bioengineering

Article
Biomechanical Gait Analysis Using a Smartphone-Based
Motion Capture System (OpenCap) in Patients with
Neurological Disorders
Yu-Sun Min 1,2,3,4 , Tae-Du Jung 1,2 , Yang-Soo Lee 1,5 , Yonghan Kwon 5 , Hyung Joon Kim 2 , Hee Chan Kim 4,6 ,
Jung Chan Lee 4,7,8 and Eunhee Park 1,2,3, *

1 Department of Rehabilitation Medicine, School of Medicine, Kyungpook National University,

Daegu 41944, Republic of Korea; [email protected] (Y.-S.M.); [email protected] (T.-D.J.);
[email protected] (Y.-S.L.)
2 Department of Rehabilitation Medicine, Kyungpook National University Chilgok Hospital,
Daegu 41404, Republic of Korea; [email protected]
3 AI-Driven Convergence Software Education Research Program, Graduate School of Computer Science and
Engineering, Kyungpook National University, Daegu 41566, Republic of Korea
4 Department of Biomedical Engineering, College of Medicine, Seoul National University,
Seoul 03080, Republic of Korea; [email protected] (H.C.K.); [email protected] (J.C.L.)
5 Department of Rehabilitation Medicine, Kyungpook National University Hospital,
Daegu 41944, Republic of Korea; [email protected]
6 Interdisciplinary Program in Bioengineering, Graduate School, Seoul National University,
Seoul 08826, Republic of Korea
7 Institute of Bioengineering, Seoul National University, Seoul 03080, Republic of Korea
8 Institute of Medical and Biological Engineering, Medical Research Center, Seoul National University,
Seoul 03080, Republic of Korea
* Correspondence: [email protected]; Tel.: +82-53-200-3867

Abstract: This study evaluates the utility of OpenCap (v0.3), a smartphone-based motion capture
system, for performing gait analysis in patients with neurological disorders. We compared kinematic
and kinetic gait parameters between 10 healthy controls and 10 patients with neurological conditions,
Citation: Min, Y.-S.; Jung, T.-D.; Lee, including stroke, Parkinson’s disease, and cerebral palsy. OpenCap captured 3D movement dynamics
Y.-S.; Kwon, Y.; Kim, H.J.; Kim, H.C.;
using two smartphones, with data processed through musculoskeletal modeling. The key findings
Lee, J.C.; Park, E. Biomechanical Gait
indicate that the patient group exhibited significantly slower gait speeds (0.67 m/s vs. 1.10 m/s,
Analysis Using a Smartphone-Based
p = 0.002), shorter stride lengths (0.81 m vs. 1.29 m, p = 0.001), and greater step length asymmetry
Motion Capture System (OpenCap) in
(107.43% vs. 91.23%, p = 0.023) compared to the controls. Joint kinematic analysis revealed increased
Patients with Neurological Disorders.
Bioengineering 2024, 11, 911.
variability in pelvic tilt, hip flexion, knee extension, and ankle dorsiflexion throughout the gait cycle in
https://doi.org/10.3390/ patients, indicating impaired motor control and compensatory strategies. These results indicate that
bioengineering11090911 OpenCap can effectively identify significant gait differences, which may serve as valuable biomarkers
for neurological disorders, thereby enhancing its utility in clinical settings where traditional motion
Academic Editor: Redha Taiar
capture systems are impractical. OpenCap has the potential to improve access to biomechanical
Received: 23 August 2024 assessments, thereby enabling better monitoring of gait abnormalities and informing therapeutic
Revised: 9 September 2024 interventions for individuals with neurological disorders.
Accepted: 9 September 2024
Published: 12 September 2024 Keywords: gait; kinematics; kinetics; motion capture; smartphone

Copyright: © 2024 by the authors.

1. Introduction
Licensee MDPI, Basel, Switzerland.
This article is an open access article Pediatric gait analysis is a critical tool for assessing motor function, particularly in
distributed under the terms and children with neurological disorders such as cerebral palsy [1]. Accurate gait analysis is
conditions of the Creative Commons essential for designing appropriate therapeutic interventions, such as botulinum toxin
Attribution (CC BY) license (https:// injections and robotic-assisted gait training [2]. However, conducting traditional gait
creativecommons.org/licenses/by/ analysis in pediatric populations presents significant challenges [3]. Children, especially
4.0/). those with neurological impairments, often struggle with cooperation and adherence to

Bioengineering 2024, 11, 911. https://doi.org/10.3390/bioengineering11090911 https://www.mdpi.com/journal/bioengineering

Bioengineering 2024, 11, 911 2 of 18

the structured protocols required for accurate data collection. For instance, young children
may find it difficult to follow instructions during gait assessment, leading to inconsistent
and unreliable data. Moreover, the necessity for assistive devices like orthoses further
complicates the process, as these devices can interfere with the natural gait pattern and
thus affect the validity of the analysis.
These challenges underline the pressing need for simpler, more adaptable gait analysis
tools that can accommodate the unique needs of pediatric populations. The traditional gait
analysis setup, typically involving a gait laboratory equipped with multiple cameras, force
plates, and motion capture systems, is not only resource-intensive but also requires sub-
stantial technical expertise and patient compliance [4]. As a result, there has been growing
interest in developing alternative methods that are less burdensome and more accessible
for use in clinical settings. Recent technological advancements have introduced innovative
tools like smartphone-based motion capture systems, which offer a more practical solution
without compromising on accuracy [5]. These systems, such as the OpenCap software,
leverage the widespread availability and user-friendly nature of smartphones to perform
3D motion analysis and musculoskeletal force estimations with minimal setup [6]. This
approach not only reduces the technical and logistical barriers associated with traditional
gait analysis but also enhances the feasibility of conducting assessments in diverse clinical
environments, including settings where full-scale gait labs are not available. Furthermore,
the use of these accessible technologies can facilitate more frequent and comprehensive
assessments, allowing clinicians to monitor patient progress more effectively and adjust
treatment plans in a timely manner. As these tools continue to evolve, they hold the poten-
tial to revolutionize gait analysis by making it more scalable and adaptable to a broader
range of clinical applications.
The development of such tools represents a significant step forward in pediatric gait
analysis, particularly for children with cerebral palsy and other neurological conditions
who may benefit from more frequent and accessible evaluations. For instance, studies have
shown that regular gait assessments in children with cerebral palsy can lead to more timely
interventions, improving mobility and overall quality of life [3,7,8]. Similarly, in stroke
and Parkinson’s disease patients, frequent gait analysis using accessible tools has been
shown to help in the early detection of gait deterioration, enabling prompt adjustments in
therapeutic strategies [7–11]. These advancements not only improve individualized care
but also support large-scale studies that can further our understanding of gait abnormalities
across different neurological conditions. As gait analysis plays a vital role in tracking the
progress of therapeutic interventions, the adoption of simpler, yet reliable, methods is
crucial for improving patient outcomes. Moreover, these innovations could facilitate large-
scale studies and clinical trials by making gait analysis more accessible and scalable, thus
contributing to the broader field of pediatric rehabilitation and biomechanics.
Quantitative gait analysis has historically been utilized primarily within limited, small-
scale groups due to the high costs, need for specialized personnel, and the significant time
and resource requirements involved [10]. This type of analysis demands the installation and
operation of expensive equipment, such as 3D motion capture systems, force plates, and
electromyography devices, all of which require skilled professionals trained in their use.
Additionally, the data collection and analysis processes are complex and time-consuming,
posing substantial economic challenges for large-scale studies or clinical applications. These
factors have hindered the widespread adoption of gait analysis, resulting in its predominant
use within specific research groups or advanced medical institutions. Recent advances in
computer vision have made it possible to perform quantitative movement analysis using
only digital videos from low-cost devices like smartphones and tablets. These technologies,
such as OpenPose, automatically identify keypoints on the human body from simple
videos, significantly improving the accessibility of movement assessment [5,12,13]. This
has led to the development of various workflows for measuring gait parameters and
joint kinematics, demonstrating the potential for widespread application in health and
performance analysis. Building on these advancements, the development of user-friendly
Bioengineering 2024, 11, 911 3 of 18

software has further enhanced the accessibility of these analytical tools, paving the way for
broader dissemination and utilization in clinical and research settings in the future.
Recently, OpenCap, an open-source, web-based software, has been developed to
enable the estimation of human 3D kinematics and dynamics from videos captured using
two or more smartphones [6]. This software is freely available and leverages decades
of advancements in computer vision and musculoskeletal simulation, allowing for the
analysis of movement dynamics without the need for specialized hardware, software, or
expertise. However, while OpenCap has been validated on musculoskeletal models of
healthy individuals, its application to patients with clinical conditions has yet to be explored.
To address this gap, we have designed a study to investigate whether this accessible and
user-friendly technology can be effectively utilized in clinical settings with actual patients,
thus extending its utility beyond the laboratory and into practical healthcare environments.
This study advances the field of gait analysis by applying the OpenCap smartphone-
based motion capture system in clinical settings to assess patients with neurological disor-
ders such as stroke, Parkinson’s disease, and cerebral palsy. We demonstrate the system’s
capability to capture significant kinematic and kinetic differences between healthy con-
trols and patients, showcasing its effectiveness in clinical environments. Additionally, this
research highlights the potential of low-cost, accessible technology to perform complex
gait analyses, thereby expanding the feasibility of biomechanical assessments in diverse
settings and supporting more frequent and comprehensive monitoring of gait abnormali-
ties in patients.

