Chapter 3
Can ground reaction force variables pre-identify the probability
of a musculoskeletal injury in collegiate distance running?
3.1 Abstract
The incidence of lower extremity injuries in collegiate distance runners is approximately 20%.
Prospective identification of a runner sustaining a potential injury remains challenging. This exploratory, cross-institutional study sought to determine if ground reaction force (GRF) characteristics during steady-state running could pre-identify competitive collegiate distance runners who
would later sustain lower-extremity injuries. Normative boundaries for ten GRF variables during
braking and propulsion were established for Non-Injured runners using median ± scaled median
absolute deviation (MADN). Runners were split into subgroups (With and Without Impact Peaks
in vertical GRF) to mitigate the influence of impact peaks on GRF variables. We hypothesized
that prior to injury, runners who later developed an injury would have more GRF variables outside
of the normative boundaries than Non-Injured runners. Using Cliff’s method, a rank-based, nonparametric method for comparing two independent groups, we found no statistically significant
difference between the number of variables outside the boundaries for Injured and Non-Injured
runners overall (p=0.17). However, Injured runners Without Impact peaks had more variables
20

outside the normative boundaries than Non-Injured runners (p<0.001). This novel analytical approach demonstrates the potential for pre-identifying collegiate distance runners Without Impact
peaks who may be at risk for injury.
3.2 Introduction
The incidence of injuries to the lower extremity in collegiate distance runners is approximately
20%, and the ability to prospectively identify the likelihood of a runner sustaining an injury remains challenging.[10–12] In addition, female track and field athletes experience 64% of bone
stress injuries (BSIs) across all female collegiate athletes, while males experience 50%.[13] The
annual incidence of BSIs among collegiate distance runners ranges from 3.9-31.3%,[14, 15] with
female runners having higher incidence rates than males.[16] BSIs most commonly occur in the
tibia or metatarsal[16]; however, BSIs of the femur[16], pelvis[17], and sacrum[18] are also prevalent and serious. In addition to BSIs, soft-tissue injuries can cause major disruptions in training
for runners.[19, 20] A systematic review of musculoskeletal injuries found that medial tibial stress
syndrome, Achilles tendinopathy, and plantar fasciitis are some of the most common injuries in
distance runners.[20] Early detection of mechanical loading characteristics contributing to lower
extremity injuries would provide valuable and timely insights for coaches preparing their runners
for competition.
Previous studies have investigated the relationship of univariate biomechanical measures to injury risk in collegiate runners with mixed results.[21–23] These investigations have studied how the
individual biomechanical measures may relate to injury, including ground reaction forces (GRFs),
joint and segment kinematics, and net-joint moments during foot contact. Even with the use of
these advanced measures, there have been contradicting results about the relationship of these
measures to lower-extremity injuries in competitive runners.[24–27] The reasons why a collegiate
runner sustains a lower-extremity injury are multifaceted. However, identifying individuals who
may be training without sufficient recovery time for tissue adaptations to unrepaired microdamage
21

may be possible.[28–30] A recent review of the relationship between GRFs and running-related
injuries highlighted the need for new approaches to address the prevalence of these injuries.[31]
The purpose of this longitudinal, exploratory study was to determine if competitive collegiate
distance runners who went on to experience a lower extremity injury demonstrated significant
differences in only their GRF characteristics prior to the onset of the injury. Our work aims to
characterize measures that could identify all types of running-related injuries for collegiate runners, and this approach differs from previous work by looking concurrently at multiple aspects
of GRF-time curves rather than individual biomechanical measures. To achieve this goal, we first
characterized normative boundaries for Non-Injured, competitive, collegiate runners using 10 GRF
variables.[32] Then, we used these normative boundaries to determine if runners who went on to
develop an injury (Injured runners) could be distinguished from Non-Injured runners. Horizontal
and vertical components of the GRF during braking and propulsion phases of multiple foot contacts
were used to characterize the mechanical loading each runner experienced. Normative boundaries
for these 10 GRF variables were represented as the median ± scaled median absolute deviation
(MADN) for Non-Injured runners. We expected that the combination of these 10 GRF measures
during the braking and propulsion phases would be sufficient to detect an oncoming lower extremity injury. We hypothesized that Injured runners would have more GRF variables outside of the
normative boundaries than Non-Injured runners prior to injury. We also tested the same hypothesis
on two subgroups of runners, With and Without Impact Peaks, in the vertical GRF to mitigate the
possibility of artificially large normative boundaries created by grouping all runners together.
3.3 Methods
3.3.1 Participants
70 competitive, collegiate distance runners volunteered to participate in this three-year crossinstitutional study. Two consecutive data collections were necessary for this analysis to determine
the With and Without Impact peak groups, so 33 (22 females, 11 males) of the 70 runners were
22

included in this analysis. Some of the runners could not participate in consecutive collections due
to schedule restraints or graduated before data could be collected. These athletes were National
Collegiate Athletics Association (NCAA) Division I distance runners from four Pacific (PAC)-12
Conference universities, including the University of Colorado Boulder, the University of Oregon,
the University of Southern California, and Stanford University. Runners provided informed consent in accordance with the Institutional Review Board of the lead institution (University of Oregon).
Injuries that disrupted training were reported by the respective athletic training staff for nine
runners (9 Injured, 24 Non-Injured). These injuries occurred within 5-20 weeks after data collection. Injury was defined as a BSI or musculoskeletal injury to the lower-extremity that caused a
participation restriction affecting at least one practice. All injuries were identified and recorded by
the athletic training staff and reported to the research team. We then divided the runners into Injured and Non-Injured groups. If a runner experienced both a soft tissue injury and a BSI, the data
prior to and closest to the BSI occurrence/diagnosis was analyzed. There were 9 Injured runners
(5 females, 4 males), and of those 9, there were 7 runners who experienced BSIs and 2 who experienced soft-tissue injuries. Of the 7 runners with BSIs, there were 2 metatarsal injuries, 3 tibia
injuries, 1 fibula injury, and 1 femur injury. Of the 2 runners who sustained soft-tissue injuries,
one experienced heel tendinopathy, while the other experienced an ankle lateral ligament sprain in
addition to hamstring dysfunction and a psoas muscle strain on the same leg.
3.3.2 Data Collection
Baseline data collections were conducted during collegiate competitive seasons (cross-country,
indoor track, and outdoor track). We collected GRF data at 1000 Hz using over-ground force plates
(Kistler USA, Amherst, MA) or instrumented treadmills (Treadmetrix, Park City, UT or Bertec,
Columbus, OH) based on the equipment available at each university. All runners were injuryfree at the time of data collection and cleared by their athletic training staff for participation. Both
female and male athletes ran at a common pace of 3.83 m/s (7 min/mi) to replicate a typical training
23

pace. For athletes running over-ground, two single beam timing gates (Brower Timing Systems,
Draper, UT, USA) placed 3.05 m apart and centered on the force plates at hip height were used
to verify pacing was within ±11% of 3.83 m/s at the time of data collection. For athletes running
on instrumented treadmills, 30-second segments were collected at the 3.83 m/s pace. We collected
GRF data for at least eight ground contacts of each leg for every runner.
3.3.3 Analysis
We analyzed GRF data using custom Python (3.9) and R (3.6.1) scripts. Data was filtered
using a fourth-order recursive Butterworth filter to remove noise from the source signal specific
to each data collection site. A cutoff frequency of 50 Hz was implemented for the instrumented
treadmill data, while the cutoff frequency for over-ground force plate data was 75 Hz. The lower
cutoff frequency for the instrumented treadmill data (50 Hz) was chosen to reduce high-frequency
mechanical noise due to the treadmill motor, which is not present in over-ground force plate data.
3.3.4 Ground Reaction Force Variables
A minimum of eight ground contacts at 3.83 m/s were used to characterize 10 GRF variables
measuring time, average GRFs, and change in center of mass velocity (∆v) during both braking
and propulsion.[32] We selected ten variables to characterize the GRFs experienced by the whole
body in the vertical and anterior-posterior directions during the braking and propulsion phases
of ground contact. Contact time (ms) was defined as the interval when the vertical GRF curve
exceeded a threshold of 20 N. Braking time (ms) was the time spent during contact when the
anterior-posterior horizontal GRF was negative. Average vertical GRF during braking (Avg vGRF
Brake) and propulsion (Avg vGRF Prop), and average anterior-posterior GRF during braking (Avg
apGRF Brake) and propulsion (Avg apGRF Prop) were calculated from GRFs during contact. To
characterize impulse generation in runners of different body mass, the change in center of mass
(CoM) velocity during braking and propulsion was calculated using the net impulse applied in the
vertical (∆vv Brake, ∆vv Prop) and horizontal (∆vap Brake, ∆vap Prop) directions (Figure 3.1).
24

Figure 3.1: Example depictions and definitions of the 10-ground reaction force (GRF) variables
used to characterize timing, force, and change in velocity (∆v). GRF curves (black traces) were
used to calculate the given variable. The dashed gray vertical line separates the braking and propulsion phases. The green horizontal lines show the length of braking and contact times. The blue
horizontal lines show the average GRFs in the vertical and anterior-posterior directions during
each phase. The blue area under the GRF curve represents the impulse in the vertical and anteriorposterior directions during each phase. The gray areas represent the vertical impulse due to body
weight (BW), which is subtracted to calculate the change in vertical velocity in each phase. Abbreviations: Avg, average; Prop, propulsion; Brake, braking; v, vertical; ap, anterior-posterior.
25

Previous research has explored various ways to characterize vertical GRF loading rate in the initial
phase of foot contact with mixed results.[24–27] We opted to use Avg vGRF Brake and ∆vv Brake
to capture increases in vertical impulse associated with higher loading rates.
3.3.5 With and Without Impact Peak Subgroups
Recognizing that GRF characteristics during braking are associated with foot strike patterns25
and that impulse may be distributed throughout ground contact, runners were further divided into
subgroups With and Without Impact peaks. A transient vertical GRF impact peak, hereafter referred to as an impact peak, is a change in the positive slope of the vertical GRF versus time curve,
leading to a local maximum.[33] Absence of an impact peak is generally associated with forefoot
striking. We considered runners with vertical GRF impact peaks that exceeded 1.2 times body
weight (BW) within the first 0.035 s of contact in at least 50% of all foot contacts during two separate data collections to have impact peaks. There were 10 runners classified as With Impact Peaks
(7 Non-Injured, 3 Injured) and 17 runners classified as Without Impact Peaks (12 Non-Injured,
5 Injured). Of the 33 individuals, 6 (5 Non-Injured, 1 Injured) could not be definitively classified as With or Without Impact Peaks due to differences in GRF characteristics between their data
collections or between legs, so their data was not included in the subgroup analysis.
3.3.6 Statistics
Statistical analysis was conducted using a custom R script (3.6.1) with functions from the package WRS2.[34] To characterize GRF variables for Non-Injured runners, median values for each
GRF variable were calculated for each leg with a minimum of eight foot contacts per leg. In
addition to the Injured and Non-Injured groups, runners were also split into subgroups With and
Without Impact Peaks. The scaled median average deviation (MADN) was used as a measure of
variance to characterize the boundaries of GRF values. Median and MADN were chosen as robust
statistical measures to replace classic methods using mean and standard deviation, which are biased by the presence of a single outlier.[35] Boundaries were established as the median ± MADN
26

for Non-Injured runners overall and then for subgroups of Non-Injured runners With and Without
Impact Peaks (Figure 3.2). For normally distributed data, 67.45% of Non-Injured runners would
lie within the median ± MADN. We did not detect a statistical difference in the GRF variables
between the Non-Injured runners’ left and right legs, so both were included for defining the NonInjured boundaries (24 runners, 48 legs). There was also no statistical difference between legs for
any of the 10 GRF variables for each subgroup, so both Non-Injured legs were included (With Impact Peaks: 12 runners, 24 legs, Without Impact Peaks: 7 runners, 14 legs). The tests used to verify
there was no statistically significant difference between legs for each GRF variable were the sign
test and percentile bootstrap methods for comparing marginal trimmed means and trimmed means
of difference scores.[35, 36] We tested that the median values of the With and Without Impact Peak
subgroups were the same using a percentile bootstrap method.
Figure 3.2: Non-Injured boundaries for contact time shown on exemplar ground reaction force
(GRF) curves. The solid vertical line is the median contact time, and the dashed vertical lines
are the Non-Injured boundaries, which are ± 1 scaled median absolute deviation (MADN). The
left graph represents the Non-Injured runner boundaries with two sample GRF curves, the middle
graph is an exemplar GRF curve with no transient vertical GRF impact peak, and the right graph
is an exemplar GRF with an impact peak. Abbreviations: BW, body weight.
To compare the results from Non-Injured runners to Injured runners, we determined the number
of GRF variables for each runner outside of the median ± MADN boundaries established by NonInjured runners. Both legs of the Non-Injured runners were included in the analysis, while only
the leg that experienced an injury was included for the Injured runners. A robust distance measure
(RDM) for each leg and each of the 10 GRF variables was calculated as follows:
27

