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: Theia3D version v2022.1.0.2309, as stated in the paper.
Study Overview
Design
A within-subject, concurrent comparison of marker-based motion capture (reference), Theia3D markerless motion capture, APDM Opal inertial measurement units, and a HumanTrak RGBD camera system during lunges, overhead squats, countermovement jumps, and cutting maneuvers, alongside a descriptive evaluation of each system's setup, data collection, and data analysis time requirements.
Participants
Thirty adults (18 women, 12 men; mean age 25.4 years; mean height 1.71 m; mean weight 71.6 kg) with no known illness or physical condition that would have restricted their participation in physical activity.
Data Collection
Each participant completed, in order, nine lunges per leg, nine overhead squats, nine double-leg countermovement jumps, and ten 45-degree cutting maneuvers (five to each side) at a self-selected running pace, for 46 individual trials total. Data were collected simultaneously using ten Vicon infrared cameras (marker-based reference, 128 Hz), eight synchronized Theia3D OptiTrack Prime Color cameras (128 Hz), fifteen APDM Opal inertial measurement units secured across the body, and a single Azure Kinect-based HumanTrak RGBD camera positioned in the frontal plane.
Analysis
Lower-body and, for cutting maneuvers, upper-body joint angles were compared between each alternative system and the marker-based reference using bias, root-mean-square error (RMSE), precision, maximum absolute error, and intraclass correlation coefficients (ICC). A joint angle comparison was considered acceptably valid if it met both an RMSE of 5 degrees or less and an ICC of 0.75 or greater. Because the HumanTrak system only measured discrete joint angles at specific movement events rather than continuous time-series data, its comparisons used average bias with 95% confidence intervals instead. Usability was evaluated descriptively by timing the setup, data collection, and data analysis stages for each system across the first five participants and documenting technical issues encountered.
Key Findings
- Theia3D showed the highest overall agreement with the marker-based reference of the three alternative systems tested, though error varied widely by joint and plane of motion. Sagittal-plane RMSEs for Theia3D ranged from 3.20 to 15.66 degrees, frontal-plane RMSEs ranged from 2.12 to 9.14 degrees, and transverse-plane RMSEs ranged from 3.16 to 56.61 degrees, generally the lowest range of the three alternative systems in each plane.
- Only two joint angle measures met the study's full acceptable-validity threshold (RMSE ≤ 5° and ICC ≥ 0.75) for Theia3D across every movement tested: sagittal-plane ankle angle during the overhead squat (RMSE = 3.20°, ICC = 0.850) and frontal-plane trunk angle during the cutting maneuver (RMSE = 3.70°, ICC = 0.750). Neither the IMU system nor the RGBD system met this full threshold for any joint angle measure across all movements, though Theia3D and the IMU system each had several additional measures that satisfied at least one of the two individual criteria.
- For the knee, hip, and ankle in the sagittal plane, Theia3D and the IMU system both showed their strongest agreement with the marker-based reference, while neither system showed acceptable agreement in the frontal or transverse planes for these same joints. Theia3D's sagittal-plane ICCs ranged from 0.92 to 0.97 for the knee, 0.61 to 0.92 for the hip, and 0.64 to 0.93 for the ankle, each somewhat higher than the corresponding range for the IMU system.
- The RGBD system showed poor agreement with the marker-based reference across every measure evaluated (ICC below 0.5 throughout), with its highest agreement reaching only an ICC of 0.26 for maximum sagittal-plane hip angle after landing in the countermovement jump, leading the authors to conclude it was not currently a viable option for fast or accurate motion capture.
- Theia3D required substantially less active, hands-on staff time per participant than marker-based motion capture, though it carried the largest data storage burden of the systems compared. Once camera setup was complete, data collection and analysis for Theia3D required roughly 1 minute of active staff time and about 11 minutes of passive computer processing per trial, compared with roughly 72 minutes of active participant and staff time for the marker-based system per participant, while Theia3D's raw video data occupied over 4 terabytes across the study, substantially more storage than the marker-based or IMU systems required.
What This Means for Movement Science Researchers and Lab Methodologists
For teams choosing between motion capture technologies for dynamic, high-demand movements, this study offers a useful, side-by-side answer: Theia3D markerless motion capture provided the closest overall agreement with marker-based motion capture of the systems tested, particularly for sagittal-plane lower-limb kinematics, while requiring less ongoing hands-on staff time once cameras were set up, though not every joint angle and plane reached the study's acceptable-validity bar, and transverse-plane and some frontal-plane measures remain areas of caution. The authors' broader framing, that system selection should be based on the specific priorities of accuracy, efficiency, portability, and cost for a given application, is a useful way to weigh Theia3D's strengths (overall validity, moderate technical demands once set up) against its tradeoffs (data storage requirements, and plane-specific gaps in measurement accuracy) relative to the other systems evaluated.
The Role of Theia3D in Movement Analysis
Theia3D was one of three alternative motion capture systems directly evaluated against a marker-based reference in this study, tested across four demanding, dynamic movements using an eight-camera OptiTrack Prime Color setup. Of the three alternative systems compared, Theia3D achieved the highest overall agreement with the marker-based reference, and the study's detailed usability data additionally documented Theia3D's practical time and technical demands, the only pose-estimation issue it encountered was a failure to generate a valid pose during the flight phase of three countermovement jump trials for a single participant, alongside its validity findings.
Study Limitations and Future Directions
The authors note that this study concurrently compared four motion capture systems, and the marker-based reference system itself carries inherent measurement error from soft-tissue motion at the skin, which makes it difficult to fully attribute the source of any disagreement to a specific system rather than to limitations of the marker-based reference. The cross-correlation method used to synchronize time-series data between the three separate computers and software packages involved may have introduced some alignment error, although the authors visually confirmed synchronization quality for all trials. There is also some possibility that collecting data from four systems simultaneously, including reflective markers and infrared light sources from multiple systems operating in the same space, could have affected the measurements of other systems, such as the RGBD camera's depth-sensing infrared light. Participants in this study were healthy, active individuals, so these findings may not generalize to populations with distinct or atypical movement patterns. All data were collected at 128 Hz to align with the marker-based and Theia3D systems with the IMU system's default sample rate, and the authors note that collecting at a higher sample rate may affect system validity differently for faster movements, while also increasing data analysis and storage requirements. The authors identify continuing to evaluate these systems for additional movement types, including upper-extremity and activities-of-daily-living tasks, and extending validation work to new populations, as clear next steps.
Citation: Edwards, N. A., Caccese, J. B., Tracy, R. E., Hagen, J., Quatman-Yates, C. C., & Oñate, J. (2025). The validity and usability of markerless motion capture and inertial measurement units for quantifying dynamic movements. Medicine & Science in Sports & Exercise, 57(3), 641–655. https://doi.org/10.1249/MSS.0000000000003579
Link: Read the full study
Which Motion Capture System Fits Your Dynamic Movement Research?
This study directly compares Theia3D markerless motion capture against inertial measurement units and an RGBD camera system for validity and real-world usability during demanding, dynamic movements, finding Theia3D delivered the strongest overall agreement with marker-based motion capture among the alternatives tested. If your work involves choosing a motion capture system for jumps, squats, lunges, or change-of-direction movements and you want to weigh these accuracy and usability tradeoffs for your own application, talk to our team about what Theia3D could offer your research.