2. Materials and Methods

2.1. Participants and Sample Size Estimation
The experiment involved two groups: a patient group comprising 10 individuals with
neurological disorders, including stroke, Parkinson’s disease, cerebral palsy, and others,
and a control group of 10 healthy adults. The sample size was determined based on the
prior literature and statistical power considerations, aiming to achieve 80% power with
a large, anticipated effect size (d = 0.8). Functional Data Analysis (FDA) was conducted,
and bootstrap simulations were used to estimate confidence intervals for key outcome
measures [14–16]. In some instances, the confidence intervals between the patient and
control groups did not overlap, indicating that the sample size was sufficient to detect
statistically significant differences. These results suggest that the current sample size was
adequate for identifying substantial group differences, validating the use of FDA and
bootstrap methods. All participants were fully informed of the study’s procedures and
provided written consent prior to data collection. The study was approved by the ethics
review committee of KNUCH (Protocol Code: 2023-09-031).

2.2. Experimental Protocol and Equipment

The experimental setup was conducted in accordance with the guidelines provided by
the OpenCap software and hardware configuration protocol (Figure S1) [6].

2.2.1. Camera Setup and Calibration

Two iOS devices with rear-facing cameras were used to capture movement while
minimizing segment occlusions by positioning the cameras 30–45◦ off the participant’s
forward-facing line. This setup was chosen to avoid limb occlusion often caused by purely
sagittal views. Participants were first observed to ensure full visibility throughout the
task, and background distractions were minimized. Calibration was performed using a
checkerboard (210 × 175 mm) printed on A4 paper, with accurate dimensional verification.
The checkerboard was placed within the camera’s field of view to ensure correct calibration
of extrinsic parameters, maintaining visibility from all cameras within a 5 m distance to
achieve precise data collection.
Bioengineering 2024, 11, 911 4 of 18

2.2.2. Video Collection and Pose Estimation

Following calibration, videos were recorded at a resolution of 720 × 1280 pixels and a
frame rate of 60 Hz. Pose estimation was performed using two algorithms, OpenPose and
High-Resolution Network (HRNet), with 20 body keypoints selected for further analysis,
including the neck, mid-hip, shoulders, hips, knees, ankles, heels, toes, elbows, and wrists.
OpenCap uses a Direct Linear Transformation algorithm to triangulate 3D keypoint posi-
tions from synchronized 2D video data, weighted by keypoint confidence scores [12,17–20].
However, because 3D keypoints from video alone are insufficient for full biomechanical
analysis, two LSTM networks were trained to predict the positions of 43 anatomical markers
from 20 triangulated video keypoints, enhancing the robustness of joint kinematics.

2.2.3. Physics-Based Modeling and Simulation

After calibration, OpenCap guides users to record the participant in a neutral standing
pose, which is used to scale a musculoskeletal model to the participant’s anthropometry
via OpenSim’s Scale tool. OpenCap utilizes the musculoskeletal model developed by Lai
et al. [21,22], incorporating modifications to the hip abductor muscle paths as described by
Uhlrich et al. [23]. The musculoskeletal model, consisting of 33 degrees of freedom across
the pelvis (6 DOF in the ground frame), hips (3 DOF each), knees (1 DOF each), ankles
(2 DOF each), metatarsophalangeal joints (1 DOF each), lumbar region (3 DOF), shoulders
(3 DOF each), and elbows (2 DOF each), is driven by 80 muscles and 13 ideal torque motors,
with ground reaction forces simulated through foot–ground contact spheres. Once scaled,
OpenCap computes joint kinematics using OpenSim’s Inverse Kinematics tool, with users
able to visualize the resulting 3D kinematics within the web application.

2.3. Experimental Procedure

Participants were instructed to walk at a self-selected speed along a flat 4 m path after
achieving the static posture required by the equipment and establishing a connection with
the camera system. Each participant completed the walk three times, and the best perfor-
mance was selected for analysis. For participants who were unable to walk independently
or required supervision, a support individual was positioned behind them to ensure safety
while allowing the patient to appear as large as possible within the camera frame.

2.4. Data Collection and Processing

In this study, coordinate data collected from experimental marker trajectories were
exported as Track Row Column (TRC) files and analyzed using the Inverse Kinematics
tool in OpenSim 4.5 (SimTK, Stanford University, Stanford, CA, USA) [24]. Temporal
variations in the kinematics of the pelvis, hip, knee, and ankle were examined throughout
the gait cycle. Subsequently, the Inverse Dynamics tool was employed to calculate joint
forces, offering insights into the time-dependent changes in joint kinetics. Importantly, the
inverse dynamics analysis was performed under the explicit assumption of an external
force-free condition. This approach enabled us to isolate the internal joint moments and
forces generated by the system itself, thereby providing a more focused investigation into
the intrinsic dynamics of movement.

2.5. Statistical Analysis

The differences in baseline characteristics between the patient and control groups were
statistically evaluated. Continuous variables were analyzed using the Mann–Whitney U
test, while categorical variables were assessed using Fisher’s Exact test. A p-value of less
than 0.05 was considered statistically significant. Functional Data Analysis (FDA) was
applied to time-series kinematic and kinetic data for key gait parameters (pelvis, hip, knee,
and ankle) to model continuous joint movements across the entire gait cycle [14,25–27].
Each subject’s gait cycle was normalized to 100% to facilitate comparisons between the
patient and control groups. Bootstrap resampling with 1000 iterations was used to generate
95% confidence intervals (CIs) for each group, assessing the variability and statistical
Bioengineering 2024, 11, 911 5 of 18

significance of joint kinematics and kinetics [15,16]. Non-overlapping regions of the CIs
between the two groups were identified as statistically significant, highlighting phases
of the gait cycle where kinematic and kinetic deviations occurred. All analyses were
performed using custom scripts in Python 3.11.4, employing statistical libraries such as
statsmodels, scipy, numpy, and matplotlib for resampling and visualization, ensuring a
robust comparison of the time-series data.

3. Results
3.1. Patient Characteristics
The demographic and clinical characteristics of the control and patient groups are
presented in Table 1. The patient group consisted of four individuals diagnosed with
stroke, two with Parkinson’s disease, and four with other neurological disorders, including
cerebral palsy, Guillain–Barré syndrome, and cerebellar ataxia. The patient group was
significantly older (mean age: 51.60 years) compared to the control group (mean age:
31.30 years; p = 0.034). The sex distribution was similar between the groups, with four
males and six females in the patient group, and three males and seven females in the control
group (p = 1.000). No significant differences were observed between the groups in terms
of height, weight, BMI, or cadence. However, the patient group exhibited a significantly
slower gait speed (p = 0.002), shorter stride length (p = 0.001), and greater step length
asymmetry (p = 0.023). Additionally, the patient group demonstrated a significantly wider
step width compared to the control group (p = 0.045).

Table 1. The demographic and clinical characteristics of the control and patient groups.

Control (n = 10) Patient (n = 10)

p-Value
Mean (SD) Mean (SD)
Age (years) 31.30 (11.55) 51.60 (24.45) 0.034 *
Sex (male/female) 3M/7F 4M/6F 1.000
Height (m) 1.68 (0.09) 1.59 (0.21) 0.230
Weight (kg) 60.50 (16.13) 62.50 (18.91) 0.802
BMI (kg/m2 ) 21.18 (3.83) 24.26 (3.95) 0.093
Gait speed (m/s) 1.10 (0.13) 0.67 (0.31) 0.002 *
Stride length (m) 1.29 (0.15) 0.81 (0.31) 0.001 *
Step width (cm) 12.17 (3.10) 15.58 (3.92) 0.045 *
Cadence (step/min) 104.60 (9.93) 94.70 (28.92) 0.328
Double support (%cycle) 29.35 (2.72) 36.69 (12.50) 0.100
Step length asymmetry (%) 91.23 (11.70) 107.43 (16.76) 0.023 *
* p-value < 0.05 is considered statistically significant.