RDM =
Healthy Group Median−Athlete Median
Healthy Group MADN
This RDM was used to determine if a GRF variable for the Injured leg of a runner was outside
the Non-Injured boundaries defined for that particular GRF variable using data from both legs of
Non-Injured runners. For example, an RDM of 3 for Avg vGRF Brake indicates that a runner’s
median value for that variable was 3 MADNs away from the Non-Injured group median. Cliff’s
method28 was used to test the probability that a randomly selected runner who went on to experience an injury would have a greater number of variables outside the boundaries than a randomly
sampled Non-Injured runner. Cliff’s method is a rank-based, non-parametric method for comparing two independent groups. The method aids in making inferences about the p-value when sample
sizes are small. It computes a heteroscedastic confidence interval:
δ = Probability (Non-Inj < Inj)−Probability (Non-Inj > Inj)
We also used Cliff’s method to compare Non-Injured and Injured runners for the With and
Without Impact Peak subgroups.
3.4 Results
3.4.1 GRF Variable Boundaries for Non-Injured Collegiate Runners
We calculated the 10 GRF variable boundaries from both legs of Non-Injured runners overall
(Table 3.1). Further, we separately calculated the boundaries of Non-Injured runners With and
Without Impact Peaks using both legs but found no significant difference between the two subgroups for any individual variable (Table 3.1).
28

Table 3.1: Median and scaled median absolute deviation (MADN) for 10 ground reaction force
(GRF) variables of Non-Injured runners. Note. The group boundary values are presented for NonInjured runners (left), Non-Injured runners Without transient vertical GRF impact peaks (middle),
and Non-Injured runners With Impact Peaks (right).
Abbreviations. BW, body weight; Avg, average; Prop, propulsion; Brake, braking; ∆v, change in
velocity; v, vertical; ap, anterior-posterior.
3.4.2 GRF Variables Outside Normative Boundaries for Injured and NonInjured Runners
Overall, we did not detect a significant difference in the number of GRF variables outside of the
normative boundaries between the 24 Non-Injured runners and the 9 Injured runners (5 females, 4
males) (p=.17, Table 3.2)). Among the 9 Injured legs, 3 had 5 or more GRF variables outside the
boundaries for Non-Injured legs (Figure 3.3). The median number of GRF variables outside the
normative boundaries was 3 for Non-Injured runners and 4 for Injured runners (Figure 3.5).
29

Table 3.2: Statistical comparison using Cliff’s method of the number of ground reaction force
(GRF) variables outside of Non-Injured normative boundaries between both legs of Non-Injured
runners versus the injured legs of Injured runners.
Note. Cliff’s method was used as a rank-based, non-parametric technique to compare the legs
of two independent groups (Non-Injured legs vs. Injured legs). Both legs were included for
the Non-Injured runners, however, only the leg that experienced an injury was included for the
Injured runners. Probabilities indicate the likelihood of a randomly sampled Injured runner’s leg
having more, equal, or fewer of the ten ground reaction force variables outside of the Non-Injured
boundaries than a randomly sampled Non-Injured runner’s leg. Boundaries were defined as the
median ± scaled median absolute deviation (MADN) for each GRF variable for the Non-Injured
runner group. A heteroscedastic confidence interval (δ) is reported. Results are broken down for
runners With (right) and Without transient vertical GRF impact peaks (middle). Abbreviations.
Non-Inj, Non-Injured; Inj, Injured; Prob, probability; ∗, Statistically significant result p<0.05.
However, a different story emerged when we examined only the runners Without Impact Peaks
(i.e., forefoot strikers). Among the Injured legs Without Impact Peaks, 4 of the 5 exhibited 6 or
more GRF variables outside the normative boundaries (Figure 3.4, Orange). For this sub-group,
we did detect a significantly greater number of GRF variables outside of the normative boundaries
for the Injured vs. Non-Injured runners (p<0.001, Table 3.2). Among all of the 12 runners Without
Impact Peaks, the median number of variables outside the normative boundaries was 4, compared
to 6 among the Injured runners (1 female, 4 males) (Figure 3.5). Of the 24 Non-Injured legs
Without Impact Peaks, 3 legs had more than 5 GRF variables outside of the normative boundaries
(Figure 3.5).
Just as our overall analysis showed, among the runners With Impact Peaks, we did not detect
significant differences between the 7 Non-Injured and 3 Injured runners (all female) in terms of
the number of GRF variables outside of the normative boundaries (p=0.41, Table 3.2). The median
30

value of the number of GRF variables outside the normative boundaries was 3 for Non-Injured
runners (Figure 3.5). Only 1 of the 3 legs had 5 variables outside of the normative boundaries
(Figure 3.4, Purple), which was more than what was typical for Non-Injured runners With Impact
Peaks (Figure 3.5, Purple Dashed Lines).
3.5 Discussion
In this longitudinal, exploratory study, we used a cross-institutional approach to investigate if
competitive collegiate distance runners who experienced lower-extremity injuries could be identified using only their GRF characteristics prior to the onset of injury. We first characterized normative GRF variable boundaries for Non-Injured, competitive, collegiate athletes during steady-state
running at a typical training pace. These 10 GRF variables were comprised of times, average
GRFs, and changes in velocity during braking and propulsion, and our findings were comparable
with similar measures reported previously for Non-Injured collegiate runners.[37] We found that
among the Without Impact Peak subgroup, Injured runners could be identified from Non-Injured
runners by the number of GRF variables outside the Non-Injured boundaries. For the subgroup
of runners With Impact Peaks, we did not find evidence to indicate that Injured runners had more
GRF variables outside of the MADN boundaries.
This exploratory study is a valuable next step in establishing normative GRF characteristics
for Non-Injured competitive collegiate distance runners, which may provide beneficial comparative data for coaching and athletic training staff. The results of this study are promising because
they provide preliminary evidence that runners Without Impact Peaks who are about to sustain an
injury could be distinguished from other Non-Injured collegiate distance runners using only GRF
variables. Runners in this study Without Impact Peaks had more GRF variables outside of the
normative boundaries than Non-Injured runners. The statistical analysis using Cliff’s method indicates an 86% probability that an Injured runner Without Impact Peaks would have more variables
outside normative boundaries than a Non-Injured runner (Table 3.2). Because the absence of an
31

impact peak is associated with forefoot striking, athletic training staff and university biomechanists
could use this GRF variable approach to screen runners with forefoot strike patterns for potential
future injuries. While we did not find a statistical difference for runners With Impact Peaks (primarily midfoot and rearfoot strikers), there were only 3 Injured runners out of 10 total runners in
this subgroup, whereas there were 5 Injured runners out of 17 total runners Without Impact Peaks.
Future studies with larger sample sizes are needed to confirm the results of the present study.
Of interest in this exploratory study was that some Non-Injured runners had a high number of
GRF variables outside of the normative boundaries but avoided injury (Figure 3.5). Despite having
many GRF characteristics that were different than their Non-Injured peers, there may be factors
enabling these runners to avoid lower-extremity injuries. Additionally, some Injured runners had
few GRF variables outside of the normative boundaries but nonetheless sustained an injury (Figure
3.5). This may be due to other factors leading to insufficient tissue adaptation to repetitive loading.
Further studies that integrate biomechanics along with physiological measures may provide more
context as to why some runners had atypical GRF measures but didn’t sustain an injury or vice
versa.
Despite the participation and successful coordination of four universities using an integrated
biomechanics informatics system, these findings may be limited by a relatively small sample
size.[4] A greater sample size may also reveal whether runners with different injury types have
different GRF variables that exceed boundaries defined by Non-Injured runners. Our statistical
power could have also been strengthened with more frequent data collection so that the GRF variables analyzed for the Injured groups were collected closer to the time of reported injuries. More
frequent data collections may have reduced temporal variability and provided more evidence for
the potential association between biomechanical variables and musculoskeletal injuries. Even with
these potential limitations, our multi-institutional approach presents a novel framework for using
GRF variables to predict the potential of an oncoming injury in individuals Without Impact Peaks.
We showed that this new analytical approach, which takes multiple characteristics of the GRF
32

time curves into account, shows promise for prospectively identifying different types of lowerextremity injuries for collegiate runners Without Impact Peaks. Our theoretical framework forming
the basis of this study is that multiple GRF-time curve variables that characterize times, average
forces, and changes in center of mass velocity may be able to pre-identify athletes who go on to
sustain running-related musculoskeletal injuries. Rather than assuming oncoming injuries may be
associated with one biomechanical measure, this framework accounts for the influence of multiple
factors or a combination of factors that may identify different types of injuries. The results of this
study provide preliminary evidence that runners Without Impact Peaks who are more likely to be
injured could be identified from Non-Injured runners because they are likely to have more GRF
characteristics outside of normative boundaries established by Non-Injured runners.
33

Figure 3.3: Top Row: Exemplar Non-Injured runner with 1 ground reaction force (GRF) variable
outside the Non-Injured boundaries (blue). Bottom Rows: Nine Injured runners with varying
GRF variables outside the Non-Injured boundaries. The green region indicates the Non-Injured
boundaries defined as the median ± scaled median absolute deviation (MADN) of all 24 NonInjured runners. The circles represent the robust distance measurement (RDM) from the median
of each runner for each GRF variable. RDM of zero indicates that the runner’s individual GRF
variable matched the median value from the Non-Injured runners. Positive RDM indicates greater
forces, longer contact and braking times, and greater changes in velocity than the median NonInjured value. Negative RDM values indicate smaller forces, shorter braking and contact times,
and smaller changes in velocity than the median Non-Injured value. Abbreviations: Avg, average;
Prop, propulsion; Brake, braking; ∆v, change in velocity; v, vertical; ap, anterior-posterior.
34

Figure 3.4: Eight runners with different lower extremity injuries with varying GRF variables outside the Non-Injured boundaries. The orange region indicates the Non-Injured boundaries defined
as the median ± scaled median absolute deviation (MADN) of the 12 Non-Injured runners Without
impact peaks. The purple region indicates the Non-Injured boundaries defined by the 7 runners
With impact peaks. The circles represent the robust distance measurement (RDM) from the median of each runner for each GRF variable. RDM of zero indicates that the runner’s individual GRF
variable matched the median value from the Non-Injured runners. Positive RDM indicates greater
forces, longer contact and braking times, and greater changes in velocity than the median NonInjured value. Negative RDM values indicate smaller forces, shorter braking and contact times,
and smaller changes in velocity than the median Non-Injured value. Abbreviations: Avg, average;
Prop, propulsion; Brake, braking; ∆v, change in velocity; v, vertical; ap, anterior-posterior.
35

Figure 3.5: The number of Non-Injured runners’ legs (open bars) and Injured (Inj) runners’ injured
legs (grey bars) with ground reaction force (GRF) variables outside the normative boundaries from
Non-Injured (Non-Inj) runners. Solid vertical lines indicate the median (Med) for each group, and
dashed vertical lines indicate each group’s scaled median absolute deviation (MADN). Top: Green
lines indicate the distribution of the 48 Non-Injured runners’ legs, and gray lines indicate the distribution of the 9 Injured runners’ injured legs. Middle: Orange lines indicate the distribution of
the 24 Non-Injured runners’ legs, and gray lines indicate the distribution of the 5 Injured runners’
injured legs Without transient vertical GRF impact peaks. Bottom: Purple lines indicate the distribution of the 14 Non-Injured runners’ legs, and gray lines indicate the distribution of the 3 Injured
runners’ injured legs With transient vertical GRF impact peaks.
36

Chapter 4
Comparison of impulse regulation and lower extremity control
between full jump and popoff takeoffs in the long jump
4.1 Abstract
During competition, a long jumper uses their capabilities to maximize horizontal jump distance
beyond the scratch line by regulating impulse generation during the last contact prior to flight
(takeoff). In training, jumpers perform popoffs, a task believed to replicate the run-jump transition
of a full jump without landing in the sand. To determine differences in impulse regulation and
lower extremity control of full jumps and popoffs, ground reaction forces (GRFs) and segment
kinematics measured during takeoff performed by competitive elite jumpers (n=15) during training
were compared. At the whole-body level, no significant differences in the change in center of
mass (CM) velocity (vertical and horizontal) during takeoff between full jumps and popoffs were
observed for the group. For both tasks, greater leg angles at initial contact were associated with a
larger loss in horizontal CM velocity and greater gains in vertical CM velocity during takeoff. No
significant differences in magnitude or percent distribution of net joint moment impulse across the
ankle, knee, and hip were observed during impact, post-impact, or foot contact between full jumps
and popoffs. The results for these participants suggest that their popoffs impose similar impulse
regulation and lower extremity control challenges as in their full jumps.
37

