Study at a glance
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: This abstract directly compares two Theia3D software versions, v2022.1.0.2309 and v2023.1.0.3161, against one another rather than against a marker-based reference system.
Study Overview
Design
A direct comparison of trunk and lower-body gait curves derived from the same set of video recordings, processed separately using two successive Theia3D software versions, in typically developing children.
Participants
Fourteen typically developing children (mean age 9.7 years).
Data Collection
Each participant completed five barefoot walks at a self-selected speed, recorded using ten Qualisys Miqus cameras at 1080p resolution and 100 Hz. The same video data were processed twice, once using Theia3D version 2022.1.0.2309 and once using version 2023.1.0.3161, with both outputs exported to Visual3D to generate gait reports for comparison.
Analysis
Trunk and lower-body gait curves for matching gait cycles from both software versions were compared using the average root-mean-square error (RMSE) and the linear fit method, which reports the strength of linear agreement (R²), any offset or shift between curves (a0), and the relative variation of one curve compared to the other (a1). Identical waveforms between versions would be expected to produce an RMSE of 0 degrees, an R² of 1, an offset of 0 degrees, and a slope of 1.
Key Findings
- The greatest difference between the two Theia3D software versions was observed at the hip joint in the transverse plane (RMSE = 4.7°, R² = 0.2, offset = -4.5°, slope = 0.2), indicating comparatively weaker agreement for this specific joint and plane between the two releases.
- Angular differences between versions elsewhere ranged from 0.8° at the thorax to 5.6° at the ankle, both in the transverse plane, with the ankle transverse-plane comparison representing the single largest RMSE value observed across all measures.
- Strong agreement (R² = 0.7-1.0) was observed between versions for sagittal-plane hip, knee, and ankle angles; frontal-plane thorax, pelvis, and hip angles; and transverse-plane thorax, pelvis, and ankle angles, indicating that the large majority of trunk and lower-body gait curves were well preserved between the two software versions.
- Weaker agreement was also observed for the sagittal-plane pelvis specifically (R² = 0.2), despite a comparatively small RMSE of 1.9°, illustrating that R² and RMSD can highlight different aspects of agreement and that both metrics are useful when comparing software versions.
- The authors attribute the largest observed differences, at the hip and ankle in the transverse plane, to specific updates made to how the software estimates the hip joint center and ankle pose between these two releases, rather than to a general degradation in measurement consistency across the software update.
What This Means for Movement Science Researchers and Longitudinal Assessment Programs
For researchers and teams running longitudinal gait research or ongoing assessment programs using Theia3D, this abstract offers a concrete, practical reference point: most trunk and lower-body gait measures showed strong agreement between these two successive software versions, but the hip and ankle in the transverse plane specifically showed differences large enough to be worth accounting for when comparing data collected across a software update. The authors' practical recommendations, to monitor software development closely and to consider re-analyzing existing data when version changes occur, are a useful standard practice for any program using markerless motion capture over an extended period, and recording software version as part of standard reporting is a simple step that supports this kind of comparison going forward.
The Role of Theia3D in Movement Analysis
Theia3D was the markerless motion capture software evaluated in this abstract, with two successive versions of the software compared directly against one another using the same underlying video recordings. Rather than evaluating Theia3D's accuracy against a marker-based reference system, this abstract offers a different and complementary kind of evidence: a direct look at how much measured outcomes can be expected to shift between two specific software releases, information that is particularly useful for research programs and assessment services planning to track gait data over time as the underlying software continues to be updated.
Study Limitations and Future Directions
As a brief conference abstract, this summary does not include an extended discussion of study limitations or statistical significance testing beyond the reported agreement metrics. The sample size of 14 children is relatively small, and the comparison was limited to two specific, successive Theia3D versions rather than a broader range of releases, so these findings may not generalize to every version transition. The authors' own stated recommendations, to monitor ongoing software development and to re-analyze existing data where appropriate, point toward an implicit need for this kind of version-comparison evidence to be revisited as Theia3D continues to be updated in the future.
Citation: Kearney, J., Greaves, H., Robinson, M., Barton, G., Gibbon, K., O'Brien, T., Wright, D., Pinzone, O., & Foster, R. (2024). Changes in trunk and lower body gait kinematics in children following a Theia3D update. Gait & Posture, 113, 113-114 (ESMAC Abstracts 2024). https://doi.org/10.1016/j.gaitpost.2024.07.127
Link: Read the full abstract
How Much Does a Theia3D Software Update Change Measured Gait Kinematics?
This conference abstract directly compares two successive Theia3D software versions in the same group of children, finding strong agreement for most trunk and lower-body gait measures alongside specific, explainable differences at the hip and ankle in the transverse plane. If your work involves longitudinal gait research or an ongoing assessment program using Theia3D, talk to our team about best practices for tracking data consistency across software versions.