Verification of validity of gait analysis systems during treadmill walking and running using human pose tracking...

Area of Science:

• Biomechanics
• Human Motion Analysis
• Computer Vision

Background:

• OpenPose is a human tracking algorithm capable of detecting joint points and measuring angles.
• The validity of OpenPose for gait analysis has not been previously established.

Purpose of the Study:

• To assess the validity of OpenPose for gait analysis by comparing its measurements to a gold-standard system.
• To determine the accuracy of OpenPose in measuring pelvic and lower limb joint angles during walking and running.

Main Methods:

• Twenty-four healthy participants underwent gait analysis (walking and running) on a treadmill.
• Simultaneous data collection using VICON motion capture and digital cameras processed by OpenPose.
• Statistical validation included linear regression, intraclass correlation coefficients (ICCs), and Bland-Altman analysis.

Main Results:

• OpenPose demonstrated high accuracy (large coefficients of determination, good to excellent ICCs) for lower limb range of motion in the sagittal plane.
• Measurements of hip, knee, and ankle angles in the sagittal plane showed good agreement with VICON.
• Poorer agreement and significant biases were observed for pelvic and hip parameters in the frontal plane.

Conclusions:

• OpenPose-based motion analysis is a valid method for measuring lower limb range of motion in the sagittal plane during gait.
• Markerless systems like OpenPose provide economical and convenient alternatives to traditional marker-based motion capture.
• Further research is needed to improve accuracy for frontal plane parameters.