Concurrent validity and test reliability of the deep learning markerless motion capture system during the overhead...

Area of Science:

• Biomechanics
• Human Movement Analysis
• Digital Health Technology

Background:

• Marker-based optical motion capture is crucial for understanding human movement but is time-consuming and labor-intensive.
• Existing systems require highly trained operators and complex data acquisition processes.
• There is a need for more accessible and efficient motion capture solutions in ecological settings.

Purpose of the Study:

• To develop and validate a deep learning-based 3D markerless motion capture system named "Ergo".
• To assess the concurrent validity and test-retest reliability of the Ergo system.
• To evaluate the system's performance against a standard marker-based motion capture system for joint kinematics.

Main Methods:

• Developed the "Ergo" system utilizing deep learning for markerless 3D motion capture.
• Collected overhead squat movement data using both Ergo and a gold standard marker-based system.
• Analyzed whole-body joint kinematics, including time series and peak joint angles.

Main Results:

• Ergo demonstrated excellent agreement for time series joint angles (ICC = 0.88-0.99).
• Peak joint angles showed excellent agreement (ICC = 0.75-1.0) compared to the gold standard.
• High test-retest reliability was observed for Ergo measurements (ICC = 0.92-0.99).

Conclusions:

• The deep learning-based markerless Ergo system achieves comparable accuracy and reliability to gold standard marker-based systems.
• Ergo offers a rapid (5-min data collection/processing) and usable solution for motion capture.
• The system is highly accessible for diverse users and ecological digital healthcare environments.