4.2 Introduction
During a long jump competition, the performance objective is to maximize the horizontal distance the jumper travels from the scratch line.[38] Maximizing the center of mass (CM) horizontal
displacement during the flight phase can be achieved with various horizontal and vertical CM velocity combinations at final contact prior to flight. The individual’s ability to generate horizontal
velocity during the run-up and the ability to regulate impulse during takeoff will define what combination of CM vertical and horizontal velocities at the last contact will produce their maximum
jump distance.[39] One way to increase the amount of vertical CM velocity generated during takeoff is to increase incoming horizontal CM velocity. However, the jumper must be able to effectively
regulate impulse generation and multi-joint control of the support leg during takeoff to successfully
use the horizontal CM velocity generated during the approach.[40]
Prior studies of the takeoff phase of the long jump have shown that as more horizontal momentum is lost during the last foot contact prior to flight, more vertical velocity is typically gained.[41–
43] To convert the horizontal momentum generated in the last steps of the approach to vertical momentum, an effective control strategy is to place the foot more anterior to the CM to create a larger
leg angle at initial contact of takeoff prior to flight. This allows the jumper to generate more negative horizontal impulse (also known as braking impulse).[44] However, individual control strategies need to be considered because increases in leg angle at initial contact did not increase vertical
CM velocities at final contact for every jumper.[44] While trajectory of the CM during flight is
determined by the horizontal and vertical CM velocities at final contact of takeoff, jumpers must
also strategically control their body segments in flight to regulate their angular momentum in ways
that add to their measured jump distance during landing.[45]
In training, coaches ask jumpers to perform popoffs, a task believed to replicate the run-jump
transition of a full jump without landing in the sand. Popoffs consist of jumpers completing their
run-up approach and takeoff as done in a full jump without performing the landing in the sand
as completed in a full jump. In this study, we sought to determine whether the performance of
popoffs in training sufficiently replicates the impulse regulation and multi-joint control of the lower
38

extremity used by the individual during their full jump. This work will provide value to coaches
and athletes by determining if popoffs sufficiently replicate the whole-body and multi-joint control
needed for an effective takeoff in the long jump.
By advancing our understanding of differences in the whole-body or multi-joint control between full jumps or popoffs for long jumpers during takeoff we can provide important insights to
guide coaching decisions made in practice.[39, 46] On a whole-body level, we expected differences
in the way jumpers achieve the performance objectives when performing full jumps as compared
to popoffs.[47] For this group of elite jumpers, we hypothesized the loss in horizontal CM velocity
and gain in vertical CM velocity during takeoff would be consistent for full jumps and popoffs.
We also hypothesized that a greater leg angle at initial contact would contribute to greater losses
in horizontal CM velocity and greater gains in vertical CM velocity during takeoff for both full
jumps and popoffs. At the multi-joint control level, we expected that jumpers would use different
strategies for controlling the lower extremity between full jumps and popoffs because of the need
to land the full jump. To address this, we hypothesized that jumpers would have greater net joint
moments (NJMs) impulse magnitudes at the ankle, knee, and hip during full jumps than popoffs
during impact, post-impact, and foot contact. We also hypothesized that NJM impulse percent
distributions across the ankle, knee, and hip would differ for full jumps and popoffs for the same
phases. This study will help coaches better understand how regulation of impulse and control of
the leg during the takeoff of popoffs during practice can prepare their jumpers for the mechanical
demand imposed at whole-body and joint levels of a full jump.
4.3 Methods
4.3.1 Participants
Data from 11 female (age = 25.1 ± 3.2 years, mass = 62.6 ± 5.7 kg, Long Jump PR = 6.61 ± 0.15,
mean ± standard deviation) and 4 male long jumpers (age = 26.1 ± 1.6 years, mass = 72.7 ± 5.7
kg, Long Jump PR = 7.93 ± 0.56) were used in accordance with the Institutional Review Board for
39

research involving human subjects. One participant was a healthy unilateral trans-tibial amputee
who jumped off of their prosthetic leg at takeoff. Using the framework outlined by McKay and
colleagues, jumpers were classified as elite (8) and world-class (7).[48]
4.3.2 Tasks
Each jumper completed their regular warm-up routine prior to completing full jumps or popoff
attempts. During the popoff, the jumpers performed the long jump without completing the landing
in the sand as done during competition. During the full jump, jumpers performed the long jump as
done during competition.
4.3.3 Data Collection
Data was collected at the Chula Vista Elite Athlete Training Center over a period of eight years
using an instrumented jump runway.[49] The timing of the data collections during the competitive
season were selected by the jumper’s coach and jumpers took part in multiple collections, depending on their availability. Kinetic and kinematic data were collected during the takeoff of the full
jumps and popoffs. Kinematic data was collected using high-speed video at 240 Hz (240 fps, Panasonic LUMIX GH5S, Osaka, Japan or Casio, Dover, NJ). Kinetic data was collected using force
plates (0.6 x 0.9 m) sampled at 1200 Hz (Kistler 9287A, Amherst, NY, USA). Force and kinematic
data were synchronized at initial contact of takeoff.
4.3.4 Analysis
Jumpers who completed both a full jump and a popoff were included in this analysis. Regulation of horizontal and vertical CM velocity and multi-joint control of the lower extremity during
takeoff for full jumps and popoffs were analyzed using the following variables. A 60 N threshold
was used to determine initial contact (IC) and final contact (FC) using the vertical component of
the ground reaction force (GRF) measured during takeoff which is defined as the last foot contact
40

Figure 4.1: Ground reaction force (GRF) time-curves and filmstrips for an exemplar jumper’s full
jump and popoff. Bold gray vertical line divide impact and post-impact. Green arrows on the
filmstrips show the GRF vector at that particular event. Abbreviations: Fv: Vertical GRF; Fh:
Horizontal GRF.
of the run-jump transition (Figure 1). GRFs were split into two phases: impact and post-impact
(Figure 1).[50] Impact was defined from IC in the vertical GRF to the local minimum after the peak
(Figure 1). Post-impact was defined from the local minimum in the vertical GRF to FC (Figure 1).
Change in horizontal CM velocity and vertical during takeoff was calculated from the GRF-time
curves and body weight GRF measures from the same day using the net impulse-change in momentum relationship (Figure 1). Leg angle was calculated as the angle from hip to ankle relative
to the right horizontal at IC of the takeoff (Figure 1).[44]
Body segment kinematics were calculated from landmarks digitized by one individual from
high-speed video. Software for markerless motion capture (VAMA 1.5, Southwest Research Institute) was implemented, inspected visually, and revised manually for accuracy. The landmarks
digitized on the body were head apex, C7, shoulder, elbow, wrist, hand, hip, knee, ankle, heel, and
toe.[51] These sagittal plane digitized coordinates were filtered using a zero-phase lag, Butterworth
filter using a cutoff frequency of 10 Hz. To estimate the total body CM and segment kinematics,
body segments of an athletic population were used.[51, 52] Kinematic data was synchronized at
41

initial contact and interpolated to match the force data sampling rate. Net joint moments (NJMs)
over time were calculated for each attempt using inverse dynamics and are the minimum moments
required at the ankle, knee, and hip to obtain the observed kinematics.[53] NJM impulse was calculated as the total positive and negative area under the moment-time curve for the ankle, knee, and
hip. Multi-joint control was represented as the magnitude of the NJM impulse at each joint and a
percentage. This percentage was the relative contribution of each joint to the total lower-extremity
moment impulse. Magnitudes and percentages of NJM impulses for each joint were calculated
during impact, post-impact, and foot contact.
4.3.5 Statistics
The following statistical analyses examined how the whole-body and multi-joint control of full
jumps and popoffs compared for this group of elite to world-class long jumpers. Statistical analysis was conducted using R and the package Rallfun-v43.[35] Checks for bad leverage points were
made using a regression method aimed at accomplishing this goal. If none were found, a correlation method was used that accounted for the overall structure of the data without specifying
a particular regression estimator.[36] A Theil-Sen estimator was used to fit a linear model based
on the variables of interest. The data from the unilateral amputee jumper was excluded from the
correlations and regression equations. Correlations between leg angle at IC and change in vertical
and horizontal CM velocity during takeoff were also determined for full jumps and popoffs.
To better understand the potential difference in the multi-joint control of the lower extremity
during full jumps and popoffs, the median and normalized median absolute deviation (MADN) of
the magnitudes and relative magnitudes of the ankle, knee, and hip, during impact, post-impact,
and foot contact were reported within jumper and for the group. Full jumps and popoff were compared using a statistical test comparing two dependent groups by computing the trimmed means of
the difference scores with a percentile bootstrap method.[35] Difference scores were computed for
full jump and popoff NJM impulse magnitudes and percent distributions for the ankle, knee, and
hip, during impact, post-impact, and foot contact.
42

4.4 Results
Figure 4.2: In full jumps and popoffs, 15 elite to world-class long jumpers generated similar
changes in center of mass (CM) horizontal and vertical velocity during takeoff (Top). There was
a statistically significant correlation between the horizontal velocity lost and the vertical velocity
gained during takeoff for both full jumps and popoffs (Top). The same data was plotted with a
color gradient to show incoming horizontal velocity, demonstrating that similar changes in CM
vertical velocity can be achieved with lower incoming horizontal velocities (Bottom). Despite
coming in with lower incoming horizontal velocities, there were female long jumpers who generated comparable changes in CM velocity during the takeoff to their male counterparts in both
full jumps and popoffs (Bottom). The two data points in the top right corner are from a unilateral
trans-tibial amputee jumper with a prosthesis on their takeoff leg. This jumper was not included
in the correlations or regression equations for the 14 jumpers without prosthetics. The unilateral
amputee jumper was able to lose less horizontal CM velocity while generated similar changes in
vertical CM velocity to the jumpers without prosthetics (Bottom). The solid gray line corresponds
to the full jump regression, and the black dotted line corresponds to the popoff regression. Note.
*: Statistically significant.
43

4.4.1 Whole-Body: Group
For both full jumps and popoffs, there was a statistically significant correlation between the loss
in horizontal CM velocity and the gain in vertical CM velocity during a long jump takeoff for
this group of jumpers (FJ: -.89*, Popoff: -.85*, Figure 2, Top). While there was a statistically
significant correlation (.83*) between larger leg angles at initial contact and losing more horizontal
CM velocity for full jumps, there were no other statistically significant correlations between leg
angle and change in CM velocity for full jumps and popoffs. Jumpers with the highest incoming
horizontal CM velocities near 10 m/s were typically males and increased their vertical CM velocity
by 3 m/s or more during takeoff (Figure 2, Bottom). However, there were female jumpers with
incoming horizontal CM velocities around 8 m/s who increased their vertical CM velocity by 3.2
m/s or more on both full jumps and popoffs (Figure 2, Bottom). The unilateral amputee jumper
gained 3.2 m/s or more vertical CM velocity while losing less horizontal CM velocity than jumpers
without prosthetics during takeoff. However, this jumper also had lower incoming horizontal CM
velocities than the other male jumpers without prosthetics for their full jump and popoff (Figure 2,
Bottom). For the jumpers without prosthetics, no statistical difference was detected in contact time
between full jumps and popoffs. The median contact time for full jumps was 130 ms (MADN: 11
ms), and for popoffs, the median was 127 ms (MADN: 12 ms).
4.4.2 Whole-Body Analysis
For the majority of the jumpers, a larger leg angle at initial contact contributed to larger losses in
horizontal CM velocity and greater gains in vertical CM velocity during takeoff of both full jumps
and popoffs (Figure 3). For 12 of the 15 jumpers, initiation of contact with a larger leg angle at
initial contact resulted in a greater corresponding change in CM velocity during takeoff, regardless
of whether that attempt was a full jump or popoff (Figure 3). Nine of the 15 jumpers generated a
greater change in horizontal and vertical CM velocity on their full jump than their popoff, including
the unilateral amputee jumper with a prosthetic on their takeoff leg (Figure 3, Left, Red Arrows).
There were 5 jumpers who generated a smaller change in horizontal and vertical CM velocity on
44

Figure 4.3: Arrows connect each long jumper’s full jump to their popoff attempt. Red arrows
indicate that jumpers generated more vertical CM velocity and lost more horizontal CM velocity
on their full jump takeoff than their popoff (8 jumpers), while blue arrows indicated that jumpers
generated less vertical CM velocity and lost less horizontal CM velocity on their full jump than
their popoff (5 jumpers, Left). Gray arrows indicate that a different pattern was found. Numbers
above the full jump indicate each jumper’s participant number. Jumper 7 is a unilateral transtibial amputee with a prosthetic on their takeoff leg. Starting with a larger leg angle at initial
contact seemed to contribute to the greater loss in horizontal CM velocity and greater gain in
vertical CM velocity during takeoff for 12 of 15 jumpers during full jumps and popoffs, as shown
by the matching arrow colors on each plot (Left, Right). If a jumper had a larger leg angle at
initial contact, they typically generated a greater loss in horizontal CM velocity and greater gain in
vertical CM velocity during takeoff for both full jumps and popoffs (Right). Red arrows indicate
that 7 jumpers lost more horizontal CM velocity and started with a larger leg angle at initial contact,
while 4 jumpers did the opposite (Right).
their full jump than their popoffs (Figure 3, Left, Blue Arrows). One jumper had essentially the
same change in CM velocity during takeoff for their full jump and popoff. There were two jumpers
that did not follow the two predominant patterns, and the leg angle at initial contact did not explain
differences in their change in CM velocity (Figure 3, Right, Gray Arrows).
45