3.2. Kinematic Findings during the Gait Cycle in the Control Group
Figure 1 illustrates the kinematic data of 10 healthy individuals during a gait cycle,
representing the angular movement of key body segments across nine distinct parameters:
pelvic tilt, pelvic list, pelvic rotation, hip flexion/extension, hip adduction/abduction, hip
internal/external rotation, knee flexion/extension, ankle dorsiflexion/plantarflexion, and
subtalar inversion/eversion. Each graph plots the average angular movement across the
gait cycle, with the blue line representing the mean values and the shaded region indicating
the standard deviation (±1 SD). The data reveal that pelvic tilt oscillates within a range
of approximately 5◦ , while pelvic list demonstrates lateral shifting within ±5◦ , and pelvic
rotation fluctuates up to ±10◦ . Hip flexion/extension exhibits the most significant angular
variation, spanning from approximately −20◦ to 30◦ , with a marked flexion phase during
the early stance followed by extension. Hip adduction/abduction oscillates around ±10◦ ,
while internal/external hip rotation remains relatively stable within ±10◦ . Knee flexion
peaks at around 70◦ during the early stance, while ankle dorsiflexion/plantarflexion ranges
between −10◦ and 10◦ , reflecting typical ankle movement during gait. Lastly, subtalar
inversion/eversion exhibits subtle oscillations within ±10◦ , consistent with normal foot
movement during walking. These kinematic patterns collectively represent the typical
 ion peaks at around 70° during the early stance, while ankle dorsiflexion/plantarflexio
ranges between −10° and 10°, reflecting typical ankle movement during gait. Lastly, sub
talar inversion/eversion exhibits subtle oscillations within ±10°, consistent with norm
Bioengineering 2024, 11, 911 foot movement during walking. These kinematic patterns collectively represent6 of 18 the typ
cal biomechanical behavior of healthy individuals during gait, providing a robust refe
ence for analyzing deviations in pathological gait conditions.
biomechanical behavior of healthy individuals during gait, providing a robust reference for
analyzing deviations in pathological gait conditions.

Figure 1. Joint-specific kinematic parameters during the gait cycle, normalized for a group of controls.
Each subplot represents a specific joint movement across the gait cycle (%). The blue line indicates
the mean
Figure kinematic angle,
1. Joint-specific with shaded
kinematic parameters during±the
areas representing 1 standard deviation
gait cycle, (SD). Thefor
normalized following
a group of co
movements
trols. are shown:
Each subplot pelvic tilt,
represents list, andjoint
a specific rotation; hip flexion/extension,
movement across the gaitadduction/abduction,
cycle (%). The blue line ind
andthe
cates internal/external
mean kinematic rotation (IR/ER);
angle, withknee flexion/extension;
shaded ankle dorsiflexion/plantarflexion;
areas representing ±1 standard deviation (SD). Th
and subtalar inversion/eversion.
following movements are shown: pelvic tilt, list, and rotation; hip flexion/extension, adduction/a
duction, and internal/external
3.3. Kinematic Findings duringrotation (IR/ER);
the Gait Cycle knee
in the flexion/extension;
Patient Group ankle dorsiflexion/planta
flexion; and subtalar inversion/eversion.
The kinematic data of the 10 patients during the gait cycle reveal notable deviations
from normal gait mechanics, characterized by increased variability and reduced control
3.3.across
Kinematic Findings
multiple during the(Figure
joint movements Gait Cycle in thetilt
2). Pelvic Patient Group
fluctuates between −10◦ and 5◦ ,
with standard
The kinematicdeviations
data reaching 15◦ , whileduring
of the 10±patients pelvic list
theshows exaggerated
gait cycle reveallateral shifts
notable deviation
between − 10 ◦ and 10 ◦ , especially in the mid to late gait phase. Pelvic rotation exhibits
from normal gait mechanics, characterized by increased variability and reduced contr
similar angular ranges as healthy controls but with greater variability, particularly towards
the latter half of the cycle. Hip flexion/extension displays a constrained range of −15◦ to
30◦ , with reduced extension during the late stance, and hip adduction/abduction shows
diminished adduction with fluctuations from −5◦ to 10◦ . Internal/external rotation of
the hip shows increased instability, with deviations up to ±12◦ , particularly during the
swing phases. Knee flexion reaches similar peaks as normal controls, but the variability is
 minished adduction with fluctuations from −5° to 10°. Internal/external rotation of the hi
shows increased instability, with deviations up to ±12°, particularly during the swin
phases. Knee flexion reaches similar peaks as normal controls, but the variability is height
Bioengineering 2024, 11, 911 ened during the early stance, while ankle dorsiflexion/plantarflexion demonstrates 7 of 18 dimin
ished dorsiflexion control, particularly during the stance phase, with deviations reachin
±10°. Subtalar inversion/eversion fluctuates more widely between −5° and 10°, with in
heightened during the early stance, while ankle dorsiflexion/plantarflexion demonstrates
creased variability of up to ±8°, suggesting instability in foot mechanics. These data indi
diminished dorsiflexion control, particularly during the stance phase, with deviations
catereaching
that the patients
±10 exhibited
◦ . Subtalar increased variability
inversion/eversion andwidely
fluctuates more less consistent
between −5joint
◦ andcontrol,
10◦ , par
ticularly during variability
with increased key phases of to
of up the
±8gait
◦ cycle, reflecting
, suggesting instabilitydiverse compensatory
in foot mechanics. strategies.
These data
indicate that the patients exhibited increased variability and less consistent joint control,
particularly during key phases of the gait cycle, reflecting diverse compensatory strategies.

Figure 2. Joint-specific kinematic parameters during the gait cycle, normalized for a group of patients.
Each subplot represents a specific joint movement across the gait cycle (%). The blue line indicates
Figure 2. Joint-specific kinematic parameters during the gait cycle, normalized for a group of pa
the mean kinematic angle, with shaded areas representing ±1 standard deviation (SD). The following
tients. Each subplot represents a specific joint movement across the gait cycle (%). The blue lin
movements are shown: pelvic tilt, list, and rotation; hip flexion/extension, adduction/abduction,
indicates the mean kinematic angle, with shaded areas representing ±1 standard deviation (SD). Th
and internal/external rotation (IR/ER); knee flexion/extension; ankle dorsiflexion/plantarflexion;
following movements are shown: pelvic tilt, list, and rotation; hip flexion/extension, adduction/ab
and subtalar inversion/eversion.
duction, and internal/external rotation (IR/ER); knee flexion/extension; ankle dorsiflexion/plantar
flexion; and subtalar
3.4. Kinematic inversion/eversion.
Differences during the Gait Cycle: A Comparison between Control and
Patient Groups
The comparison of gait kinematics between the patient and control groups revealed
significant deviations across multiple joints, particularly in pelvic tilt, hip rotation, knee
flexion, and ankle dorsiflexion during key phases of the gait cycle (Figure 3). For instance, in
the pelvic tilt, there are substantial non-overlapping regions between 0–20% and 60–80% of
the gait cycle, with patients showing increased variability and larger tilt (ranging from −10◦
to 10◦ ) compared to controls (typically within −5◦ to 5◦ ). Similarly, in hip flexion/extension,
 80% of the gait cycle, with patients showing increased variability and larger tilt (rangin
from −10° to 10°) compared to controls (typically within −5° to 5°). Similarly, in hip flex
ion/extension, a significant difference is observed between 50 and 60% of the gait cycle
where patients demonstrate reduced extension, peaking at 20°, compared to controls, wh
Bioengineering 2024, 11, 911 achieve near 30°. The knee flexion/extension also reveals significant divergence 8 of 18 aroun
20–30% of the gait cycle, where patients show reduced flexion (approximately 50°) com
pared to the control group (around 70°). These statistically significant differences, marke
a significant difference is observed between 50 and 60% of the gait cycle, where patients
by demonstrate
non-overlapping CIs, underscore the altered gait mechanics in patients, reflecting bio
reduced extension, peaking at 20◦ , compared to controls, who achieve near
mechanical impairments,
30◦ . The knee reducedalso
flexion/extension mobility,
reveals and potential
significant compensatory
divergence strategies
around 20–30% of durin
gait. ◦
the gait cycle, where patients show reduced flexion (approximately 50 ) compared to
the control group (around 70◦ ). These statistically significant differences, marked by non-
overlapping CIs, underscore the altered gait mechanics in patients, reflecting biomechanical
impairments, reduced mobility, and potential compensatory strategies during gait.

