Study at a glance

In 19 team-sport athletes performing straight running and four change-of-direction movements at three speeds, Theia3D-derived lower-limb joint moments showed substantial, systematic differences from marker-based motion capture across every joint, movement plane, direction, and speed tested, exceeding a predefined 10% agreement threshold in every case and leading the authors to conclude the two systems' joint kinetics should not be used interchangeably for these movements.

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 2023-1-0-3161, as stated in the paper. This is one of the more directly critical papers about Theia3D in this collection with respect to joint kinetics specifically, and the summary and findings below are presented without softening, consistent with our practice of flagging author-identified limitations clearly.

Study Overview

Design

A concurrent comparison study evaluating the agreement of lower-extremity joint kinetics between marker-based and markerless (Theia3D) motion capture during straight running and four change-of-direction cutting movements, across multiple cutting angles and approach speeds, in team-sport athletes.

Participants

Nineteen running-based, trauma-free team-sport athletes (6 women, 13 men; mean age 23.7 years) who volunteered and provided informed consent.

Data Collection

Participants performed a straight run and four change-of-direction cuts (at 45, 90, 135, and 180 degrees) at three individually self-selected speeds ("slow jogging," "fast jogging," and "sprinting"), each repeated three times, with the final cutting step performed on an embedded force plate. Marker-based data were captured using 21 infrared cameras with 22 reflective markers, and markerless data were captured simultaneously using eight high-speed video cameras processed through Theia3D, with hip, knee, and ankle joint moments calculated from each system's kinematic data combined with the same force-plate measurements.

Analysis

Agreement between systems was assessed for peak joint moments using an extended Bland-Altman approach with linear mixed-effects models, and across the entire stance phase using bootstrapped functional prediction bands. Both methods expressed between-system differences in absolute terms and relative to typical marker-based moment magnitudes, using a predefined threshold of 10% relative difference as a descriptive reference for acceptable agreement. Agreement was compared across joints, movement planes, the five movement directions, and the three speed categories.

Key Findings

  • Across every joint, movement plane, direction, and speed combination tested, relative differences between markerless and marker-based joint moments exceeded the predefined 10% agreement threshold, with relative limits of agreement ranging from 58% to 153% and relative prediction-band areas ranging from 48% to 252%.
  • The ankle showed the smallest absolute differences between systems but, because ankle moments are naturally smaller in magnitude, the largest relative differences (123%); the hip showed the opposite pattern, with the largest absolute differences but the smallest relative differences (98%), reflecting its typically larger moment values.
  • Differences between systems were smallest for internal and external rotation moments and largest for flexion and extension moments across joints, with the hip showing the largest absolute differences of any joint overall.
  • Agreement did not show a consistent pattern across the five movement directions, with the smallest relative limits of agreement at 135-degree cuts and the largest at 90-degree cuts; the researchers had expected agreement to worsen with larger cutting angles, based on prior straight-running research, but this pattern did not hold for change-of-direction movements specifically.
  • Agreement was only modestly and inconsistently better for slower movements: peak-moment agreement was somewhat better for slow versus fast straight running, but this speed-related pattern was not consistently present across the four change-of-direction movements.
  • As a specific, practically relevant example, the authors note that knee abduction moment, a measure frequently used in team-sport research as a surrogate for lateral knee loading relevant to anterior cruciate ligament mechanics, showed a limit of agreement (±0.82 Nm/kg) of similar magnitude to previously reported typical knee abduction moment values themselves (0.45 to 1.44 Nm/kg), meaning between-system differences could meaningfully affect how this specific measure is interpreted.

What This Means for Movement Science Researchers and Team-Sport Performance Assessment Teams

For researchers and teams working on team-sport performance and lower-limb loading assessment, this study offers a clear, specific caution: while markerless motion capture has shown increasingly strong agreement with marker-based systems for many kinematic measures, this study indicates that markerless and marker-based joint kinetics should not currently be treated as directly interchangeable for change-of-direction movements specifically, across any joint, plane, direction, or speed tested. The authors are careful to note this does not rule out markerless motion capture's independent usefulness, for example as a standalone tool for tracking how an individual athlete's own joint kinetics change over time, provided that the system's own session-to-session consistency and smallest detectable change are well matched to the specific application. Teams whose work depends on precise joint moment values during cutting or pivoting movements, particularly where a specific measure like knee abduction moment is being used as a loading indicator, should weigh this finding carefully and treat markerless- and marker-based-derived kinetic values from this kind of task as distinct, non-substitutable measurements rather than interchangeable ones.

The Role of Theia3D in Movement Analysis

Theia3D was the markerless motion capture system directly evaluated against marker-based motion capture in this study, specifically for its agreement in estimating lower-extremity joint kinetics during team sport-specific change-of-direction movements, a task category the authors note has received comparatively little attention in markerless validation research to date relative to walking and straight-line running. The study's systematic comparison across five movement directions, three speeds, three joints, and three movement planes offers a notably thorough mapping of where and how markerless- and marker-based-derived joint kinetics agree and diverge for this specific, practically important movement category.

Study Limitations and Future Directions

The authors note that their markerless data were sampled at 85 Hz, the highest resolution available for their video cameras, while their marker-based data were downsampled to match this frequency for comparison; this downsampling may have affected signal content, particularly for the fast change-of-direction movements examined, though any resulting error would likely be conservative given the richer original marker-based sampling rate. Only one marker-based marker set and model were used for comparison, so findings may not generalize to other marker configurations, biomechanical models, or camera setups, and the authors note that prior research has shown marker set and model choice alone can meaningfully influence calculated joint kinetics. They note that identifying the underlying sources of the markerless-marker-based discrepancies observed, including differences in joint-center definitions, segment center-of-mass locations, and inertial parameters between the two systems' underlying models, will require further detailed analysis. The authors also note that a software update could alter these findings, meaning the results are specific to the Theia3D version used in this study. They identify aligning joint-center definitions, segment coordinate systems, and filtering procedures between markerless and marker-based pipelines, incorporating training datasets with more high-speed multiplanar movements, and establishing application-specific (rather than generic) agreement thresholds for change-of-direction tasks as important next steps for this line of research.

Citation: Helwig, J., Vanrenterghem, J., Koska, D., Gollhofer, A., & Willwacher, S. (2026). Agreement of markerless and marker-based joint kinetics in team sport-specific change-of-direction movements. Scientific Reports, 16, 31298. https://doi.org/10.1038/s41598-026-72910-3

Link: Read the full study

Do Markerless and Marker-Based Systems Agree on Joint Kinetics During Change-of-Direction Movements?

This study of 19 team-sport athletes finds that Theia3D-derived lower-limb joint moments differ substantially from marker-based motion capture across every joint, movement plane, direction, and speed tested during change-of-direction cutting tasks, exceeding a predefined agreement threshold in every case. If your work involves team-sport performance research and you want to understand where markerless capture currently stands for joint kinetics specifically, talk to our team about applying Theia3D to your own research.

Recent Posts