Figure 4.4: Net joint moment (NJM) impulses for full jumps and popoffs organized by individual
for 14 elite to world-class long jumpers without prosthetics at the ankle, knee, and hip during foot
contact (Top), impact (Bottom Left), and post-impact (Bottom Right). Vertical gray lines separate
the data for each jumper with their full jump attempt (left) and then their popoff attempt (right).
4.4.3 Multi-Joint Control of the Leg
No significant differences were found between full jumps and popoffs in NJM impulse magnitude or percent distribution at the ankle, knee, or hip during impact, post-impact, or foot contact for
the 14 jumpers without prosthetics (Table 4.1). The confidence intervals of the difference scores
between full jumps and popoffs for all conditions were within the median difference score for the
group ± the MADN difference score for the group. Data for each individual’s full jump and popoff
is shown for impact, post-impact, and foot contact (Figure 4). Jumpers regulated the mechanical
demand imposed on the ankle, knee, and hip by generating greater NJM impulses in post-impact
compared to impact for both full jumps and popoffs (Figure 5, Middle).
46

Table 4.1: Median and normalized median absolute deviation (MADN) are reported for the net
joint moment (NJM) impulse magnitude and percent distributions to the overall moment impulse
for 14 elite to world-class long jumpers without prosthetics during impact, post-impact, and foot
contact. There were no significant differences in NJM impulse magnitude or percent distribution
between full jumps and popoffs at the ankle, knee, or hip. During impact, long jumpers controlled
the leg primarily using the hip, while in post-impact, the multi-joint control was dominated by the
knee and ankle.
During impact, the hip dominated leg control for full jumps and popoffs (Table 4.1, Figure 5;
Top). The center of pressure (COP) was located near the foot CM, but larger GRF magnitudes
contributed to the need for a larger ankle NJM to control the leg (Figure 6). The orientation of
the net joint forces (NJFs) fluctuated but generally aligned with the shank during impact, resulting
in a relatively low contribution by the knee NJM. However, the large magnitude of the NJFs in
relation to a more horizontal thigh contributed to the larger hip NJMs needed to control the leg
during impact. (Figure 6, Figure 5; Top).
In contrast, during post-impact, the ankle and knee dominated control of the leg for both full
jumps and popoffs (Table 4.1, Figure 5; Middle). As the COP shifted closer to the toe throughout
post-impact, the GRF and ankle NJF caused a positive moment that had to be controlled by a negative ankle NJM. The NJFs were less aligned with the shank throughout post-impact than impact,
causing a negative moment countered by the ankle and knee NJMs (Figure 6). Despite this difference in NJFs alignment with the shank, the NJFs were of a smaller magnitude resulting in smaller
NJM needed to control the leg during post-impact. Throughout post-impact, the orientation of the
47

Figure 4.5: Net joint moment (NJM) impulses for full jumps (Left) and popoffs (Right) for a group
of 14 elite to world-class long jumpers without prosthetics at the ankle, knee, and hip during impact
(Top), post-impact (Middle), and foot contact (Bottom). Data is ordered left to right by smallest
to largest leg angle at initial contact. No significant difference was found in the NJM impulse
magnitudes between full jumps and popoffs at the ankle, knee, or hip during impact, post-impact,
and foot contact during long jump takeoffs.
NJFs aligned more closely with the thigh, thereby requiring smaller hip NJMs to control the leg
(Figure 6).
48

Figure 4.6: Net joint moment (NJM) time curves for an exemplar long jumper’s popoff attempt
during takeoff. (Top) Filmstrip of the joint kinetics at key instances in time with associated free
body diagrams (FBDs) and NJMs at the ankle, knee, and hip. The blue lines on the FBDs are
the net joint forces, and the circles correspond to the NJMs. The circle diameter indicates the
magnitude of the NJM, while the color indicates direction.
4.5 Discussion
In this work, we sought to better understand if differences existed between full jumps and
popoffs in training for elite to world-class long jumpers on the whole-body and multi-joint control levels. On a whole-body level, group results indicated that in both full jumps and popoffs,
jumpers’ loss in horizontal CM velocity were associated with gains in vertical CM velocity during
takeoff. For individuals at the whole-body level, we found that jumpers lost more horizontal CM
velocity and gained more vertical CM velocity during takeoff of both the full jumps and popoffs
49

by positioning their foot further in front of their CM, creating a larger leg angle at initial contact
of takeoff. Differences on a whole-body level in the change in CM velocity between full jumps
and popoffs could be explained primarily by the leg angle at initial contact. From a multi-joint
control perspective, this group of jumpers did not have significant differences in the NJM impulse
magnitude or percent distributions between full jumps and popoffs for the ankle, knee, or hip during impact, post-impact, or foot contact. However, we found that for both full jumps and popoffs,
the hip dominated the control of the leg during impact, whereas the ankle and knee primarily controlled the leg during post impact. Overall, when looking at the whole-body mechanical demands
and multi-joint control of the leg during a long jump takeoff, popoffs appear to be an effective way
to prepare jumpers for the demands of their sport.
While this work was able to shed light on how full jumps and popoffs compare during takeoff
for long jumpers, there were some limitations associated with these advancements. The opportunity to measure GRFs during takeoff is a rare event, and due to these constraints, we were only able
to gather one full jump and popoff attempt for this group of fifteen jumpers. Adding to the sample
size would allow the continued exploration of underlying mechanics used by jumpers in practice.
This exploratory study also had eight jumpers who did not perform their full jump attempt on the
same day as their popoff. Because the data was collected on different days, the jumpers may have
been in a different part of their training cycle or competitive season with different neuromuscular
capabilities. While the segment lengths of the individual likely remain the same across collections,
the capacity of the individual may vary in terms of generating incoming horizontal CM velocity,
regulating the GRF impulse during takeoff, or controlling the ankle, knee, and hip of the takeoff
leg. Collecting the attempts on the same day could provide a different perspective for a more controlled “within-athlete” experiment, provided that it aligns with the practice plan of the coaches.
Reporting the NJM impulses as a magnitude and percent distribution of total lower-extremity moment impulse was an attempt to quantify the minimum NJM impulse needed to perform the task
and normalize what was needed to control the leg during impact and post-impact on the day it was
performed. Despite these limitations, this work fills a gap in the current literature and provides
50

meaningful insights to guide coaching decisions in practice.
On a whole-body level, we found that this group of elite to world-class jumpers lost a comparable amount of horizontal CM velocity and gained a comparable amount of vertical CM velocity
during takeoff for full jumps and popoffs. This was consistent with findings reported in the long
jump literature about the change in CM velocity generated during takeoff for full jumps.[44] Practically, this means that the whole-body mechanical demands of the task were met using similar
strategies regardless of if a jumper was doing a full jump or a popoff. As expected, jumpers with
faster incoming horizontal CM velocities generated greater changes in CM velocity during takeoff.
However, it was interesting that female jumpers who initiated contact with horizontal CM velocities near 8 m/s were able to generate changes in CM velocity that were similar to male jumpers
who came in at 10 m/s. This indicates that female jumpers are capable of withstanding the high
forces and converting their horizontal CM velocity to vertical CM velocity during the takeoff at a
level comparable to male jumpers. It was also interesting that the unilateral amputee jumper came
in with lower incoming horizontal CM velocities than the male jumpers without prosthetics but lost
less horizontal CM velocity while generating similar increases in vertical CM velocity. Therefore,
the prosthetic may enable more vertical CM velocity to be generated during takeoff without the
same proportional loss of horizontal CM velocity. However, a jumper with a prosthetic on one
leg may be at a disadvantage in generating faster incoming horizontal CM velocities, which was
consistent with previous literature studying unilateral amputees.[54]
At the whole-body level, our within-individual analysis revealed that leg angle at initial contact
influenced the change in CM velocity generated during takeoff for both full jumps and popoffs.
If a jumper positioned their leg further out in front of their CM at initial contact of takeoff, they
lost more horizontal CM velocity and gained more vertical CM velocity when completing a full
jump or popoff. This was consistent with findings for the group trends from previous literature
investigating full jumps.[44] For 12 of the 15 jumpers in this study, if they positioned their foot
further in front of their CM at initial contact, they had a larger change in CM velocity during takeoff, regardless of whether that attempt was a full jump or a popoff. The jumper with a prosthetic
51

generated more vertical CM velocity during their full jump attempt by positioning the prosthetic
further in front of their CM at initial contact than in their popoff attempt. The leg angle at initial
contact for this jumper was substantially smaller than jumpers with intact legs during both the full
jump and popoff because of the prosthetic shape. Overall, for this group, whole-body mechanical differences during takeoff found between full jumps and popoff attempts within an individual
could be explained largely by the leg angle at initial contact, which is similar mechanically to how
the influence of touchdown distance at initial contact of foot contact in a sprint acceleration affects
whole-body mechanics.[55, 56]
When looking at the multi-joint control for the group, there was no statistically significant
difference between NJM impulse magnitudes or percent distributions for full jumps and popoffs.
For coaches aiming to prepare their jumpers for the demands of a full jump, this indicates that
for high-level jumpers, popoffs may be an effective way of training for the multi-joint control demands during a full jump takeoff. We also found that the multi-joint control demands and priorities
shifted from impact to post-impact for both full jumps and popoffs. Jumpers generated more of
the NJM impulse magnitude needed to control the leg during post-impact of takeoff. This greater
demand was driven by the longer post-impact time duration because the magnitude of the NJMs in
post-impact at each was smaller for most of the jumpers. Regarding multi-joint control priorities,
during impact, jumpers used a hip-dominant strategy to control the leg for both full jumps and
popoffs. However, later in post-impact, the knee and ankle dominated the control of the leg for
both the full jump and popoff. Consistent with other studies, segment orientation, GRF differences,
and adjacent NJMs influenced the differences in control strategies during impact and post-impact
across the lower-extremity joints.[57, 58] These findings indicate that popoffs can be an effective
way to prepare elite jumpers for the demands of a full long jump takeoff from a multi-joint control
perspective.
This work investigated how long jumpers regulate the whole-body and multi-joint demands of
full jumps and popoffs and advanced our understanding of how popoffs done in training prepare
elite jumpers for competition. For this group of elite to world-class jumpers, there were similar
52

regression equations and correlation values for the horizontal CM velocity lost and vertical CM velocity gained during takeoff. On an individual level, differences in a jumper’s full jump and popoff
could primarily be explained by the leg angle at initial contact. When looking at the multi-joint
control of the leg during takeoff, there was no statistically significant difference between NJM impulse magnitudes or percent distributions of full jumps or popoffs at the ankle, knee or hip during
impact, post-impact, or foot contact. While the demands on the leg were similar for full jumps and
popoffs, we found that in impact, the hip primarily dominated control of the leg, while in postimpact, the ankle and knee dominated control of the leg. These findings on the whole-body and
multi-joint control level indicate that popoffs completed in practice seem to be an effective way to
prepare this group of elite to world-class jumpers for the demands of full jumps in training. This
study successfully added new information about the mechanical demands of popoffs to support
coaches in making appropriate training decisions for their jumpers.
53

Chapter 5
Multi-joint control strategies sprinters use to regulate impulse
generation during the first foot contact out of the blocks
5.1 Abstract
Quick generation of horizontal impulse in the first foot contact out of the blocks is needed by
sprinters to accelerate the body toward the finish line. The purpose of this study was to determine
how body configuration at initial foot contact and impulse regulation was correlated to changes
in total body horizontal momentum and multi-joint control during the first foot contact out of the
blocks by highly-trained sprinters. Measurements of ground reaction forces and segment kinematics during sprint starts performed during a training session revealed that positioning the foot further
behind the center of mass (CM) at initial contact significantly correlated with faster contact times,
greater average horizontal forces, and increased mechanical demand (net joint moment impulse)
on the knee and hip during impact. Increases in average horizontal forces were significantly correlated with smaller magnitudes of shank angular velocity during impact and faster times to peak
thigh angular velocity during post-impact. We also found that we could use these same measures
to provide timely, relevant, usable, and easy-to-understand (TRUE) feedback regarding an athlete’s
start mechanics to coaches for use within a training session on the track.
54