Figure 3. Bootstrap confidence bands and non-overlapping regions for hip, knee, and ankle pa-
rameters
Figure in patientsconfidence
3. Bootstrap and controls. This figure
bands illustrates the bootstrap
and non-overlapping confidence
regions for hip,intervals (CI) ankle
knee, and and param
regions of statistically significant differences between patients (blue) and controls (green) for key
eters in patients and controls. This figure illustrates the bootstrap confidence intervals (CI) and re
gait parameters. The shaded regions around each curve represent the 95% confidence intervals
gions of statistically significant differences between patients (blue) and controls (green) for key ga
generated through 1000 bootstrap resamples. Red-highlighted vertical spans indicate areas where the
parameters. The shaded regions around each curve represent the 95% confidence intervals gene
confidence intervals of the two groups do not overlap, suggesting statistically significant differences
ated through 1000 bootstrap resamples. Red-highlighted vertical spans indicate areas where th
in these regions.
confidence intervals of the two groups do not overlap, suggesting statistically significant difference
in these Table
regions.
2 summarizes the peak joint angles for the hip, knee, ankle, and subtalar joints
in both the control and patient groups, including their respective means and standard
deviations (SD). The comparison of the maximum, minimum, and mean values of the
kinematic parameters across the gait cycle is presented in Table S1. A significant difference
was observed in hip extension (p = 0.007), indicating a marked discrepancy between the
control and patient groups in this movement. No statistically significant differences were
found for the other joint movements.
Bioengineering 2024, 11, 911 9 of 18

Table 2. Comparison of peak joint angles between control and patient groups.

Control Mean Patients Mean

Joint Peak Value (Degree) p-Value
(SD) (SD)
Hip Flexion 24.225 (2.348) 26.146 (10.074) 0.57
Extension −21.736 (5.938) −11.52 (8.642) 0.007 *
Adduction 10.599 (5.291) 8.431 (3.068) 0.281
Abduction −7.937 (2.402) −7.544 (2.465) 0.722
Internal Rotation −1.208 (3.307) 1.121 (6.209) 0.313
External Rotation −13.663 (5.101) −13.629 (6.114) 0.989
Knee Flexion 61.492 (5.702) 53.952 (12.767) 0.113
Extension 1.152 (1.013) 2.168 (2.458) 0.25
Ankle Dorsiflexion 20.416 (10.514) 16.253 (6.256) 0.299
Plantarflexion −14.189 (13.056) −6.667 (8.03) 0.142
Subtalar Inversion 17.496 (7.34) 14.758 (8.779) 0.459
Eversion −15.709 (6.618) −11.021 (12.62) 0.316
* p-value < 0.05 is considered statistically significant.

3.5. Gait Cycle Kinematics in Stroke Patients versus Controls: A Subgroup Comparison
Figure 4 compares the kinematic gait data of 10 healthy individuals and 10 stroke
patients, focusing on key parameters such as the pelvis, hip, knee, and ankle joints. “In
pelvic tilt, there are substantial non-overlapping regions between 0–40% and 50–60% of
the gait cycle, indicating significant differences in tilt control, with stroke patients showing
greater fluctuations, tilting from −2.5◦ to 5◦ , compared to the controls, whose tilt remained
more stable around ±2◦ . The pelvic list shows less pronounced variability between groups,
although the stroke group exhibits a broader fluctuation, especially during the mid-phase
of the gait cycle. In pelvic rotation, significant differences emerge towards the end of the
gait cycle (90–100%), where stroke patients show reduced rotation control, with ranges
from 5◦ to −7.5◦ , while controls maintain a more steady pattern between 5◦ and −5◦ .
For hip flexion/extension, significant differences occur around 40–60% of the gait cycle,
where stroke patients display a more restricted range of motion (hip flexion peaking at
20◦ compared to the control group at nearly 30◦ ). In hip adduction/abduction, notable
deviations are observed near the end of the gait cycle (80–100%), where stroke patients
demonstrate reduced adduction capacity, with values around −5◦ to −7.5◦ compared to
controls. The hip internal/external rotation shows increased variability across the entire
cycle for stroke patients, though without marked regions of non-overlapping significance.
The knee flexion/extension reveals significant differences between 40 and 60% of the gait
cycle, where stroke patients exhibit less flexion (around 40–50◦ ) compared to the control
group (approaching 70◦ ). In ankle dorsiflexion/plantarflexion, non-overlapping regions are
present at both the early and late stages of the gait cycle (0–20% and 80–100%), indicating
substantial impairments in stroke patients, with their dorsiflexion being markedly reduced,
ranging between −5◦ and 5◦ compared to the broader range of −10◦ to 10◦ seen in controls.
Lastly, in subtalar inversion/eversion, although both groups exhibit similar trends, the
stroke patients display greater variability, particularly in the mid-stance phase, with more
pronounced deviations around inversion control. These differences in kinematic patterns
suggest significant impairments in joint mobility and stability in stroke patients, especially
in pelvic control, hip and knee flexion/extension, and ankle motion. The regions of
non-overlapping confidence intervals underline the severity of gait deviations in stroke
patients, likely due to neurological deficits that affect motor control and coordination
during ambulation.
 The regions of non-overlapping confidence intervals underline the severity of gait devia-
tions in stroke patients, likely due to neurological deficits that affect motor control and
coordination during ambulation.
Bioengineering 2024, 11, 911 10 of 18

Figure 4. Comparison of bootstrap confidence intervals and non-overlapping regions for hip flexion,
Figure 4. Comparison of bootstrap confidence intervals and non-overlapping regions for hip flexion,
hip adduction, knee angle, and ankle angle in stroke patients and normal controls. The blue and
hip adduction, knee angle, and ankle angle in stroke patients and normal controls. The blue and
green lines represent the mean curves for stroke patients and healthy controls, respectively, while the
green lines represent the mean curves for stroke patients and healthy controls, respectively, while
shaded regions indicate the 95% bootstrap confidence intervals (CI) calculated across the gait cycle.
the shaded regions indicate the 95% bootstrap confidence intervals (CI) calculated across the gait
Red-highlighted regions indicate statistically significant differences where the confidence intervals
cycle. Red-highlighted regions indicate statistically significant differences where the confidence in-
between the two groups do not overlap.
tervals between the two groups do not overlap.
3.6. Gait Cycle Kinematics in Parkinson’s Disease Patients versus Healthy Controls:
A Subgroup
3.6. Gait Cycle Comparison
Kinematics in Parkinson’s Disease Patients versus Healthy Controls: A
The Comparison
Subgroup comparison of gait kinematics between Parkinson’s disease patients and controls
revealed significant deviations across multiple joints during key phases of the gait cycle
The comparison of gait kinematics between Parkinson’s disease patients and controls
(Figure 5). Notably, pelvic tilt does not exhibit any statistically significant differences
revealed
between significant
the groups, deviations
as evidenced across
by themultiple
absence ofjoints during key regions,
red-highlighted phases indicating
of the gait cycle
(Figure 5). Notably, pelvic tilt does not exhibit any statistically significant
that both groups maintain similar pelvic tilt patterns throughout the gait cycle. However, differences be-
tween thelist,
in pelvic groups, as evidenced
significant differencesbyarethe absence
observed of red-highlighted
between 40 and 50% of theregions, indicating
gait cycle, with that
both groups maintain
the Parkinson’s similar pelvic
group displaying greatertiltlateral
patterns throughout
deviations comparedthe gait cycle. However,
to controls. Pelvic in
rotation shows a statistically significant difference around 60% of the gait
pelvic list, significant differences are observed between 40 and 50% of the gait cycle, with cycle, with the
Parkinson’s
the Parkinson’s group exhibiting
group reducedgreater
displaying rotational control.
lateral Differences
deviations are also prominent
compared to controls. in Pelvic
hip flexion/extension between 60 and 75% of the gait cycle, where Parkinson’s patients
rotation shows a statistically significant difference around 60% of the gait cycle, with the
show reduced hip extension. In hip adduction/abduction, significant deviations occur
Parkinson’s group exhibiting reduced rotational control. Differences are also prominent
around 40–60% of the cycle, indicating reduced adduction in the Parkinson’s group. For hip
ininternal/external
hip flexion/extension
rotation,between
there are 60 and 75%
significant of the gait
differences cycle, where
at 50–60% Parkinson’s
and 90–100% patients
of the cy-
show
cle, reflecting greater variability in the Parkinson’s group. Similarly, knee flexion/extension occur
reduced hip extension. In hip adduction/abduction, significant deviations
around 40–60% of the
differs significantly cycle, 50
between indicating
and 60%,reduced adduction
with reduced flexionininthe Parkinson’s
Parkinson’s group. For
patients.
hip internal/external
Ankle rotation, there
dorsiflexion/plantarflexion showsaresignificant
significant differences
deviations at 50–60%
at three and
phases of the90–100%
gait of
cycle—70%
the and 80–90%—with
cycle, reflecting greater Parkinson’s
variabilitypatients
in thedemonstrating
Parkinson’s reduced
group.dorsiflexion
Similarly, knee
control. Lastly, subtalar inversion/eversion presents statistically significant differences in
early phases of the cycle, indicating greater instability in foot mechanics among Parkinson’s
patients. These findings suggest that individuals with Parkinson’s disease experience signif-
 duced dorsiflexion control. Lastly, subtalar inversion/eversion presents statistically signif-
icant differences in early phases of the cycle, indicating greater instability in foot mechan-
ics among Parkinson’s patients. These findings suggest that individuals with Parkinson’s
disease experience significant motor impairments across multiple joints during walking,
Bioengineering 2024, 11, 911 with increased variability and reduced control particularly evident in hip, knee, and
11 of 18 ankle
movements.

icant motor impairments across multiple joints during walking, with increased variability
and reduced control particularly evident in hip, knee, and ankle movements.

