Human Movement & Mobility Research
Not specified

Differences in centre of mass measurements between markerless and marker-based motion capture systems during balance and mobility assessments in individuals with chronic and sub-acute stroke

medRxiv
•
2026
•
Preprint (not peer-reviewed)
Theia3D version:
v2023.1.0.3161
More Articles

Study at a glance

In 17 adults with chronic or sub-acute stroke performing balance and mobility tasks, Theia3D showed strong agreement with marker-based motion capture for centre of mass position in the anterior-posterior and medial-lateral directions, with larger differences in the vertical direction that narrowed considerably once the markerless model's segment structure was matched to the marker-based reference.

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 v2023.1.0.3161, as stated in the paper. The paper does not name a specific camera brand/model for the markerless system. This paper is a preprint posted to medRxiv and has not yet completed peer review.

Study Overview

Design

A concurrent validity study comparing centre of mass position between a marker-based reference system and a markerless (Theia3D) system, using two different markerless body-model configurations, during balance and mobility tasks in individuals with chronic and sub-acute stroke.

Participants

Seventeen adults with stroke (12 men, 5 women; 14 chronic, 3 sub-acute), drawn by convenience from participants in a separate reactive-balance-training study, able to stand unassisted for 30 seconds and walk 10 meters without a walking aid. One participant wore an ankle-foot orthosis and one participant wore an arm sling during testing.

Data Collection

Testing took place in a dedicated balance-research laboratory, with participants wearing a safety harness attached to an overhead gantry throughout testing. A marker-based system (67 reflective markers) and an 8-camera markerless system processed through Theia3D recorded simultaneously at 100 Hz. Each participant completed sit-to-stand, rising onto their toes, a backward reactive step, quiet standing, and six unperturbed walking trials, drawn from a session that also included additional balance tasks as part of the larger parent study.

Analysis

Three body models were built: a marker-based reference model (11 segments) in Visual3D, a "default" markerless model (17 segments) representing Theia3D's standard output, and a "fit" markerless model (11 segments) constructed to match the marker-based model's segment count and anthropometry. Centre of mass position in each model was filtered and compared between systems in three directions, anterior-posterior, medial-lateral, and vertical (superior-inferior), using the coefficient of determination (R²) and root-mean-square difference (RMSD) as the primary agreement measures.

Key Findings

  • Agreement between the markerless and marker-based systems was strong to excellent for COM position in the anterior-posterior and medial-lateral directions across all five tasks and both markerless models (anterior-posterior R² = 0.949-0.999, RMSD 1.0-2.9 cm; medial-lateral R² = 0.782-0.991, RMSD 0.4-0.7 cm).
  • Agreement in the vertical direction was comparatively weaker and more variable, particularly during quiet standing, where the "default" markerless model showed its lowest correlation of any task-direction combination (R² = 0.422, RMSD 5.86 cm), a pattern the authors attribute to the small magnitude of vertical COM movement during quiet standing relative to measurement noise, consistent with prior markerless validation findings during low-movement tasks.
  • Matching the markerless model's segment structure to the marker-based reference substantially narrowed the vertical-direction gap: the "fit" model's vertical RMSD was roughly five times smaller than the "default" model's across most tasks (for example, 1.03 cm versus 5.86 cm during quiet standing), while anterior-posterior and medial-lateral agreement remained similarly strong between the two markerless model configurations.
  • The participant wearing an arm sling showed visibly larger differences from the marker-based reference across most tasks than other participants, while the participant wearing an ankle-foot orthosis did not show a similarly consistent pattern, though the authors note they cannot determine from this data whether the sling itself altered movement or simply interfered with marker placement on the marker-based system.
  • The marker set used for the marker-based reference system was not specifically modified to match the markerless system's default segment definitions, since the markerless system was introduced partway through the parent study, which the authors note may have limited the agreement achievable with the "default" model and was part of the motivation for also building the anthropometrically matched "fit" model.

What This Means for Movement Science Researchers and Balance Assessment Teams

For researchers and teams working on balance and mobility assessment after stroke, this study offers an encouraging, specific result: Theia3D tracked centre of mass position with strong to excellent agreement against marker-based motion capture in the directions most central to balance control, anterior-posterior and medial-lateral sway, across a range of standing, transfer, and walking tasks in a population with stroke-related mobility differences. The vertical-direction findings add a useful, concrete nuance: agreement in this direction was weaker when using Theia3D's standard body-model output, particularly during quiet standing, but improved considerably once the markerless model's segment structure was more closely matched to the marker-based reference, suggesting that model configuration choices meaningfully affect how well markerless vertical COM tracking performs and warrant consideration when accuracy in that direction matters for a given application.

The Role of Theia3D in Movement Analysis

Theia3D was the markerless motion capture system evaluated in this study, compared directly against a marker-based reference system for centre of mass tracking, a whole-body balance measure rather than an individual joint angle, during functional balance and mobility tasks in adults with stroke. The study's comparison of two different Theia3D body-model configurations against the same marker-based reference also offers a useful, more general illustration of how segment-definition choices in the markerless model itself can influence agreement with marker-based measurement, independent of any difference in underlying camera-based tracking accuracy.

Study Limitations and Future Directions

The authors note that their sample, while living with stroke, was drawn from participants who met minimum standing and walking ability criteria for the parent balance-training study, making the group relatively high-functioning; they identify evaluating markerless COM accuracy in people with greater balance or mobility differences as an important next step. They also note that because the marker-based system's marker set was not purpose-built to match the markerless model, the "default" model comparison may understate the agreement achievable with a more closely matched setup, which the "fit" model results help illustrate but do not fully resolve. The authors identify further investigation into how assistive devices such as orthoses and slings affect markerless-versus-marker-based agreement, ideally in larger samples designed specifically to isolate device effects from marker-placement interference, as a useful direction for future research.

Citation: Majoni, N., Inness, E. L., Jagroop, D., Danells, C. J., & Mansfield, A. (2026). Differences in centre of mass measurements between markerless and marker-based motion capture systems during balance and mobility assessments in individuals with chronic and sub-acute stroke. medRxiv. https://doi.org/10.64898/2026.02.18.26346541

Link: Read the full preprint

Can Markerless Capture Track Centre of Mass During Balance Tasks After Stroke?

This preprint study of 17 adults with stroke finds that Theia3D tracks centre of mass position with strong to excellent agreement against marker-based motion capture during balance and mobility tasks, with vertical-direction agreement improving considerably when the markerless body model is matched to the marker-based reference. If your work involves balance or mobility assessment and you want to understand where markerless centre-of-mass tracking currently stands, talk to our team about applying Theia3D to your own balance research.

Recent Posts