5.2 Introduction
Because of the importance of time in sprint races, quick generation of horizontal impulse during
the first steps out of blocks is needed to accelerate the body towards the finish line.[59] A margin
as small as one-hundredth of a second can decide who wins the race and continues to advance in
competitive athletics. Race analyses indicate that faster athletes win races by quickly increasing
the horizontal velocity of the body’s center of mass (CM) early in the race and maintaining greater
average horizontal CM velocities throughout the race.[60]
During the first foot contact out of the blocks, a sprinter achieves the greatest change in
their horizontal CM velocity during a race by generating large horizontal ground reaction forces
(GRF).[61] An increase in the horizontal CM velocity in the forward direction results when the
anterior-directed horizontal impulse is greater than the posterior-directed impulse during foot contact.[32, 62–64] Previous research on sprint starts involving multievent athletes revealed that positioning the foot more posterior to the CM at initial contact, [65–67] reducing yield at the ankle
to the GRF, and/or increasing thigh angular velocity contributed to quicker changes in horizontal
momentum during the first foot contact out of the blocks.[55] However, positioning the foot too
far behind the CM can reduce the ability to produce the necessary force required to accelerate.[68]
A previous review article on the biomechanics of sprint starts highlighted the need to establish the
link between kinematics at initial contact and GRF features in early acceleration.[56]
In this study, our aim was to build on previous findings to advance our understanding of how
body configuration and impulse regulation during the first foot contact out of the blocks are correlated to changes in total body horizontal momentum and multi-joint control of the lower extremity.
Kinematic and kinetic data were collected during the first foot contact out of blocks performed by
highly-trained sprinters in the context of a typical training session on the track using force plates
and high-speed video. At the whole-body level, we hypothesized that initiating foot contact further behind the CM at initial contact would reduce contact time and increase the average horizontal
force generated. We also hypothesized that control of the lower extremity resulting in less yield at
the ankle and maintenance of a more stable shank segment orientation during impulse generation,
55

would correlate with greater average horizontal forces during the first foot contact.[55] Related
to this, we hypothesized that a faster time to peak thigh angular velocity in post-impact would
correlate with greater average horizontal forces. To better understand how individuals control the
lower extremity during the first step out of the blocks, we studied how the ankle, knee, and hip
contributed to overall control of the stance leg during impulse generation. We expected that individuals who positioned their foot further behind their CM at initial contact would alter the reaction
force orientation relative to the lower extremity segments, thereby altering the relative contributions of the ankle, knee, and hip during impulse regulation. We hypothesized that foot position
relative to CM at initial contact would be correlated with the relative distribution of the net joint
moment (NJM) impulse across the lower extremity during both impact and post-impact.
To advance coach and athlete understanding of sprint mechanics in ways that can facilitate
improvements in sprint performance, we used the same measures in our hypotheses to provide
timely, relevant, usable, and easy-to-understand information (TRUE feedback) to improve impulse
regulation during the first foot contact out of the blocks.[69] Analysis and immediate feedback provided within a training session were expected to advance our understanding of how highly-trained
to world-class sprinters increase their speed during the first foot contact out of the blocks from
full-body and multi-joint coordination perspectives.
5.3 Methods
5.3.1 Participants
Six female (age = 19.8 ± 1.7 years, mass = 63.6 ± 5.9 kg) and nine male (age = 20.4 ± 1.9 years,
mass = 83.0 ± 7.5 kg) highly-trained sprinters volunteered to participate in accordance with the
Institutional Review Board for research involving human subjects at the University of Southern
California. Four of the female athletes were sprinters (100m PR: 11.6 ± 0.7 s), one female was a
hurdler (100 mH PR: 13.15 s), and one female was a long jumper (LJ PR: 6.34 m). Eight male
athletes were sprinters (100 m PR: 10.2 ± 0.2 s), while one male athlete was a hurdler (110mh PR:
56

13.53 s). Athletes were classified as elite (14) and world-class (1) using the framework outlined by
McKay and colleagues.[48]
5.3.2 Experimental Design
A customized system for measuring GRFs and high-speed video of the first foot contact out of
the blocks was developed to advance our understanding of impulse regulation during a sprint start
performed on a track in a way that can provide coaches with TRUE feedback that could be used
within a session.[69, 70] As part of the design phase, we met with the coaches before the data
collection to demonstrate how measured GRFs, when synchronized with high-speed video of the
athlete’s sprint start, could facilitate improvements in sprint start mechanics during a single training session.[6] During an interactive session, the coaches used the technology and explored how
resultant GRF vectors and force-time curves synchronized with high-speed video caused observed
changes in body center of mass velocity during the first foot contact out of the blocks (Figure
5.1).[55] The coaches then determined how best to embed the use of the sprint start force feedback
system within the individual athlete’s training sessions.
5.3.3 Data Collection
Before beginning, athletes went through their regular sprint start warm-up. The athletes familiarized themselves with the experimental conditions until they were comfortable completing a start
on the instrumented runway. Each individual used their preferred block settings and their own
spikes. Athletes then performed maximal effort trials separated by at least 5 minutes of rest. Each
athlete completed 1-3 trials and accelerated through at least the 10-meter mark.
Sagittal and rear plane kinematics (240 fps, Panasonic LUMIX GH5S, Osaka, Japan) and GRFs
(1200 Hz, 0.4 x 0.6m, Kistler 9286B, Amhurst, MA) were measured for the first foot contact out of
the blocks. Kinetic data were collected using two force plates oriented long side together, creating
a 0.6 by 0.8-meter region to accommodate natural foot placement out of the blocks. Friction was
increased by adding Velcro to the feet of the portable force plates to minimize relative motion with
57

Figure 5.1: Exemplar feedback video where resultant ground reaction force (GRF) vectors, forcetime curves, and kinematics were synchronized for a sprinter’s first foot contact out of the blocks.
This feedback was provided to coaches and athletes within 3 minutes of sprint start completion.
(Left) Shows an early instant in the video where the GRF vector is directed posteriorly early in
foot contact. (Right) Demonstrates a later instant in the video of the foot contact when the GRF
vector is directed anteriorly. The blue vertical line indicates the instance in time on the force-time
curve synchronized with the athlete’s corresponding kinematic position. The following measures
were provided to the coaches and athletes; change in horizontal velocity (∆vh), change in vertical
velocity (∆vv), and contact time (s).
the track surface. The portable force plates were covered with a track runway material to accommodate sprinting spikes. In front of and behind the force plates, the same track runway material
was placed on top of a stiff vaulting foam (same thickness as the force plates) used in gymnastics
to provide a raised (0.035 m) and level runway for 10 meters of acceleration. Sagittal video was
synchronized with kinetic data at the time of initial foot contact.
5.3.4 Data Processing
After data collection, body segment landmarks were digitized by the same individual from highspeed video. Markerless motion capture software (VAMA 1.5, Southwest Research Institute) was
58

used, visually inspected, and manually revised for accuracy. The digitized body landmarks included the toe, heel, ankle, knee, hip, shoulder, elbow, wrist, hand, C7, and head apex.18 Digitized
coordinates of the body landmarks in the sagittal plane were filtered with a fourth-order Butterworth filter (zero-phase lag) using an average cut-off frequency (10 Hz) based on a method described by Jackson.[71] Body segment parameters of an athletic population were used to estimate
the total body CM and segment kinematics.[51, 52] The processed segment kinematics data was
synchronized at initial contact and then interpolated to match the sampling rate of the force data.
Net joint moments were calculated using inverse dynamics.[53] The NJM impulse was calculated
as the area under the NJM-time curve of the ankle, knee, and hip during contact.
After the athletes performed three trials, coaches were provided with videos of force-time
curves synchronized with force vector overlays (Figure 5.1).[70] This representation of the forces
in the video provided TRUE feedback so the coaches could discuss ways the individual athlete
might improve their first foot contact. Quantitative information about the change in horizontal
velocity, vertical velocity, and contact time during the first foot contact was provided to compare
results and appreciate differences in the technique used between attempts (Figure 5.1). Using a
custom Python script, GRF was represented by a resultant force vector (vertical, anterior-posterior
components) originating at the center of pressure with a length proportional to the magnitude normalized by the athlete’s body weight. Below the video, the force-time curve was displayed as a
graph with a vertical line that advanced with each video frame. Data visualization of these results
was available within three minutes of a sprinter’s trial.
5.3.5 Analysis
To test our hypotheses, the following variables were calculated using GRF and segment kinematic data. Foot contact was defined as the period of time from initial to final contact when the
vertical force was above 16 N. Ground contact was divided into impact and post-impact for analysis. Impact was defined from initial contact (IC) to the first local minimum of the vertical force.[50]
Post-impact was defined from the first local minimum of the vertical force to final contact (FC).[44]
59

Foot position relative to CM was defined as the horizontal distance in meters from the total body
center of mass (CM) to the foot CM at initial contact. A larger positive distance indicates that
the foot CM was positioned further behind the CM at initial contact in the first foot contact out
of the blocks. The average horizontal force was calculated as the average anterior-posterior component of the GRF throughout the entire foot contact. Peak shank angular velocity during impact
was measured to assess ankle yield during foot contact. A small negative value (smaller magnitude) indicated less yield. The counterclockwise rotation of the shank was reported as a negative
value in degrees per second. Time to peak thigh angular velocity in post-impact was measured
to determine how quickly the thigh rotated over the shank. A smaller time to peak thigh angular
velocity indicated the thigh rotated more quickly over the shank. NJM impulse and the relative
contributions of the ankle, knee, and hip to the total overall lower-extremity NJM moment impulse
were analyzed to address our hypotheses related to multi-joint control. Multi-joint control of the
leg was represented as the relative contribution of the NJM impulse at the ankle, knee, and hip to
the total overall lower-extremity moment impulse during impact, post-impact, and foot contact as
a percentage.
5.3.6 Statistics
Correlations between these variables of interest were used to test our hypotheses and better
understand impulse regulation in the first foot contact. Statistical analysis was conducted using
R and the package Rallfun-v43.[35] Outliers were determined for each variable of interest using
a regression method that removes bad leverage points.[36] To calculate the correlation between
variables of interest, if there were no bad leverage points then we used a correlation method that
accounted for the overall structure of the data without specifying a particular regression estimator.
If there were bad leverage points, then we used a correlation method with a robust regression
estimator that removed bad leverage points.[36] Regression equations for variables of interest were
calculated using a Theil-Sen estimator where all leverage points were removed. A method was used
to verify if there was heteroscedasticity in the data by testing the hypothesis that the error terms
60

in the linear regression model are homoscedastic while checking for outliers in the independent
variable.[72] For the hypothesis related to understanding the multi-joint control at the ankle, knee,
and hip, the median and normalized median absolute deviation (MADN) of the NJM impulse
during impact, post-impact, and foot contact were reported for the group. Magnitudes of the NJM
impulse were reported for each joint, along with the percentage of each joint to the overall sum of
the ankle, knee, and hip NJM impulse.
5.4 Results
5.4.1 Whole Body
Positioning the foot further behind the CM at initial contact (IC) was significantly correlated
with faster contact times and greater average horizontal forces (Figure 5.2). The statistically significant correlation between foot position relative to CM at IC and contact time was -0.88*. We
also found that when the foot was positioned further behind the CM at IC, there was a statistically
significant correlation of 0.82* with greater average horizontal forces (Figure 5.2, Right). For the
overall group, the median horizontal velocity gained during the first foot contact out of blocks was
1.21 m/s (Table 5.1). For the group of athletes with braking impulse, the median loss in CM horizontal velocity was 0.03 m/s. The median loss in braking velocity was 2.3% of the CM horizontal
velocity gained during propulsion (Table 5.1).
Table 5.1: Median (MADN) change in the horizontal center of mass velocity for 15 highly-trained
to world-class sprinters in the first foot contact out of blocks.
Abbreviations: MADN; normalized median absolute deviation, ∆v; change in center of mass
velocity, h: horizontal.
61

Figure 5.2: Foot position relative to the center of mass (CM) at initial contact (IC) was significantly
correlated to both contact time (Left) and average horizontal force (Right) for the first foot contact
out of the blocks for sprinters. Larger positive values indicate that the foot was positioned further
behind the CM. Blue lines indicate the group median for each variable for 15 highly-trained to
world-class sprinters. The purple highlighted dot represents Athlete A’s exemplar trial, where the
foot was positioned further behind the center of mass (CM). The yellow highlighted dot represents
Athlete B’s exemplar trial, where their foot is closer to the CM. The black line indicates the linear
regression. Note: *; Statistically significant.
To illustrate how foot position relative to CM at IC influenced contact time and average horizontal force, exemplar data are provided for two athletes (Figure 5.3). Athlete A positioned their
foot further behind their CM (21.5 cm) at initial contact compared to Athlete B (1.7 cm) (Figure
5.3, Left). When comparing the force-time curves between athletes, one can see how positioning
the foot further behind the CM at IC (by Athlete A) contributes to a reduction in contact time while
also generating a greater average horizontal force compared to Athlete B (Figure 5.3). While both
exemplar athletes generated similar changes in horizontal velocity (Athlete A: 1.28 m/s vs Athlete
B: 1.26 m/s), Athlete A generated this velocity 81 ms faster than Athlete B.
Greater average horizontal forces were significantly correlated with smaller magnitudes of
shank angular velocity in impact and faster times to peak thigh angular velocity in post-impact
(Figure 5.4, Top). The statistically significant correlation between smaller magnitudes of shank
angular velocity (less yield) in impact and greater average horizontal forces was 0.52∗ (Figure 5.4,
Top Left). There appeared to be heteroscedasticity in this correlation (p=-0.04*), and athletes with
62