Figure 5. Comparison of bootstrap confidence intervals and non-overlapping regions for hip flexion,
hip adduction,
Figure knee flexion,
5. Comparison and ankle
of bootstrap plantarflexion/dorsiflexion
confidence in the Parkinson’s
intervals and non-overlapping patients
regions andflexion,
for hip
controls.
hip The mean
adduction, kneejoint angle trajectories
flexion, and ankle for Parkinson’s patients are shown
plantarflexion/dorsiflexion in blue,
in the while the patients
Parkinson’s green and
lines represent healthy controls. The shaded regions around each curve represent the 95%
controls. The mean joint angle trajectories for Parkinson’s patients are shown in blue, while the bootstrap
confidence
green intervals (CIs).
lines represent Red-highlighted
healthy controls. Theareas indicateregions
shaded time points whereeach
around the confidence intervalsthe 95%
curve represent
do not overlap, signifying statistically significant differences between the two groups.
bootstrap confidence intervals (CIs). Red-highlighted areas indicate time points where the confi-
dence intervals do not overlap, signifying statistically significant differences between the two
3.7. Gait Cycle Kinematics in Pediatric Patients versus Healthy Controls: A Subgroup Comparison
groups.
Figure 6 compares the kinematic gait data of pediatric patients and healthy controls
across key parameters such as the pelvis, hip, knee, and ankle joints. In pelvic tilt, significant
3.7. Gait Cycle Kinematics in Pediatric Patients versus Healthy Controls: A
differences are observed across 0–90% of the gait cycle, with pediatric patients exhibiting
Subgroup Comparison ranging from −30◦ to 0◦ , compared to controls, whose tilt remains
greater fluctuations,
more Figure
stable6within −5◦ to
compares the5◦kinematic
. In pelvic gait data of pediatric
list, significant patients
differences and healthy
are observed aroundcontrols
0–10%,
across 50–60%,
key and 80–100%
parameters such asofthethepelvis,
gait cycle,
hip, with
knee,pediatric
and anklepatients
joints.showing
In pelvicgreater
tilt, signifi-
lateral
cant deviations
differences are(− 15◦ to 5◦across
observed ) compared
0–90% to of
controls 5◦ to with
the gait(−cycle, 5◦ ). For pelvic patients
pediatric rotation,exhib-
significant differences occur within the 0–30% range, with pediatric patients displaying
iting greater fluctuations, ranging from −30° to 0°, compared to controls, whose tilt re-
increased variability, fluctuating between 0◦ and 30◦ , while controls remain more stable
mains ◦more◦ stable within −5° to 5°. In pelvic list, significant differences are observed
at −5 to 5 . In hip flexion/extension, differences are noted around 0–10%, 40–50%, and
around
90–100% 0–10%, 50–60%,
of the gait cycle, and
where 80–100%
pediatricof the gait
patients cycle,
exhibit withpeak
higher pediatric patients
hip flexion showing
(around
greater
◦ lateral deviations (−15° to◦ 5°) compared to controls (−5° to 5°).
50 ) compared to controls (30–40 ). In knee flexion/extension, significant differences For pelvic rotation,
significant differences
appear around 0–20%, occur
70–80%,within the 0–30%
and 90–100% range,
of the gait with
cycle,pediatric patients
with pediatric displaying
patients
showing reduced knee flexion, peaking at 50◦ , compared to 70◦ in controls. In ankle
dorsiflexion/plantarflexion, differences are evident across a large portion of the gait cycle,
specifically 0–10% and 20–100%, with pediatric patients showing restricted movement
between −30◦ and 5◦ , compared to controls who range from −20◦ to 15◦ . Lastly, in subtalar
inversion/eversion, significant differences are observed between 20–50% and 80–100% of
 pear around 0–20%, 70–80%, and 90–100% of the gait cycle, with pediatric patients show-
ing reduced knee flexion, peaking at 50°, compared to 70° in controls. In ankle dorsiflex-
ion/plantarflexion, differences are evident across a large portion of the gait cycle, specifi-
cally 0–10% and 20–100%, with pediatric patients showing restricted movement between
Bioengineering 2024, 11, 911 −30° and 5°, compared to controls who range from −20° to 15°. Lastly, in subtalar 12 of 18 inver-
sion/eversion, significant differences are observed between 20–50% and 80–100% of the
gait cycle, where pediatric patients exhibit greater variability, with values ranging from
the gait cycle, where pediatric patients exhibit greater variability, with values ranging from
−5° to 20°, compared to controls whose range is between −10° and 10°. These findings in-
−5◦ to 20◦ , compared to controls whose range is between −10◦ and 10◦ . These findings
dicate that pediatric patients exhibit significant motor control differences across multiple
indicate that pediatric patients exhibit significant motor control differences across multiple
joints,
joints,characterized bygreater
characterized by greater variability
variability andand reduced
reduced stability
stability duringduring gait, particularly
gait, particularly in
inthe
thehip,
hip,knee,
knee, ankle,
ankle, andand subtalar
subtalar jointsjoints

Figure
Figure6.6.Comparison
Comparison of bootstrapconfidence
of bootstrap confidence intervals
intervals andand non-overlapping
non-overlapping regions
regions for
for hip, hip, knee,
knee,
andandankle
anklejoint
jointkinematics
kinematics in inchildren
childrenandand adults.
adults. This
This figure
figure presents
presents a comparison
a comparison of limb
of lower lower limb
joint
jointkinematics betweenchildren
kinematics between childrenandand healthy
healthy adultadult controls
controls throughout
throughout the gait
the gait cycle, cycle,on
focusing focusing
the on
the following key parameters: hip flexion/adduction, knee flexion, and ankle plantarflexion/dorsi-
following key parameters: hip flexion/adduction, knee flexion, and ankle plantarflexion/dorsiflexion.
flexion.
The mean Thejoint
mean joint
angle angle trajectories
trajectories for childrenfor
arechildren
shown inare shown
blue, while in blue,
those for while
healthythose
adultsfor
arehealthy
adults are displayed
displayed in green. in green.regions
Shaded Shaded regions
around eacharound
curve each curvethe
represent represent the 95%
95% bootstrap bootstrap con-
confidence
fidence intervals
intervals (CIs). Red-highlighted
(CIs). Red-highlighted areas indicate
areas indicate time pointstime points
where thewhere the confidence
confidence intervals do intervals
not do
not overlap,
overlap, denoting
denoting statistically
statistically significant
significant differences
differences betweenbetween the two groups.
the two groups.

3.8.Kinetic
3.8. KineticDifferences
Differences during
duringthe
theGait
GaitCycle: A Comparison
Cycle: between
A Comparison Control
between and Patient
Control and Groups
The analysis
Patient Groups of joint moments during the gait cycle, normalized to body weight,
demonstrated significant differences between the control and patient groups across the
hip,The
knee,analysis of joint
and ankle joints moments
(Figure S1).during
In hipthe gait cycle, normalized
flexion/extension moments, tothebody weight,
patient
demonstrated
group exhibited increased variability, particularly during the mid-stance phase (30–35% the
significant differences between the control and patient groups across
hip, knee,
of the gaitand ankle
cycle), withjoints
mean(Figure S1).ofInuphip
differences flexion/extension
to 0.1 Nm/kg compared moments, the patient
to the control
group
group. exhibited increased
Statistically variability,
significant particularly
differences were observed,during the mid-stance
particularly phaseand
in mid-stance (30–35%
the terminal stance, as reflected by non-overlapping 95% confidence intervals. In knee
flexion/extension moments, the patient group showed deviations of up to 0.05 Nm/kg
during the loading response (8–10%) and mid-stance (30–35%) phases, indicating altered
joint loading patterns. Finally, in ankle dorsiflexion/plantarflexion moments, the patient
group demonstrated reduced plantarflexion moments during the push-off phase (40–45%),
with differences reaching 0.05 Nm/kg compared to the control group. These deviations
Bioengineering 2024, 11, 911 13 of 18

were statistically significant, as the 95% confidence intervals did not overlap with zero.
Collectively, these findings suggest that patients with neurological disorders display altered
joint moment patterns during gait, likely due to compensatory mechanisms and muscle
impairments, particularly during critical phases of the gait cycle.

