Concurrent assessment of gait kinematics using marker-based and markerless motion capture

Area of Science:

• Biomechanics
• Human Movement Analysis
• Motion Capture Technology

Background:

• Marker-based optical motion capture is standard for quantifying human movement but is time-intensive and requires expert operation.
• Markerless motion capture systems present a practical alternative for kinematic data acquisition.

Purpose of the Study:

• To compare the accuracy of a deep learning-based markerless motion capture system against a conventional marker-based system for human gait kinematics.
• To evaluate the suitability of markerless systems for biomechanical research and clinical applications.

Main Methods:

• Simultaneous data collection of human gait from 30 healthy adults using both markerless (deep learning, 8 cameras) and marker-based (7 infrared cameras) systems.
• Comparison of joint center positions and lower limb segment/joint angles between the two systems.

Main Results:

• Markerless and marker-based systems showed high agreement in joint center positions (average RMSD < 2.5 cm, hip 3.6 cm).
• Global segment pose estimates were highly similar (RMSD < 5.5° for most angles).
• Similar patterns of lower limb joint angles (flexion/extension, ab/adduction, inversion/eversion) were captured by both systems.

Conclusions:

• The deep learning-based markerless motion capture system is a viable and accurate alternative to marker-based systems for human gait analysis.
• Markerless systems offer practical advantages, making them suitable for applications prioritizing efficient data collection in biomechanics.