Figure 5.3: Exemplar athlete trials, which illustrate how foot position relative to the center of
mass (CM) at initial contact (IC) influenced contact time and average horizontal force. Positioning
the foot further behind the CM at initial contact led to faster contact times and a greater average
horizontal force for Athlete A. Athlete kinematics at IC are on top and associated exemplar forcetime curves are shown on the bottom. Black dots along the red skeleton indicate the CM of the
segment, and the large yellow dot indicates the total body center of mass. The gray vertical lines
divide braking and propulsion. Abbrev: ∆vh; Change in horizontal velocity, ∆vv; Change in
vertical velocity, BW; body weight.
smaller magnitudes of shank angular velocity (less shank yield) had more variability in the average horizontal force they produced (Figure 5.4, Top Left). The statistically significant correlation
between faster times to peak thigh angular velocity and greater average horizontal forces was -
0.75* (Figure 5.4, Top Right). When examining the segment coordination pattern of the shank and
thigh, athletes with smaller magnitudes of peak shank angular velocity during impact also reduced
their times to peak thigh angular velocity (Figure 5.4, Bottom). Smaller magnitudes of peak shank
63

angular velocity and faster times to peak thigh angular velocity were achieved when the foot was
positioned further behind the CM as the thigh rotated over the shank (Figure 5.4, Bottom).
Figure 5.4: Average horizontal force generated during the first foot contact out of blocks was
improved when athletes reduced shank yield in impact (Top Left) and reached their peak thigh
angular velocity faster (Top Right). Greater average horizontal forces in the first foot contact
out of blocks in a sprint start were significantly correlated to smaller magnitudes of peak shank
angular velocities (less shank shield) during impact (Top Left, 0.52*) and faster times to peak
thigh angular velocity in post-impact (Top Right, -0.75*). Shank and thigh coordination generally
improved when the foot was positioned further behind the CM (Bottom). This is shown by the blue
dots in the bottom right corner of the plot with smaller peak shank angular velocity magnitudes in
impact and faster times to peak thigh angular velocity in post-impact. Blue lines indicate the group
median for each variable for 15 highly-trained to world-class sprinters. The purple highlighted dot
represents Athlete A’s exemplar trial, where the foot was positioned further behind the center of
mass (CM). The yellow highlighted dot represents Athlete B’s exemplar trial, where their foot is
closer to the CM. Note: *; Statistically significant.
64

5.4.2 Joint Kinetics
Positioning the foot further behind the CM at initial contact was significantly correlated with
greater magnitudes of knee and hip NJM impulses during impact (Correlations- Knee: -0.58*,
Hip: -0.66*, Figure 5.5, Top and Bottom Left). However, there was no statistically significant
correlation between the foot position relative to CM and ankle NJM impulse. In post-impact, there
was no statistically significant correlation between foot position relative to CM and NJM impulse
for the ankle, knee, or hip. The overall group analysis of the distribution of the ankle, knee, and
hip NJM impulse during impact indicated that multi-joint control of the leg was dominated by the
knee and hip (Table 5.2, Figure 5.5, Top Left and Bottom). During impact, the center of pressure
was located closer to the CM of the foot, contributing to lower NJMs at the ankle. However, the
orientation of the shank segment relative to the net joint forces (NJFs) at the ankle and knee led
to the larger NJMs seen at the knee in the impact phase. The NJFs were generally aligned with
the thigh, but the NJM at the hip was driven up to oppose the proximal knee NJM. During postimpact, the ankle became more dominant relative to hip and knee contributions (Table 5.2, Figure
5.5, Top Right). The multi-joint control changed later in contact because the center of pressure
shifted away from the CM of the foot closer to the toe, driving up the NJM at the ankle to counter
the large moment caused by the GRF and ankle NJF. However, the moment at the knee remained
small because the NJFs did not align with the shank and caused a large clockwise moment to
oppose the proximal ankle NJM. The hip moment remained small in post-impact because of the
relatively small knee NJM and alignment of the NJFs with the thigh at the knee and hip joints.
To illustrate how foot position relative to CM at IC influenced multi-joint control, results from
a third exemplar athlete (Athlete C) are provided. Athlete C had a trial where the foot was 6.4 cm
behind their CM, while another trial was 16.7 cm. For Athlete C’s trial, where the foot was further
behind the CM, the shank yield was minimized, and thigh excursion was smaller (Figure 5.6). The
ankle moment during the overall contact was larger for Athlete C’s trial where the foot was further
behind the CM. Athlete C was able to generate an average acceleration of 5.13 m/s2 when the foot
was further behind the CM compared to 4.56 m/s2 when the foot was less behind the CM. These
65

Table 5.2: Net joint moment (NJM) impulses for the ankle, knee, and hip, during impact, and postimpact, and foot contact for 15 highly-trained to world-class sprinters in the first foot contact out
of blocks. Median (Med) and normalized median absolute deviation (MADN) for the magnitude
and percent of overall lower-extremity NJM impulse are reported.
patterns demonstrated within Athlete C were also found across exemplar Athletes A and B, who
had different foot positions relative to their CM at IC.
66

Figure 5.5: Foot position further behind the center of mass (CM) at initial contact (IC) was significantly correlated to greater net joint moment (NJM) impulses at the knee and hip during impact
for the first foot contact out of blocks in a sprint start (Bottom Left). There was no statistically
significant correlation between the foot position relative to the CM at IC for the ankle during impact or for any lower-extremity joint during post-impact. Net joint moment impulses are shown
for the ankle, knee, and hip of each sprint trial during foot contact (bottom right), impact (left top
and bottom), and post-impact (top right). Trials are organized so a foot positioned further behind
the center of mass (CM) at initial contact is on the right. The bottom left is a zoomed-in version of
the NJM impulse at each lower-extremity joint during impact.
5.5 Discussion
In this work examining the first foot contact out of blocks, we found that athletes who positioned their foot further behind their CM were able to increase their horizontal velocity in a shorter
amount of time. Athletes had more successful interactions when they stabilized the shank early in
contact, which enabled the thigh to rotate over more quickly. On a multi-joint control level, when
67

Figure 5.6: Foot position relative to the center of mass (CM) at initial contact (IC) influenced
the multi-joint control of the lower extremity for exemplar Athlete C during the first foot contact
of a sprint start. When the foot was further behind the CM, the shank yield was smaller during
impact, the range of thigh excursion was smaller, and the ankle moment was larger throughout foot
contact. The images show the free body diagrams for the following events for each trial: initial
contact, horizontal force zero-crossing, peak vertical force, peak horizontal force, and final contact.
Blue lines show the direction and magnitude of the net joint forces. The diameter of the circles
indicates the magnitude of the net joint moment, and the color of the circles indicates the direction
at that instant in time.
the foot was positioned further behind the CM, the demand at the knee and hip increased during
impact. The knee and hip demand were greater for the group in impact, while the demand at the ankle became more dominant in post-impact. This study helped to advance our understanding of how
sprinters increase their speed in the first foot contact out of the blocks at a full-body and multi-joint
coordination level. By collecting force and high-speed video in a normal on-track training session,
68

we were also able to demonstrate how biomechanical information can be provided to coaches in a
timely, relevant, usable, and easy-to-understand way to support technique conversations between
coaches and athletes within a training session.[6, 70]
As with other exploratory studies, this work would benefit from an increase in sample size
and the ability to collect data during multiple training sessions with each athlete. While we were
able to provide feedback within three minutes of the athlete’s attempt, we were not able to determine how this biomechanical information contributed to improvements in sprint start performance.
However, we were able to avoid many of the pitfalls identified by Waters by hosting sessions with
the coaches prior to collections with the athletes.[69] This helped us to empower the coaches with
the information before trying to collect the data in an actual training session. The findings of the
correlational analysis from data of highly-trained sprinters, however, provided important insights
into how specific technical modifications in body configuration and control of the lower extremity
can improve regulation of impulse generation during the first foot contact out of blocks.
We found that positioning the foot behind the center of mass was advantageous for completing the whole-body mechanical objectives of generating horizontal velocity quickly in the first
foot contact out of the blocks. Positioning the foot further behind the CM at IC was significantly
correlated with faster contact times and greater average horizontal forces, indicating that athletes
who prepared their segment orientation for contact were able to generate the change in horizontal velocity needed more quickly. This was consistent with findings in previous literature which
have studied the influence of foot position relative to the center of mass.[67] Contact times, average horizontal forces, and changes in horizontal velocity during braking, propulsion, and overall
contact found in this study were comparable to another study that looked at the third foot contact
after block clearance.[64] As found previously for multi-event athletes, positioning the foot further
behind the CM at IC was significantly correlated with greater average horizontal forces and faster
contact times.[55]
The influence of positioning the foot further behind the CM on shank and thigh coordination
also suggests that this body configuration at initial contact may facilitate lower-extremity control
69

during impulse generation. Differences in the control of the lower extremity during contact were
reflected in the significant correlation between less shank yield (reflected by smaller magnitudes
of shank angular velocities in impact) and greater average horizontal force during contact. The potential heteroscedasticity in this data also revealed that individuals with less shank yield in impact
demonstrated more variability in the average horizontal force they produced. These results suggest that there are additional factors that contribute to increasing average horizontal force beyond
shank stability in the impact phase. Athletes who had smaller magnitudes of peak shank angular
velocity during impact also reduced their time to peak thigh angular velocity in post-impact during impulse generation. These athletes also generally had foot positions further behind the CM at
IC. These results inform future investigations and the role of muscle activation prior to contact to
prepare for the impending imposed moments by the GRF orientation relative to the lower extremity segments.[73] This whole-body strategy stabilizes the shank (reduces yield/collapse) in impact
and contributes to the thigh rotating over the shank more quickly in post-impact consistent with
previous literature.[55] These findings held true across the group and across trials within an exemplar Athlete C. In the trial where Athlete C’s foot was further behind the CM, they had less shank
yield, a smaller thigh excursion, and a greater average acceleration (Figure 5.6). These advantages
seemed to be supported by starting with a thigh orientation that was more vertical. Reducing shank
yield and increasing thigh angular velocity contribute to a sprinter’s ability to generate a large increase in horizontal velocity quickly in the first foot contact of a race.
On a multi-joint control level, positioning the foot further behind the CM at IC led to increased
demands on the lower extremity during impact. Positioning the foot further behind the CM at IC
increased the demand, specifically at the knee and hip, during impact. Practically, this means that
athletes need to have the neuromuscular capacity to handle this increased mechanical demand in
impact to gain the other whole-body advantages that come with positioning the foot further behind
the CM at IC. We also found that the distribution of the NJM impulse across the ankle, knee, and
hip varied in impact and post-impact. In impact, the hip and knee primarily control the leg, while
in post-impact, the ankle becomes more dominant. Differences in GRF, segment orientation, and
70

adjacent NJMs contributed to the different control strategies across the ankle, knee, and hip during
impact and post-impact, consistent with previous literature.[57, 58]
By investigating how highly-trained sprinters regulate impulse generation during the first foot
contact out of the blocks during a training session, we have advanced our understanding of how
body configuration at initial contact is significantly correlated with contact time, average horizontal force during foot contact, and lower extremity multi-joint control during impact. At the
whole-body level, positioning the foot further behind the center of mass (CM) at initial contact
was found to be significantly correlated with faster contact times and greater average horizontal
forces. Increases in average horizontal forces during contact were correlated with smaller magnitudes of shank angular velocity during impact and faster times to peak thigh angular velocity
during post-impact. Positioning the foot further behind the CM at initial contact was also significantly correlated with increases in mechanical demand (net joint moments) on the knee and hip
during impact. This work successfully facilitated the implementation of immediate biomechanics
feedback using integrated technology for coaches aiming to improve their athlete’s first foot contact out of the blocks in a sprint start. Timely, relevant, understandable, and easy-to-understand
feedback was provided to coaches about an individual’s sprint start performance within a training
session.
71