4. Discussion
This study demonstrated that OpenCap, an innovative open-source program, is highly
effective in the simple and efficient measurement of kinematics and kinetics in both healthy
individuals and patient populations. OpenCap’s ability to accurately capture movement
dynamics using only video data from smartphones made it particularly useful in clinical
settings where traditional motion capture systems are either impractical or unavailable.
The software’s user-friendly interface and robust algorithms enabled the collection of
comprehensive gait data, which were successfully applied to various patient groups,
including those with neurological conditions such as stroke, Parkinson’s disease, and
cerebral palsy. The program’s adaptability to different patient needs, even in cases of
impaired mobility, highlights its potential as a valuable tool for clinical gait analysis and
for tracking patient progress over time.

4.1. Comparison of Temporospatial Gait Parameters and Efficiency of Data Acquisition in

Neurological Conditions
The temporospatial gait parameters observed in our patient group closely align with
findings reported in studies involving individuals with neurological conditions such as
stroke, Parkinson’s disease, and cerebral palsy. For instance, our patient group’s mean gait
speed was 0.67 m/s, which is consistent with the reduced gait speeds typically reported in
stroke patients (0.4–0.8 m/s) and Parkinson’s disease patients (0.6–0.9 m/s) [7,9,11,28–32].
Similarly, the stride length in our patient group averaged 0.81 m, which is markedly shorter
than the control group’s 1.29 m, and is comparable to the reduced stride lengths observed
in patients with cerebral palsy (0.7–1.0 m) and Parkinson’s disease (0.8–1.1 m) [4,8,33,34].
Additionally, the increased step length asymmetry in our patient group (107.43%) is in
line with findings from stroke and cerebral palsy patients, where asymmetry percentages
typically exceed 100%, reflecting the challenges in maintaining symmetrical gait patterns in
these populations.
These comparisons underscore the validity of our findings and suggest that the gait
characteristics of our patient group mirror those seen in individuals with well-documented
neurological impairments. Moreover, the data acquisition process in our study was notably
efficient, with all measurements being completed within a brief period of under 10 min. This
efficiency stands in stark contrast to traditional gait analysis methods, which often require
extensive setup and prolonged data collection times. The use of iOS-based motion capture
technology in our study allowed for rapid, yet accurate, acquisition of gait parameters,
demonstrating not only the practicality of this approach but also its potential for broad
clinical application. The ability to quickly and accurately assess gait characteristics in
patients, especially those with neurological conditions, is critical for timely diagnosis
and intervention, and our method offers a valuable tool for achieving this in diverse
clinical settings.

4.2. Comparative Analysis of Kinematic Gait Patterns in Controls and Patients with
Neurological Disorders
This study provided a comparative analysis of gait patterns between individuals
with neurological conditions, such as stroke and Parkinson’s disease, and healthy controls.
Our findings highlighted several key deviations in gait kinematics among neurological
patients, consistent with prior studies using traditional marker-based systems [7–9]. In
stroke patients, we observed a reduction in hip flexion during the swing phase, where
the hip flexion decreased by up to 10 degrees compared to the normal range of 35 to
40 degrees [7,28,35]. This reduction contributes to the asymmetric gait patterns commonly
observed in stroke patients, leading to increased joint loading on the unaffected side, as
Bioengineering 2024, 11, 911 14 of 18

previously documented in studies that observed a 15–20% increase in joint loading on the
contralateral side [29].
In comparison to previous studies on Parkinson’s disease (PD) gait patterns, the results
from this study reinforce and add specificity to our understanding of motor impairments
during walking. The reduction in hip extension observed between 60 and 75% of the gait
cycle in this study aligns with prior research that has quantified reduced hip extension
by up to 9–11 degrees in PD patients compared to healthy controls [36]. Similarly, ankle
dorsiflexion impairments were prominent in this study, particularly between 70 and 90%
of the gait cycle, reflecting the findings by Nanhoe-Mahabier et al., who reported that PD
patients exhibit reduced dorsiflexion by approximately 6–8 degrees during the terminal
stance phase [30]. Moreover, the increased variability in hip internal/external rotation and
subtalar inversion/eversion at different phases of the gait cycle observed here aligns with
earlier work by Frenkel-Toledo et al., who found increased variability in gait metrics, such
as step length and time, and irregular joint movements [31]. Their research indicated that
this variability is 20–30% higher in PD patients than in controls, contributing to balance
instability. Finally, the deviations in pelvic list and rotation reported in the current study
echo findings by Nieuwboer et al., which showed that PD patients have a 3–5 degree
greater lateral pelvic list during the gait cycle, affecting overall gait symmetry [32]. By
quantifying the significant differences across multiple joints, this study corroborates the
extent of movement impairment in Parkinson’s disease, while these numbers help specify
how PD patients deviate from normal gait patterns.
Despite the overall consistency with previous studies, there are some discrepancies,
particularly in the estimation of joint forces and muscle activation patterns. Traditional
gait analyses that incorporate force plates provide more precise measurements of ground
reaction forces (GRF), which are crucial for accurate kinetic calculations. In our study, the
absence of direct GRF measurements may have introduced some degree of inaccuracy in
the joint force estimates, particularly during dynamic phases of the gait cycle. Studies using
marker-based systems with GRF have been able to produce more refined joint force data,
which may explain some of the differences observed [13].

4.3. Comparative Analysis of Kinematic Gait Patterns between Healthy Controls and Pediatric
Patients with Neurological Disorders
In comparing pediatric patients with neurological conditions, such as cerebral palsy,
to healthy controls, we found consistent deviations in gait kinematics, echoing results
from previous research. Specifically, pediatric patients exhibited reduced hip extension
by 5–10 degrees during the late stance phase. This reduction aligns closely with findings
from studies on children with cerebral palsy, which reported hip extension reductions of
up to 12 degrees [37]. Additionally, our study showed a reduction of 10–15 degrees in knee
flexion during the swing phase, which closely matches the reductions reported in children
with spastic diplegia, where knee flexion was reduced by 10–14 degrees. These reductions
in knee flexion contribute to difficulties in limb clearance during gait, a common issue
in pediatric neurological populations that often leads to a characteristic “stiff-knee” gait
pattern [38]. Ankle kinematics in pediatric patients revealed restricted dorsiflexion during
the early stance (10–20%) and push-off, with reductions of 10 degrees. This is in agreement
with prior studies that reported a 6–8 degree reduction in dorsiflexion in children with
cerebral palsy [33]. Restricted dorsiflexion limits the ability to achieve a smooth heel-
to-toe transition during walking, which can negatively affect the efficiency and fluidity
of gait. Collectively, these deviations highlight the significant impact that neurological
impairments have on gait stability, range of motion, and overall mobility. These deviations
likely result from compensatory mechanisms related to weakened musculature and poor
neuromotor control, as previously suggested in studies [3,4,39]. Addressing these specific
deviations through targeted rehabilitation programs, particularly those aimed at improving
pelvic stability, joint mobility, and neuromuscular coordination, could lead to significant
improvements in functional gait for pediatric patients.
Bioengineering 2024, 11, 911 15 of 18

4.4. Limitations of the Study

One significant limitation of this study is the absence of direct ground reaction force
(GRF) measurements, which are crucial for accurate kinetic analysis. GRF data provide
essential insights into the external forces acting on the body during movement, and without
these measurements, estimates of joint forces and muscle activations may lack precision,
particularly during dynamic phases of the gait cycle [40,41]. The assumption of an external
force-free condition in the inverse dynamics analysis further departs from real-world
conditions, especially during the stance phase, where GRF plays a substantial role in
influencing joint kinetics. As a result, the subtle yet important variations typically observed
between these phases were less pronounced, complicating the interpretation of the data.
Future studies could address this by integrating OpenCap with portable force plates
or in-shoe sensors to capture GRF during movement [42]. Additionally, incorporating
muscle-driven simulations with algorithmic differentiation would allow for more accurate
estimations of joint dynamics by accounting for muscle activations, skeletal dynamics, and
external forces [43]. Leveraging such approaches, supported by optimal control methods
and tools like CasADi, would significantly enhance the accuracy of dynamic analyses,
particularly during the load-bearing phases of movement [44].
While the camera-based motion capture system used in this study offers flexibility
and ease of use, such markerless systems generally exhibit reduced accuracy compared
to traditional marker-based systems, with factors such as occlusion, lighting, and cam-
era angles potentially introducing errors in kinematic data, especially during complex
or abnormal gait patterns [6,45–47]. Future research should include a validation of the
accuracy of the OpenCap system in specific patient populations, such as those with stroke,
Parkinson’s disease, and pediatric patients. Currently, the default model provided by
OpenCap, the LaiUlrich2022 model, is a modified version of the Rajagopal model based on
adult data [6,21]. Incorporating a pediatric-specific model, which accounts for differences
in height, weight, and muscle proportions, could improve the accuracy of kinematic and
dynamic estimations for pediatric populations, providing more reliable information across
diverse age groups and clinical conditions.
Before initiating the gait analysis using the OpenCap system, a short period of static
posture—either neutral standing or any pose—must be maintained for model scaling,
typically within 5 s. This process requires the individual to be captured from two camera
angles while maintaining a static position unaided, which poses challenges for patients with
compromised balance or pediatric populations. Furthermore, many patients undergoing
rehabilitation are unable to walk independently and rely on assistive devices such as canes
or walkers, rendering them unable to undergo proper evaluation due to camera vision
occlusion. Even for patients who do not use assistive devices but require close supervision
to prevent falls, the lack of multi-person identification support in the OpenCap system
presents difficulties in ensuring accurate measurements. In the future, the development
of protocols that support multi-camera setups capable of tracking markers from the rear
or software that allows for multi-person identification could greatly expand the usability
of the OpenCap system for a broader range of patients, including those with neurological
conditions who utilize assistive devices.
The relatively small sample size also limits the generalizability of our findings. With
only 20 participants across different patient groups, the study may not fully represent the
broader population of individuals with neurological impairments. A larger sample size
would provide more statistical power, allowing for a more nuanced analysis and increased
confidence in detecting subtle differences between the patient groups.

