Video-based Clinical Gait Analysis in Parkinson's Disease: A Novel Approach Using Frontal Plane Videos and Machine...

Area of Science:

• Neurology
• Biomedical Engineering
• Computer Science

Background:

• Neurological conditions like Parkinson's disease (PD) significantly impair gait.
• Traditional instrumented gait analysis is accurate but costly and inaccessible for widespread clinical use.
• There is a need for accessible, cost-effective gait analysis methods.

Purpose of the Study:

• To develop and validate a machine learning-based method for clinical gait analysis using readily available video devices.
• To assess the feasibility of detecting Parkinson's disease (PD) gait impairments from video recordings.
• To provide a user-friendly and cost-effective solution for remote gait assessment.

Main Methods:

• Utilized open-source body pose (MediaPipe) and metric depth (ZoeDepth) estimators for video analysis.
• Calculated spatiotemporal parameters (STP) of gait from frontal plane videos.
• Validated the method against traditional marker-based motion capture (MoCap) using healthy controls and people with PD (PWP).

Main Results:

• The video-based method demonstrated high agreement and correlation with marker-based motion capture (MoCap).
• Results align with findings from high-end instrumental gait analysis techniques.
• The methodology successfully supported the detection of Parkinson's disease (PD) from gait videos.

Conclusions:

• The proposed method offers an innovative, user-friendly, and cost-effective solution for clinical gait analysis.
• Automated video-based gait analysis can enable telemedicine evaluations for movement decline in individuals with gait impairments.
• This approach has the potential for early PD detection, remote monitoring, and improved management of fall risk.