Chapter 6
Conclusion
The development of the Integrated Biomechanics Informatics System (IBIS) aimed to standardize data management practices in biomechanics to provide timely and accurate feedback to
coaches, athletes, clinicians, and patients. One application of the IBIS system outlined in Chapter
2 was to use bioinformatics to identify injury risk, improve performance, and facilitate learning for
track and field athletes. In this work, the organized data sets facilitated by the IBIS system were
leveraged to fill essential knowledge gaps for coaches and athletes in steady-state running, long
jump, and the first step out of blocks in a sprint.
In Chapter 3, the IBIS system was used to integrate bioinformatics acquired from competitive collegiate distance runners from multiple institutions to determine if this information could
contribute to the prospective identification of injury risk. Collegiate distance runners experience
lower-extremity injuries at rates as high as 20%. In the cross-institutional, exploratory study conducted, we wanted to determine if steady-state running ground reaction force (GRF) characteristics
could identify collegiate distance runners who would go on later to sustain lower-extremity injuries.
For Non-Injured runners, we established normative boundaries for 10 GRF variables during braking and propulsion. These boundaries were defined as the median +/- the scaled median absolute
deviation (MADN). Two subgroups were studied, With and Without Impact Peaks, to account for
the potential influence of impact peaks on the GRF variables calculated. Prior to injury, we hypothesized that runners that developed an injury after data collection would have more GRF variables outside of the normative boundaries that Non-Injured runners. A rank-based, non-parametric
72

method for comparing two independent groups called Cliff’s method, was used and there was
no statistically significant difference between Injured and Non-Injured runners for the number of
variables outside of the boundaries (p=0.17). However, for the Without Impact peak group, the
Injured runners have more variables outside of the normative boundaries that Non-Injured runners
(p¡0.001). This new approach demonstrates that it may be possible to use GRF characteristics to
identify collegiate distance runners Without Impact peaks who may be at risk prior to injury.
In Chapter 4, the IBIS system was used to analyze kinetic and kinematic long jump data
acquired over 8 years to determine if whole-body and multi-joint control in popoffs replicate an
individual’s full jump. Long jumpers maximize their horizontal jump distance from the scratch line
by using their individual capabilities to regulate impulse generation during the last contact prior
to flight known as takeoff. Popoffs are a task performed by jumpers in training and are believed
to replicate a full jump’s run-jump transition without landing in the sand. Ground reaction forces
(GRFs) and segment kinematics were measured in training to determine differences in impulse
regulation and lower extremity control between full jumps and popoff takeoffs for 15 elite competitive jumpers. No significant differences at the whole-body level were found in the change in
center of mass (CM) velocity (horizontal or vertical) for the group between full jumps and popoffs.
Greater leg angles at initial contact for both tasks were related to larger losses in horizontal CM
velocity and greater gains in vertical CM velocity during takeoff. At the multi-joint control level,
no significant differences in net joint moment impulses at the ankle, knee, or hip during impact,
post-impact, and foot contact were found for magnitude or percent distribution. This study suggest
that for these jumpers, popoffs impose lower extremity control and impulse regulation challenges
that are similar to full jumps.
Finally, in Chapter 5, the IBIS system facilitated the ability to provide coaches and athletes
with biomechanical feedback about their sprint start within one practice session. For sprinters, it
is important to generate horizontal impulse quickly in the initial steps out of the blocks to accelerate toward the finish line. In this work, we aimed to determine how impulse regulation and body
segment configuration at initial foot contact in the first foot contact out of blocks by highly-trained
73

sprinters was correlated to total body changes in horizontal momentum and multi-joint control.
Segment kinematics and ground reaction forces (GRFs) were measured during the first foot contact out of blocks showed that positioning the foot further behind the center of mass (CM) at initial
contact correlated significantly with greater average horizontal forces, shorter contact times and
more mechanical demand on the knee and hip during impact. Smaller shank angular velocity
magnitudes and faster times to peak thigh angular velocity during post-impact were significantly
correlated with increases in average horizontal forces during foot contact. These biomechanical measures helped provide feedback that was timely, relevant, usable, and easy-to-understand
(TRUE) about an athlete’s start mechanics to coaches within a training session.
Building off of these exploratory studies, there are many opportunities for expanding upon
these findings in future studies. In Chapter 3, the IBIS system facilitated the ability to increase
the sample size of participants by working across institutions. Future studies will be able to determine if the methods outlined in Chapter 3 are able to identify runners who go on to experience
lower extremity injuries. Chapter 4 demonstrated how the IBIS system enabled longitudinal data
to be collected for athletes over many years. Additionally, other training drills could be studied
to determine how various drills do or do not successfully prepare athletes for different aspects of
competition. Building off the work started in Chapter 5, it would be helpful for athletes to receive
feedback and have a chance to do additional attempts within a practice session. This will give them
the opportunity to see how various adjustments they make to their mechanics will influence their
performance. Overall, continuing the work started in these exploratory studies, it would be helpful
to continue investigating patterns within individuals rather than assuming group results represent
everyone’s behavior. The IBIS system demonstrated that the use of best practices for standardizing
biomechanics data using databases and remote servers accessible in multiple locations will enable
difficult questions that require large datasets to be answered.
Together this body of work demonstrates how the IBIS system helped facilitate biomechanical studies to better understand musculoskeletal injuries in collegiate long distance runners, how
74

training drills done in practice prepare long jumpers for competition, and enable technique discussions during a single practice session for sprinters. Adopting best practices from the medical
imaging community to create the IBIS system allowed us to advance what could normally be accomplished by a single biomechanics lab by working across institutions, studying athletes over
time, and providing integrated feedback to improve performance in practice.
75

References
1. Wilkinson, M. D., Dumontier, M., Aalbersberg, I. J., Appleton, G., Axton, M., Baak, A., et al.
The FAIR Guiding Principles for scientific data management and stewardship. en. Scientific
Data 3, 160018. doi:10.1038/sdata.2016.18 (2016).
2. Torres-Esp´ın, A., Almeida, C. A., Chou, A., Huie, J. R., Chiu, M., Vavrek, R., et al. Promoting
FAIR Data Through Community-driven Agile Design: the Open Data Commons for Spinal
Cord Injury (odc-sci.org). en. Neuroinformatics 20, 203–219. doi:10.1007/s12021-021-
09533-8 (2022).
3. Liu, J., Wang, X., Verma, S. K., McNitt-Gray, J. & Liu, B. J. The development of an imaging
informatics-based platform and data viewer interface to support sports performance and
injury prevention in track and field athletes in Medical Imaging 2019: Imaging Informatics
for Healthcare, Research, and Applications (eds Bak, P. R. & Chen, P.-H.) (SPIE, San Diego,
United States, 2019), 3. doi:10.1117/12.2513774.
4. Liu, J., Stewart, H., Wiens, C., Mcnitt-Gray, J. & Liu, B. Development of an integrated biomechanics informatics system with knowledge discovery and decision support tools for research
of injury prevention and performance enhancement. en. Computers in Biology and Medicine,
105062. doi:10.1016/j.compbiomed.2021.105062 (2021).
5. McNitt-Gray, J. L. 2016 American Society of Biomechanics Hay Award Lecture: Preparing for Impact—Biomechanical Inspirations From Olympic Sports. en. Journal of Applied
Biomechanics 34, 249–257. doi:10.1123/jab.2017-0270 (2018).
6. McNitt-Gray, J. L., Sand, K., Ramos, C., Peterson, T., Held, L. & Brown, K. Using technology and engineering to facilitate skill acquisition and improvements in performance. en.
Proceedings of the Institution of Mechanical Engineers, Part P: Journal of Sports Engineering and Technology 229, 103–115. doi:10.1177/1754337114565381 (2015).
7. Meur, Y. L. & Torres-Ronda, L. 10 Challenges Facing Today’s Applied Sport Scientist. en, 7
(2019).
8. Vincent, J., Stergiou, P. & Katz, L. The Role of Databases in Sport Science: Current Practice
and Future Potential. en. International Journal of Computer Science in Sport 8, 18.
76

9. Codd, E. F. A relational model of data for large shared data banks. en. Communications of
the ACM 26, 64–69. doi:10.1145/357980.358007 (1983).
10. Macera, C. A., Pate, R. R., Powell, K. E., Jackson, K. L., Kendrick, J. S. & Craven, T. E. Predicting lower-extremity injuries among habitual runners. eng. Archives of Internal Medicine
149, 2565–2568 (1989).
11. Marti, B., Vader, J. P., Minder, C. E. & Abelin, T. On the epidemiology of running injuries:
The 1984 Bern Grand-Prix study. en. The American Journal of Sports Medicine 16, 285–294.
doi:10.1177/036354658801600316 (1988).
12. Van Gent, R. N., Siem, D., van Middelkoop, M., van Os, A. G., Bierma-Zeinstra, S. M. A.,
Koes, B. W., et al. Incidence and determinants of lower extremity running injuries in long
distance runners: a systematic review. en. British Journal of Sports Medicine 41, 469–480.
doi:10.1136/bjsm.2006.033548 (2007).
13. Bennell, K. L. & Brukner, P. D. Epidemiology and site specificity of stress fractures. en.
Clinics in Sports Medicine 16, 179–196. doi:10.1016/S0278-5919(05)70016-8 (1997).
14. Bennell, K. L., Malcolm, S. A., Thomas, S. A., Wark, J. D. & Brukner, P. D. The incidence
and distribution of stress fractures in competitive track and field athletes: a twelve-month
prospective study. en. The American Journal of Sports Medicine 24, 211–217. doi:10.1177/
036354659602400217 (1996).
15. Snyder, R. A., Koester, M. C. & Dunn, W. R. Epidemiology of stress fractures. en. Clinics in
Sports Medicine 25, 37–52. doi:10.1016/j.csm.2005.08.005 (2006).
16. Koenig, S. J., Toth, A. P. & Bosco, J. A. Stress fractures and stress reactions of the diaphyseal
femur in collegiate athletes: an analysis of 25 cases. en. American Journal of Orthopaedics,
5.
17. Nusselt, T., Klinger, H.-M., Schultz, W. & Baums, M. H. Fatigue stress fractures of the pelvis
: A rare cause of low back pain in female athletes. en. Acta Orthopaedica Belgica 76, 7
(2010).
18. Hameed, F. & McInnis, K. C. Sacral stress fracture causing radiculopathy in a female runner:
a case report. en. PM&R 3, 489–491. doi:10.1016/j.pmrj.2010.10.009 (2011).
19. Quarmby, A., Monnig, J., Mugele, H., Henschke, J., Kim, M., Cassel, M., ¨ et al. Biomechanics
and lower limb function are altered in athletes and runners with achilles tendinopathy compared with healthy controls: A systematic review. en. Frontiers in Sports and Active Living
4, 1012471. doi:10.3389/fspor.2022.1012471 (2023).
20. Lopes, A. D., Junior, L. C. H., Yeung, S. S. & Costa, L. O. P. What are the main runningrelated musculoskeletal injuries? en. Sports Med (2012).
77

21. Kuhman, D. J., Paquette, M. R., Peel, S. A. & Melcher, D. A. Comparison of ankle kinematics and ground reaction forces between prospectively injured and uninjured collegiate cross
country runners. en. Human Movement Science 47, 9–15. doi:10.1016/j.humov.2016.01.
013 (2016).
22. Dudley, R. I., Pamukoff, D. N., Lynn, S. K., Kersey, R. D. & Noffal, G. J. A prospective
comparison of lower extremity kinematics and kinetics between injured and non-injured collegiate cross country runners. en. Human Movement Science 52, 197–202. doi:10.1016/j.
humov.2017.02.007 (2017).
23. Kliethermes, S. A., Stiffler-Joachim, M. R., Wille, C. M., Sanfilippo, J. L., Zavala, P. &
Heiderscheit, B. C. Lower step rate is associated with a higher risk of bone stress injury: a
prospective study of collegiate cross country runners. eng. British Journal of Sports Medicine.
doi:10.1136/bjsports-2020-103833 (2021).
24. Davis, I. S., Bowser, B. J. & Mullineaux, D. R. Greater vertical impact loading in female
runners with medically diagnosed injuries: a prospective investigation. en. British Journal of
Sports Medicine 50, 887–892. doi:10.1136/bjsports-2015-094579 (2016).
25. Schmida, E. A., Wille, C. M., Stiffler-Joachim, M. R., Kliethermes, S. A. & Heiderscheit,
B. C. Vertical loading rate is not associated with running injury, regardless of calculation
method. en. Medicine & Science in Sports & Exercise Publish Ahead of Print. doi:10 .
1249/MSS.0000000000002917 (2022).
26. Messier, S. P., Martin, D. F., Mihalko, S. L., Ip, E., DeVita, P., Cannon, D. W., et al. A 2-
year prospective cohort study of overuse running injuries: the runners and injury longitudinal
study (trails). en. The American Journal of Sports Medicine 46, 2211–2221. doi:10.1177/
0363546518773755 (2018).
27. Davis, J. J. & Gruber, A. H. Leg stiffness, joint stiffness, and running-related injury: evidence
from a prospective cohort study. en. Orthopaedic Journal of Sports Medicine 9. Publisher:
SAGE Publications Inc, 23259671211011213. doi:10.1177/23259671211011213 (2021).
28. Troy, K. L., Davis, I. S. & Tenforde, A. S. A narrative review of metatarsal bone stress injury
in athletic populations: etiology, biomechanics, and management. en. PM&R 13, 1281–1290.
doi:10.1002/pmrj.12518 (2021).
29. Winters, M., Burr, D. B., van der Hoeven, H., Condon, K. W., Bellemans, J. & Moen, M. H.
Microcrack-associated bone remodeling is rarely observed in biopsies from athletes with
medial tibial stress syndrome. en. Journal of Bone and Mineral Metabolism 37, 496–502.
doi:10.1007/s00774-018-0945-9 (2019).
30. Schilcher, J., Bernhardsson, M. & Aspenberg, P. Chronic anterior tibial stress fractures in
athletes: No crack but intense remodeling. en. Scandinavian Journal of Medicine & Science
in Sports 29, 1521–1528. doi:10.1111/sms.13466 (2019).
78

