Study at a glance

In 21 to 26 adults running on a treadmill, Theia3D markerless motion capture agreed closely with a marker-based system across running kinematics and kinetics, with differences between the two systems consistently larger than any difference caused by switching between form-fitting and loose athletic clothing.

Summary

Disclaimer: The following summary details independent academic research. Theia3D is intended for research, sports performance and educational use only. It is not cleared as a medical device and is not intended to be used for the diagnosis, treatment, mitigation or prevention of any disease, injury or medical condition.

Editor's note: Theia3D version v2022.1.0.2309, as stated in the paper.

Study Overview

Design

A within-subject comparison of treadmill running biomechanics with two parts: a concurrent comparison between marker-based and markerless motion capture recorded simultaneously in form-fitting clothing, and an asynchronous comparison between markerless motion capture recorded in form-fitting clothing versus each participant's own athletic clothing on a separate trial.

Participants

Thirty adults were recruited. After excluding participants with inconsistent camera frame counts, 26 (12 women, 14 men; mean age 23.5) were included in the kinematic comparison. A further exclusion of participants with erroneous center-of-pressure data left 21 (10 women, 11 men; mean age 23.3) in the kinetic comparison.

Data Collection

Participants ran at a self-selected pace on an instrumented treadmill under two conditions: form-fitting marker-based motion-capture clothing (with simultaneous marker-based motion capture, markerless motion capture, and ground reaction force data), and their own self-selected athletic clothing (with markerless motion capture and ground reaction force data only). Both camera systems recorded at 85 Hz; ground reaction forces were recorded at 425 Hz. Running speed was held constant across both conditions without being disclosed to participants.

Analysis

Joint center positions, segment angles, joint angles, and joint moments were calculated in Visual3D and time-normalized to one full running gait cycle. Differences between the concurrent marker-based and markerless data, and between the two markerless clothing conditions, were summarized across participants as root-mean-square differences (RMSDs). Two alternative marker-based model definitions were also tested to examine how sensitive the comparison was to how body segments and landmarks were defined.

Key Findings

  • Joint center positions differed by less than 3 centimeters on average between the concurrent marker-based and markerless data. Mean RMSDs were 2.12 cm at the ankle, 2.15 cm at the knee, and 2.96 cm at the hip.
  • Sagittal-plane joint angles, the movements most central to running mechanics, showed the closest agreement between systems. Mean RMSDs were 5.2 degrees or less at the ankle, knee, and hip, comparable to previously published minimal detectable change values for marker-based treadmill running kinematics.
  • Transverse-plane rotations and certain frontal-plane angles showed larger differences between systems than sagittal-plane angles did. Concurrent joint angle RMSDs reached 9.8 degrees for hip internal/external rotation and 9.4 degrees for knee internal/external rotation, noticeably higher than the sagittal-plane values.
  • Differences caused by switching clothing conditions were consistently smaller than differences between the two motion-capture systems. Asynchronous RMSDs between the two markerless clothing conditions were smaller than the concurrent marker-based-versus-markerless RMSDs for nearly every joint angle and moment measured, with hip flexion as the one exception, where the clothing-condition difference was slightly larger.
  • How a marker-based model defines body segments and landmarks materially affects how much it appears to agree with markerless data. Testing two alternative marker-based model definitions showed that a meaningful share of the apparent disagreement between marker-based and markerless systems could be shifted simply by changing those definitions, independent of any true difference in measured movement.

What This Means for Running Biomechanics Researchers and Movement Science Teams

For teams considering markerless capture for running research, this study's central practical finding is reassuring: a runner's choice of clothing had only a modest effect on Theia3D's measurements, consistently smaller than the baseline difference between marker-based and markerless systems. That supports using markerless capture to study running in a person's own gear rather than requiring specialized clothing, which could make longitudinal or field-based running research more accessible. The study's methodological finding is just as important for anyone running this kind of comparison themselves: because marker-based model definitions can shift the apparent level of agreement with markerless data, the authors recommend using Theia3D's own automated coordinate system definitions to standardize comparisons across multiple sites or studies, rather than treating any single marker-based model as a fixed ground truth.

The Role of Theia3D in Movement Analysis

Theia3D was the markerless system evaluated throughout this study, measuring joint kinematics and kinetics from treadmill running trials under two different clothing conditions without requiring the specialized attire that marker-based capture needs. By comparing Theia3D's own asynchronous clothing-condition data against its concurrent agreement with a marker-based reference, the study isolated how much of the total measurement difference was attributable to clothing versus the inherent difference between capture systems.

Study Limitations and Future Directions

The authors note that their treadmill running speeds were self-selected and consequently lower than typical overground running speeds, consistent with known differences in perceived effort between treadmill and overground running; they anticipate that a higher running speed recorded at a higher camera frame rate would produce comparable results to those reported here. The 85 Hz camera frame rate used was lower than is typical for running studies, a limitation of the hardware used to maximize image resolution and synchronization. The order of the two clothing conditions was divided evenly across participants but not randomized, though the authors note this would not affect the comparison of simultaneous marker-based and markerless data, and the absence of clothing-order effects in the results suggests this was not an issue. The precise segment coordinate system definitions used internally by Theia3D are not publicly available, so the marker-based definitions used for comparison were designed to approximate rather than exactly replicate them. The authors disclose that two of the study's co-authors are employed by Theia Markerless Inc. and a third serves as the company's CEO, and note that the underlying data were collected prior to those authors' employment with the company.

Citation: Kanko, R. M., Outerleys, J. B., Laende, E. K., Selbie, W. S., & Deluzio, K. J. (2024). Comparison of concurrent and asynchronous running kinematics and kinetics from marker-based and markerless motion capture under varying clothing conditions. Journal of Applied Biomechanics, 40(2), 129–137. https://doi.org/10.1123/jab.2023-0069 This study builds on the same research group's earlier comparison of marker-based and markerless motion capture during walking (Kanko, R. M., Laende, E. K., Davis, E. M., Selbie, W. S., & Deluzio, K. J. (2021). Concurrent assessment of gait kinematics using marker-based and markerless motion capture. Journal of Biomechanics, 127, 110665. https://doi.org/10.1016/j.jbiomech.2021.110665).

Link: Read the full study

Does What You Wear Change What Markerless Capture Measures?

This study suggests that for running specifically, clothing has a smaller effect on markerless measurements than the baseline difference between motion-capture systems, a reassuring result for anyone hoping to study running biomechanics in the field rather than the lab. If your work involves running or field-based movement research and you want to understand how markerless capture performs across real-world clothing conditions, talk to our team about what that could mean for your own data collection.

Recent Posts