Poster R7 - Accurate Markerless Motion Capture Kinematic Comparison During Unilateral and Bilateral Landings

Introduction: The crucial need to capture joint kinematics and conduct movement analysis outside a laboratory setting is often accomplished through markerless motion capture. This technology can be used widely to assist in rehabilitation, sport performance, and clinical applications. Theia3D (Theia Markerless Inc., Kingston, ON) is a markerless motion capture system that provides valid and reliable data when compared with marker-based motion capture1,2,3. However, there is a desire to find lower cost and more portable technology, as Theia3D is expensive and requires extensive set-up. Several markerless motion capture platforms have been developed, many of which are open source and require a minimum of 2 cameras. There have been no prior studies validating these systems. OpenCap (Stanford, USA) is a new, low-cost, and extremely portable system that has been independently validated for lower extremity kinematics against marker-based capture4. This system boasts a 25x decrease in usage time while assessing and processing biomechanical metrics4. These advancements could aid in the incorporation of this technology into clinical settings. However, prior to implementation, there is a need to determine differences between various markerless motion capture systems with Theia3D serving as the reference or gold-standard. Currently, many markerless applications are not being utilized in clinical settings due to the insufficient accuracy of their joint angle results5. The goal of this work was to assess lower extremity joint angle differences between OpenCap and Theia3D and document potential limitations of the two systems. We hypothesize that no difference will exist in lower extremity joint angles between the two systems.

Materials and

Methods: 14 young healthy adults were used in this validation study. Testing was completed using Theia3D and OpenCap to collect video data simultaneously, at 60 fps and 1080p during several dynamic movements. Participants were instructed to perform 7 repetitions each of a single leg hop (SLH), drop vertical jump (DVJ) beginning on a 30cm box, and a countermovement jump (CMJ). A custom MATLAB code was used to compare the OpenCap and Theia kinematic data. The right sagittal plane hip, knee, ankle, and frontal plane hip angles were compared within each participant across each trial for every task. These outcome measures were selected as they are the only lower extremity kinematics that are generated by OpenCap without using OpenSim (Stanford, USA) to calculate the other angles using musculoskeletal modeling. A linear mixed effects model (α=0.05) was used to determine differences in joint angles between OpenCap and Theia3D using JMP software (SAS Institute INC., Cary, NC) with α=0.05 being used to determine statistical differences.

Results, Conclusions, and Discussions: Peak ankle dorsiflexion was different between Theia3D and OpenCap for all tasks (CMJ: p< 0.001; DVJ: p< 0.001; SLH: p< 0.001). OpenCap reported a 7.2°±5.1 greater angle across all tasks. Peak ankle plantarflexion was different for the CMJ (p < 0.001) and DVJ (p < 0.001). OpenCap, reported a 2.8°±4.7 greater angle during the CMJ and DVJ. No differences were found in peak knee flexion between the two systems across all tasks. Peak knee extension was different for all tasks (CMJ: p=0.003; DVJ: p< 0.001; SLH: p< 0.001), with OpenCap reporting 3.1°±3.8 greater angles. Peak hip flexion was different for the CMJ (p < 0.001) and DVJ (p < 0.001), with OpenCap reporting 5.3 ±10.6 smaller angles. Differences were found in peak hip flexion during SLH landing (p < 0.001), with OpenCap reporting 3.8°±8.6 smaller angles. No difference in peak hip flexion was found during SLH take-off. Peak hip extension was different for all tasks (CMJ: p< 0.001; DVJ: p< 0.001; SLH: p=0.003), with OpenCap reporting 4.2°±4.9 greater angles. Peak hip abduction was different during the SLH (p < 0.001). OpenCap reported a 4.3°±5.9 greater angle during the SLH. Peak hip adduction was different during all tasks (CMJ: p< 0.001; DVJ: p=0.043; SLH: p< 0.001), with OpenCap reporting 1.9°±3.3 greater angles.
With differences in the performance between the two systems, there are several metrics for which OpenCap can provide valid (within 2-5°)6 results during unilateral and bilateral landing tasks. While OpenCap can provide some valid lower extremity joint kinematics, the system will not function if more than one participant moves into the camera field of view during data collection, which is a major limitation in the clinical environment. Future studies should include curve analysis comparison across the camera systems, and more movement tasks to identify differences in time-series and discrete metrics.
Within the 2-5° range, OpenCap reliably measures peak ankle plantarflexion, knee flexion/extension, hip extension and abduction/adduction during the CMJ, DVJ, and SLH. However, OpenCap may measure greater peak hip flexion and ankle dorsiflexion across these tasks.

Acknowledgements (Optional): Support for this study was provided by a grant from the National 

Institutes of Health: R01 AR078811-03.