31. Gruber, A. H. The “impacts cause injury” hypothesis: Running in circles or making new
strides? en. Journal of Biomechanics 156, 111694. doi:10 . 1016 / j . jbiomech . 2023 .
111694 (2023).
32. Munro, C. F. & Miller, Doris I. Ground reaction forces in running: a reexamination. en.
Journal of Biomechanics 20, 147–155. doi:10.1016/0021-9290(87)90306-x (1987).
33. Samaan, C. D., Rainbow, M. J. & Davis, I. S. Reduction in ground reaction force variables with instructed barefoot running. en. Journal of Sport and Health Science 3, 143–151.
doi:10.1016/j.jshs.2014.03.006 (2014).
34. Mair, P. & Wilcox, R. Robust statistical methods in R using the WRS2 package. en. Behavior
Research Methods 52, 464–488. doi:10.3758/s13428-019-01246-w (2020).
35. Wilcox, R. Modern Statistics for the Social and Behavioral Sciences 2nd. en (Chapman &
Hall, 2017).
36. Wilcox, R. R. A guide to robust statistical methods eng. OCLC: 1407112136 (Springer,
Cham, 2023).
37. Evans, R. J., Moffit, T. J., Mitchell, P. K. & Pamukoff, D. N. Injury and performance related biomechanical differences between recreational and collegiate runners. en. Frontiers in
Sports and Active Living 5, 1268292. doi:10.3389/fspor.2023.1268292 (2023).
38. Hay, J. G. The biomechanics of the long jump. eng. Exercise and Sport Sciences Reviews 14,
401–446 (1986).
39. Mathiyakom, W., McNitt-Gray, J. & Wilcox, R. Lower extremity control and dynamics during backward angular impulse generation in backward translating tasks. en. Experimental
Brain Research 169, 377–388. doi:10.1007/s00221-005-0150-7 (2006).
40. Bobbert, M. F., Gerritsen, K. G. M., Litjens, M. C. A. & Van Soest, A. J. Why is countermovement jump height greater than squat jump height?: en. Medicine & Science in
Sports & Exercise 28, 1402–1412. doi:10.1097/00005768-199611000-00009 (1996).
41. Yu, B. & Hay, J. G. Optimum phase ratio in the triple jump. en (1996).
42. Liu, H. & Yu, B. Effects of phase ratio and velocity conversion coefficient on the performance
of the triple jump. en. Journal of Sports Sciences 30, 1529–1536. doi:10.1080/02640414.
2012.713502 (2012).
43. Christopher Daniel Ramos. Regulation of linear and angular impulse generation: implications for athletic performance PhD thesis (University of Southern California, 2017).
79

44. Ramos, C. D., Ramey, M., Wilcox, R. R. & McNitt-Gray, J. L. Generation of linear impulse
during the takeoff of the long jump. en. Journal of Applied Biomechanics 35, 52–60. doi:10.
1123/jab.2017-0249 (2019).
45. Bedi, J. & Cooper, J. Take off in the long jump—Angular momentum considerations. en.
Journal of Biomechanics 10, 541–548. doi:10.1016/0021-9290(77)90034-3 (1977).
46. Mathiyakom, W., McNitt-Gray, J. & Wilcox, R. Lower extremity control and dynamics during backward angular impulse generation in forward translating tasks. en. Journal of Biomechanics 39, 990–1000. doi:10.1016/j.jbiomech.2005.02.022 (2006).
47. Held, L. A., Flashner, H. & McNitt-Gray, J. L. en. in Handbook of Human Motion (eds
Muller, B., Wolf, S. I., Brueggemann, G.-P., Deng, Z., McIntosh, A., Miller, F., ¨ et al.) 1–31
(Springer International Publishing, Cham, 2017). doi:10.1007/978-3-319-30808-1_127-
1.
48. McKay, A. K., Stellingwerff, T., Smith, E. S., Martin, D. T., Mujika, I., Goosey-Tolfrey, V. L.,
et al. Defining training and performance caliber: a participant classification framework. en.
International Journal of Sports Physiology and Performance 17, 317–331. doi:10 . 1123 /
ijspp.2021-0451 (2022).
49. Ramos, C., Ramey, M. R. & McNitt-Gray, J. L. en. in Biomechanical Principles and Applications in Sports (eds Pallis, J. M., McNitt-Gray, J. L. & Hung, G. K.) 17–42 (Springer
International Publishing, Cham, 2019). doi:10.1007/978-3-030-13467-9_2.
50. Ramey, M. R. Force relationships of the running long jump. eng. Medicine and Science in
Sports 2, 146–151 (1970).
51. De Leva, P. Adjustments to Zatsiorsky-Seluyanov’s segment inertia parameters. en. Journal
of Biomechanics 29, 1223–1230. doi:10.1016/0021-9290(95)00178-6 (1996).
52. Cherny˘ı, G. G. & Regirer, S. A. Contemporary problems of biomechanics eng. OCLC: 21972338
(Mir Publishers ; CRC Press, Moscow, Boca Raton, 1990).
53. Winter, D. A. Biomechanics and motor control of human movement 4th ed. OCLC: ocn318408191
(Wiley, Hoboken, N.J, 2009).
54. Grabowski, A. M., McGowan, C. P., McDermott, W. J., Beale, M. T., Kram, R. & Herr, H. M.
Running-specific prostheses limit ground-force during sprinting. en. Biology Letters 6, 201–
204. doi:10.1098/rsbl.2009.0729 (2010).
55. Kathleen Elizabeth Costa. Control and dynamics during horizontal impulse generation PhD
thesis (University of Southern California, 2004).
80

56. Bezodis, N. E., Willwacher, S. & Salo, A. I. T. The Biomechanics of the Track and Field
Sprint Start: A Narrative Review. en. Sports Medicine 49, 1345–1364. doi:10.1007/s40279-
019-01138-1 (2019).
57. Mathiyakom, W., Wilcox, R. & McNitt-Gray, J. L. Generation of forward angular impulse
with different initial conditions. en. Human Movement Science 87, 103035. doi:10.1016/j.
humov.2022.103035 (2023).
58. Mathiyakom, W., McNitt-Gray, J., Requejo, P. & Costa, K. Modifying center of mass trajectory during sit-to-stand tasks redistributes the mechanical demand across the lower extremity
joints. en. Clinical Biomechanics 20, 105–111. doi:10.1016/j.clinbiomech.2004.08.
005 (2005).
59. Hay, J. G. The biomechanics of sports techniques 4th ed (Prentice-Hall, Englewood Cliffs,
N.J, 1993).
60. Mann, R. & Herman, J. Kinematic Analysis of Olympic Sprint Performance: Men’s 200
Meters. International Journal of Sport Biomechanics 1, 151–162. doi:10.1123/ijsb.1.2.
151 (1985).
61. Colyer, S. L., Nagahara, R. & Salo, A. I. T. Kinetic demands of sprinting shift across the
acceleration phase: Novel analysis of entire force waveforms. en. Scandinavian Journal of
Medicine & Science in Sports 28, 1784–1792. doi:10.1111/sms.13093 (2018).
62. Haugen, T. A., Breitschadel, F. & Seiler, S. Sprint mechanical variables in elite athletes: Are ¨
force-velocity profiles sport specific or individual? en. PLOS ONE 14 (ed Peyre-Tartaruga, ´
L. A.) e0215551. doi:10.1371/journal.pone.0215551 (2019).
63. Clark, K. P. & Weyand, P. G. Sprint running research speeds up: A first look at the mechanics
of elite acceleration. en. Scandinavian Journal of Medicine & Science in Sports 25, 581–582.
doi:10.1111/sms.12520 (2015).
64. Von Lieres Und Wilkau, H. C., Bezodis, N. E., Morin, J.-B., Irwin, G., Simpson, S. & Bezodis, I. N. The importance of duration and magnitude of force application to sprint performance during the initial acceleration, transition and maximal velocity phases. en. Journal of
Sports Sciences 38, 2359–2366. doi:10.1080/02640414.2020.1785193 (2020).
65. Ridderikhoff, A., Batelaan, J. H. & Bobbert, M. F. Jumping for distance: control of the external force in squat jumps: en. Medicine & Science in Sports & Exercise 31, 1196–1204.
doi:10.1097/00005768-199908000-00018 (1999).
66. Bobbert, M. F. & Van Zandwijk, J. P. Dynamics of force and muscle stimulation in human
vertical jumping: en. Medicine & Science in Sports & Exercise 31, 303–310. doi:10.1097/
00005768-199902000-00015 (1999).
81

67. Mero, A., Komi, P. & Gregor, R. Biomechanics of Sprint Running: A Review. en. Sports
Medicine 13, 376–392. doi:10.2165/00007256-199213060-00002 (1992).
68. Bezodis, N. E., Trewartha, G. & Salo, A. I. T. Understanding the effect of touchdown distance
and ankle joint kinematics on sprint acceleration performance through computer simulation.
en. Sports Biomechanics 14, 232–245. doi:10.1080/14763141.2015.1052748 (2015).
69. Waters, A., Phillips, E., Panchuk, D. & Dawson, A. The coach–scientist relationship in highperformance sport: Biomechanics and sprint coaches. en. International Journal of Sports
Science & Coaching 14, 617–628. doi:10.1177/1747954119859100 (2019).
70. McNitt-Gray, J. L., Held, L., Mathiyakom, W., Peterson, T. J. & Zaferiou, A. M. Make an
impact: going full circle together. en. Sports Biomechanics, 1–20. doi:10.1080/14763141.
2024.2369206 (2024).
71. Jackson, K. M. Fitting of Mathematical Functions to Biomechanical Data. IEEE Transactions
on Biomedical Engineering BME-26, 122–124. doi:10.1109/TBME.1979.326551 (1979).
72. Lyon, J. D. & Tsai, C.-L. A Comparison of Tests for Heteroscedasticity. The Statistician 45,
337. doi:10.2307/2988471 (1996).
73. McNitt-Gray, J., Hester, D., Mathiyakom, W. & Munkasy, B. Mechanical demand and multijoint control during landing depend on orientation of the body segments relative to the
reaction force. en. Journal of Biomechanics 34, 1471–1482. doi:10.1016/S0021-9290(01)
00110-5 (2001).
82

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Development of an integrated biomechanics informatics system (IBIS) with knowledge discovery and decision support tools based on imaging informatics methodology

PDF

A medical imaging informatics based human performance analytics system

PDF

Lower extremity control and dynamics during the golf swing

PDF

Regulation of linear and angular impulse generation: implications for athletic performance

PDF

An electronic patient record (ePR) system for image-assisted minimally invasive spinal surgery

PDF

Knowledge‐driven decision support for assessing radiation therapy dose constraints

PDF

Imaging informatics-based electronic patient record and analysis system for multiple sclerosis research, treatment, and disease tracking

PDF

Upper extremity control and dynamics during manual wheelchair propulsion at different speeds and wheelchair configurations

PDF

Using biomechanics to understand how an individual accomplishes a task in a variety of contexts

PDF

Understanding reactive balance control strategies in non-disabled and post-stroke gait

PDF

Investigation of preclinical testing methods for total ankle replacements

PDF

Modeling human regulation of momentum while interacting with the environment

PDF

Mining an ePR system using a treatment plan navigator for radiation toxicity to evaluate proton therapy treatment protocol for prostate cancer

PDF

Using nonlinear feedback control to model human landing mechanics

PDF

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

PDF

Decision support system in radiation therapy treatment planning

PDF

Modeling and vibration analysis of wheelchair-users

PDF

Bio-inspired tendon-driven systems: computational analysis, optimization, and hardware implementation

PDF

Pathomechanics of femoroacetabular impingement syndrome: utilizing subject-specific modeling approaches to investigate the influence of hip joint morphology and neuromuscular control

PDF

Integration of digital twin and generative models in model-based systems upgrade methodology

Asset Metadata

Creator Stewart, Harper Elizabeth (author)

Core Title Identifying injury risk, improving performance, and facilitating learning using an integrated biomechanics informatics system (IBIS)

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Biomedical Engineering

Degree Conferral Date 2024-08

Publication Date 08/19/2024

Defense Date 08/15/2024

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

Tag Athletics,biomechanics,impulse regulation,multi-joint control,Track and field

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor McNitt-Gray, Jill L. (committee chair), Finley, James (committee member), Khoo, Michael (committee member), Liu, Brent (committee member), Wilcox, Rand R. (committee member)

Creator Email harperestewart7@gmail.com,hestewar@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC1139991ZM

Unique identifier UC1139991ZM

Identifier etd-StewartHar-13402.pdf (filename)

Legacy Identifier etd-StewartHar-13402

Document Type Dissertation

Format theses (aat)

Rights Stewart, Harper Elizabeth

Internet Media Type application/pdf

Type texts

Source 20240819-usctheses-batch-1199 (batch), 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 author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

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

Repository Email cisadmin@lib.usc.edu