Microsoft Kinect can distinguish differences in over-ground gait between older persons with and without Parkinson's disease

Abstract

Gait patterns differ between healthy elders and those with Parkinson's disease (PD). A simple, low-cost clinical tool that can evaluate kinematic differences between these populations would be invaluable diagnostically; since gait analysis in a clinical setting is impractical due to cost and technical expertise. This study investigated the between group differences between the Kinect and a 3D movement analysis system (BTS) and reported validity and reliability of the Kinect v2 sensor for gait analysis. Nineteen subjects participated, eleven without (C) and eight with PD (PD). Outcome measures included spatiotemporal parameters and kinematics. Ankle range of motion for C was significantly less during ankle swing compared to PD (p=0.04) for the Kinect. Both systems showed significant differences for stride length (BTS (C 1.24±0.16, PD=1.01±0.17, p=0.009), Kinect (C=1.24±0.17, PD=1.00±0.18, p=0.009)), gait velocity (BTS (C=1.06±0.14, PD=0.83±0.15, p=0.01), Kinect (C=1.06±0.15, PD=0.83±0.16, p=0.01)), and swing velocity (BTS (C=2.50±0.27, PD=2.12±0.36, p=0.02), Kinect (C=2.32±0.25, PD=1.95±0.31, p=0.01)) between groups. Agreement (Range_ICC =0.93-0.99) and consistency (Range_ICC =0.94-0.99) were excellent between systems for stride length, stance duration, swing duration, gait velocity, and swing velocity. The Kinect v2 can was sensitive enough to detect between group differences and consistently produced results similar to the BTS system.

Keywords: Gait analysis; Kinect; Motion capture; Parkinson's disease.