5. Conclusions
This study demonstrates that OpenCap, a smartphone-based motion capture system,
is a feasible and cost-effective tool for clinical gait analysis in patients with neurological
disorders. The system was able to capture significant differences in gait parameters be-
tween healthy controls and patients with conditions such as stroke, Parkinson’s disease,
Bioengineering 2024, 11, 911 16 of 18

and pediatric patients. These findings highlight the potential of OpenCap to enhance
accessibility to biomechanical assessments, offering a practical alternative for gait analysis
in clinical settings where traditional motion capture systems may not be viable.

Supplementary Materials: The following supporting information can be downloaded at:

https://www.mdpi.com/article/10.3390/bioengineering11090911/s1, Figure S1. Overview of the
Experimental Setup and Data Analysis Workflow.; Figure S2. Joint moments during the gait cycle
normalized to body weight for the control and patient groups.; Table S1. Comparative analysis of
kinematic parameters during gait cycle between control and patient groups.
Author Contributions: Conceptualization, Y.-S.M., H.C.K. and J.C.L.; methodology, E.P.; soft-
ware, E.P.; validation, T.-D.J. and Y.-S.L.; formal analysis, Y.-S.M.; investigation, Y.K. and H.J.K.;
resources, Y.K. and H.J.K.; data curation, Y.K. and H.J.K.; writing—original draft preparation, Y.-S.M.;
writing—review and editing, J.C.L.; visualization, Y.-S.M.; supervision, H.C.K.; project administra-
tion, J.C.L. All authors have read and agreed to the published version of the manuscript.
Funding: This study was supported by the BK21 FOUR project (AI-driven Convergence Software
Education Research Program) funded by the Ministry of Education, School of Computer Science and
Engineering, Kyungpook National University, Korea (4120240214871).
Institutional Review Board Statement: The study was conducted in accordance with the Declaration
of Helsinki, and approved by the Institutional Review Board of Kyungpook National University
Chilgok Hospital (KNUCH-protocol code 2023-09-031).
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
Written informed consent has been obtained from the patients to publish this paper.
Data Availability Statement: The data presented in this study are available on request from the
corresponding author.
Conflicts of Interest: The authors declare no conflicts of interest.

References
1. Gage, J.R. Gait Analysis in Cerebral Palsy; Cambridge University Press: Cambridge, UK, 1991; ISBN 978-0-521-41277-3.
2. Heinen, F.; Desloovere, K.; Schroeder, A.S.; Berweck, S.; Borggraefe, I.; van Campenhout, A.; Andersen, G.L.; Aydin, R.; Becher,
J.G.; Bernert, G.; et al. The Updated European Consensus 2009 on the Use of Botulinum Toxin for Children with Cerebral Palsy.
Eur. J. Paediatr. Neurol. 2010, 14, 45–66. [CrossRef] [PubMed]
3. Shrader, W.; Shih, C.; McDonald, T. Instrumented Gait Analysis in the Care of Children with Cerebral Palsy: Current Concept
Revew. J. Pediatr. Orthop. Soc. N. Am. 2021, 3, 237. [CrossRef]
4. Rasmussen, H.M.; Pedersen, N.W.; Overgaard, S.; Hansen, L.K.; Dunkhase-Heinl, U.; Petkov, Y.; Engell, V.; Baker, R.; Holsgaard-
Larsen, A. The Use of Instrumented Gait Analysis for Individually Tailored Interdisciplinary Interventions in Children with
Cerebral Palsy: A Randomised Controlled Trial Protocol. BMC Pediatr. 2015, 15, 202. [CrossRef] [PubMed]
5. Rodrigues, T.B.; Salgado, D.P.; Catháin, C.Ó.; O’Connor, N.; Murray, N. Human Gait Assessment Using a 3D Marker-Less
Multimodal Motion Capture System. Multimed. Tools Appl. 2020, 79, 2629–2651. [CrossRef]
6. Uhlrich, S.D.; Falisse, A.; Kidziński, Ł.; Muccini, J.; Ko, M.; Chaudhari, A.S.; Hicks, J.L.; Delp, S.L. OpenCap: Human Movement
Dynamics from Smartphone Videos. PLoS Comput. Biol. 2023, 19, e1011462. [CrossRef]
7. Cimolin, V.; Vismara, L.; Ferraris, C.; Amprimo, G.; Pettiti, G.; Lopez, R.; Galli, M.; Cremascoli, R.; Sinagra, S.; Mauro, A.; et al.
Computation of Gait Parameters in Post Stroke and Parkinson’s Disease: A Comparative Study Using RGB-D Sensors and
Optoelectronic Systems. Sensors 2022, 22, 824. [CrossRef]
8. Steffensen, E.A.; Magalhães, F.; Knarr, B.A.; Kingston, D.C. Comparison of Markerless and Marker-Based Motion Capture of Gait
Kinematics in Individuals with Cerebral Palsy and Chronic Stroke: A Case Study Series. Res. Sq. 2023, rs.3.rs-2557403. [CrossRef]
9. Mohan, D.M.; Khandoker, A.H.; Wasti, S.A.; Ismail Ibrahim Ismail Alali, S.; Jelinek, H.F.; Khalaf, K. Assessment Methods of
Post-Stroke Gait: A Scoping Review of Technology-Driven Approaches to Gait Characterization and Analysis. Front. Neurol. 2021,
12, 650024. [CrossRef]
10. Alberto, S.; Cabral, S.; Proença, J.; Pona-Ferreira, F.; Leitão, M.; Bouça-Machado, R.; Kauppila, L.A.; Veloso, A.P.; Costa,
R.M.; Ferreira, J.J.; et al. Validation of Quantitative Gait Analysis Systems for Parkinson’s Disease for Use in Supervised and
Unsupervised Environments. BMC Neurol. 2021, 21, 331. [CrossRef]
11. di Biase, L.; Di Santo, A.; Caminiti, M.L.; De Liso, A.; Shah, S.A.; Ricci, L.; Di Lazzaro, V. Gait Analysis in Parkinson’s Disease: An
Overview of the Most Accurate Markers for Diagnosis and Symptoms Monitoring. Sensors 2020, 20, 3529. [CrossRef]
12. Cao, Z.; Hidalgo, G.; Simon, T.; Wei, S.-E.; Sheikh, Y. OpenPose: Realtime Multi-Person 2D Pose Estimation Using Part Affinity
Fields. IEEE Trans. Pattern Anal. Mach. Intell. 2021, 43, 172–186. [CrossRef] [PubMed]
Bioengineering 2024, 11, 911 17 of 18

13. Kanko, R.M.; Laende, E.K.; Davis, E.M.; Selbie, W.S.; Deluzio, K.J. Concurrent Assessment of Gait Kinematics Using Marker-Based
and Markerless Motion Capture. J. Biomech. 2021, 127, 110665. [CrossRef] [PubMed]
14. Ramsay, J.O.; Silverman, B.W. Functional Data Analysis; Springer Series in Statistics; Springer: New York, NY, USA, 2005;
ISBN 978-0-387-40080-8.
15. Carpenter, J.; Bithell, J. Bootstrap Confidence Intervals: When, Which, What? A Practical Guide for Medical Statisticians. Stat.
Med. 2000, 19, 1141–1164. [CrossRef]
16. Bland, J.M.; Altman, D.G. Statistics Notes: Bootstrap Resampling Methods. Br. Med. J. 2015, 350, h2622. [CrossRef]
17. Ren, S.; He, K.; Girshick, R.; Sun, J. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks. IEEE
Trans. Pattern Anal. Mach. Intell. 2016, 39, 1137–1149. [CrossRef]
18. Sun, K.; Xiao, B.; Liu, D.; Wang, J. Deep High-Resolution Representation Learning for Human Pose Estimation. In Proceedings
of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, CA, USA, 20 June 2019;
pp. 5686–5696.
19. Jin, S.; Xu, L.; Xu, J.; Wang, C.; Liu, W.; Qian, C.; Ouyang, W.; Luo, P. Whole-Body Human Pose Estimation in the Wild. arXiv 2020,
arXiv:2007.11858.
20. Zhang, F.; Zhu, X.; Dai, H.; Ye, M.; Zhu, C. Distribution-Aware Coordinate Representation for Human Pose Estimation. arXiv
2019, arXiv:1910.06278.
21. Rajagopal, A.; Dembia, C.L.; DeMers, M.S.; Delp, D.D.; Hicks, J.L.; Delp, S.L. Full-Body Musculoskeletal Model for Muscle-Driven
Simulation of Human Gait. IEEE Trans. Biomed. Eng. 2016, 63, 2068–2079. [CrossRef]
22. Lai, A.K.M.; Arnold, A.S.; Wakeling, J.M. Why Are Antagonist Muscles Co-Activated in My Simulation? A Musculoskeletal
Model for Analysing Human Locomotor Tasks. Ann. Biomed. Eng. 2017, 45, 2762–2774. [CrossRef]
23. Uhlrich, S.D.; Jackson, R.W.; Seth, A.; Kolesar, J.A.; Delp, S.L. Muscle Coordination Retraining Inspired by Musculoskeletal
Simulations Reduces Knee Contact Force. Sci. Rep. 2022, 12, 9842. [CrossRef]
24. Seth, A.; Hicks, J.L.; Uchida, T.K.; Habib, A.; Dembia, C.L.; Dunne, J.J.; Ong, C.F.; DeMers, M.S.; Rajagopal, A.; Millard, M.; et al.
OpenSim: Simulating Musculoskeletal Dynamics and Neuromuscular Control to Study Human and Animal Movement. PLoS
Comput. Biol. 2018, 14, e1006223. [CrossRef] [PubMed]
25. Ferraty, F.; Vieu, P. Nonparametric Functional Data Analysis; Springer Series in Statistics; Springer: New York, NY, USA, 2006;
ISBN 978-0-387-30369-7.
26. Edwards, W.B.; Derrick, T.R.; Hamill, J. Time Series Analysis in Biomechanics. In Handbook of Human Motion; Müller, B., Wolf, S.I.,
Brueggemann, G.-P., Deng, Z., McIntosh, A., Miller, F., Selbie, W.S., Eds.; Springer International Publishing: Cham, Switzerland,
2017; pp. 1–24, ISBN 978-3-319-30808-1.
27. Ullah, S.; Finch, C.F. Applications of Functional Data Analysis: A Systematic Review. BMC Med. Res. Methodol. 2013, 13, 43.
[CrossRef] [PubMed]
28. Balaban, B.; Tok, F. Gait Disturbances in Patients with Stroke. Phys. Med. Rehabil. 2014, 6, 635–642. [CrossRef] [PubMed]
29. Burpee, J.L.; Lewek, M.D. Biomechanical Gait Characteristics of Naturally Occurring Unsuccessful Foot Clearance during Swing
in Individuals with Chronic Stroke. Clin. Biomech. 2015, 30, 1102–1107. [CrossRef] [PubMed]
30. Nanhoe-Mahabier, W.; Snijders, A.H.; Delval, A.; Weerdesteyn, V.; Duysens, J.; Overeem, S.; Bloem, B.R. Walking Patterns in
Parkinson’s Disease with and without Freezing of Gait. Neuroscience 2011, 182, 217–224. [CrossRef]
31. Frenkel-Toledo, S.; Giladi, N.; Peretz, C.; Herman, T.; Gruendlinger, L.; Hausdorff, J.M. Treadmill Walking as an External
Pacemaker to Improve Gait Rhythm and Stability in Parkinson’s Disease. Mov. Disord. 2005, 20, 1109–1114. [CrossRef]
32. Nieuwboer, A.; Kwakkel, G.; Rochester, L.; Jones, D.; van Wegen, E.; Willems, A.M.; Chavret, F.; Hetherington, V.; Baker, K.; Lim,
I. Cueing Training in the Home Improves Gait-Related Mobility in Parkinson’s Disease: The RESCUE Trial. J. Neurol. Neurosurg.
Psychiatry 2007, 78, 134–140. [CrossRef]
33. Õunpuu, S.; Davis, R.B.; DeLuca, P.A. Joint Kinetics: Methods, Interpretation and Treatment Decision-Making in Children with
Cerebral Palsy and Myelomeningocele. Gait Posture 1996, 4, 62–78. [CrossRef]
34. Adde, L.; Helbostad, J.L.; Jensenius, A.R.; Taraldsen, G.; Grunewaldt, K.H.; Støen, R. Early Prediction of Cerebral Palsy by
Computer-Based Video Analysis of General Movements: A Feasibility Study. Dev. Med. Child. Neurol. 2010, 52, 773–778.
[CrossRef]
35. Mizuta, N.; Hasui, N.; Kai, T.; Inui, Y.; Sato, M.; Ohnishi, S.; Taguchi, J.; Nakatani, T. Characteristics of Limb Kinematics in the
Gait Disorders of Post-Stroke Patients. Sci. Rep. 2024, 14, 3082. [CrossRef]
36. Morris, M.E. Movement Disorders in People With Parkinson Disease: A Model for Physical Therapy. Phys. Ther. 2000, 80, 578–597.
[CrossRef] [PubMed]
37. Arnold, A.S.; Liu, M.Q.; Schwartz, M.H.; Ounpuu, S.; Delp, S.L. The Role of Estimating Muscle-Tendon Lengths and Velocities of
the Hamstrings in the Evaluation and Treatment of Crouch Gait. Gait Posture 2006, 23, 273–281. [CrossRef] [PubMed]
38. Goldberg, S.R.; Ounpuu, S.; Arnold, A.S.; Gage, J.R.; Delp, S.L. Kinematic and Kinetic Factors That Correlate with Improved Knee
Flexion Following Treatment for Stiff-Knee Gait. J. Biomech. 2006, 39, 689–698. [CrossRef] [PubMed]
39. Ziziene, J.; Daunoraviciene, K.; Juskeniene, G.; Raistenskis, J. Comparison of Kinematic Parameters of Children Gait Obtained by
Inverse and Direct Models. PLoS ONE 2022, 17, e0270423. [CrossRef]
40. Mundt, M.; Born, Z.; Goldacre, M.; Alderson, J. Estimating Ground Reaction Forces from Two-Dimensional Pose Data: A
Biomechanics-Based Comparison of AlphaPose, BlazePose, and OpenPose. Sensors 2023, 23, 78. [CrossRef]
Bioengineering 2024, 11, 911 18 of 18

41. Colyer, S.; Needham, L.; Evans, M.; Wade, L.; Cosker, D.; Cazzola, D.; McGuigan, P.; Bilzon, J. Estimation of Ground Reaction
Forces from Markerless Kinematics and Comparison Against Measured Force Plate Data. ISBS Proc. Arch. 2023, 41, 23.
42. Solanki, D.; Lahiri, U. Design of Instrumented Shoes for Gait Characterization: A Usability Study With Healthy and Post-Stroke
Hemiplegic Individuals. Front. Neurosci. 2018, 12, 459. [CrossRef]
43. Falisse, A.; Serrancolí, G.; Dembia, C.L.; Gillis, J.; De Groote, F. Algorithmic Differentiation Improves the Computational Efficiency
of OpenSim-Based Trajectory Optimization of Human Movement. PLoS ONE 2019, 14, e0217730. [CrossRef]
44. Andersson, J.; Åkesson, J.; Diehl, M. CasADi: A Symbolic Package for Automatic Differentiation and Optimal Control. In Recent
Advances in Algorithmic Differentiation; Springer: Berlin/Heidelberg, Germany, 2012; pp. 297–307.
45. Stenum, J.; Hsu, M.M.; Pantelyat, A.Y.; Roemmich, R.T. Clinical Gait Analysis Using Video-Based Pose Estimation: Multiple
Perspectives, Clinical Populations, and Measuring Change. PLoS Digit. Health 2024, 3, e0000467. [CrossRef]
46. Werling, K.; Bianco, N.A.; Raitor, M.; Stingel, J.; Hicks, J.L.; Collins, S.H.; Delp, S.L.; Liu, C.K. AddBiomechanics: Automating
Model Scaling, Inverse Kinematics, and Inverse Dynamics from Human Motion Data through Sequential Optimization. PLoS
ONE 2023, 18, e0295152. [CrossRef]
47. Needham, L.; Evans, M.; Wade, L.; Cosker, D.P.; McGuigan, M.P.; Bilzon, J.L.; Colyer, S.L. The Development and Evaluation of a
Fully Automated Markerless Motion Capture Workflow. J. Biomech. 2022, 144, 111338. [CrossRef] [PubMed]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.