TELEREHABILITATION ASSESSMENT OF KNEE RANGE OF MOVEMENT; COMPARISON OF HUMAN POSE ESTIMATION TOOL, PHYSIOROM, VS 3D MOTION CAPTURE SYSTEM

Abstract

INTRODUCTION AND AIMS Telerehabilitation is increasingly acknowledged, though uptake remains limited by an inability to accurately undertake objective measurement. This study sought to determine the validity of a human pose estimation tool (PhysioROM, Coviu) assessed via two-dimensional (2D) video to measure active knee flexion and extension movement during common rehabilitation exercises, versus gold-standard three-dimensional (3D) motion capture. METHODS Fifty participants were recruited (aged 18 to 30), each performing three common knee rehabilitation exercises; sit-to-stand, seated and reclined active knee flexion. Movements were recorded using the 12-camera, 3D Vicon motion capture system simultaneously with 2D video simulating a home telehealth consultation. Joint kinematics were calculated using the Vicon Plug-in Gait model (3D Vicon) or the human pose estimation tool within the PhysioROM algorithm (2D PhysioROM). RESULTS Mean peak knee flexion was 3.1°, 9.9° and 24.8° (p<0.001) greater for 3D Vicon versus 2D PhysioROM (sit-to-stand, seated flexion and reclined flexion, respectively). For peak knee extension, Vicon 3D was 2.4° and 1.93° greater than 2D PhysioROM in sit-to-stand (p=0.020) and seated flexion (p< 0.001), whilst 3.6° less than PhysioROM for reclined knee flexion (p=0.030). Statistical parametric mapping revealed significant differences between Vicon 3D and PhysioROM for 71% and 42% of timepoints (p<0.001) for seated and reclined movements, clustered around full knee flexion. Significant differences were also found at an additional 37% (p<0.001) of timepoints clustered around knee extension for reclined postures. Overall mean absolute deviation was 8.7°, 7.6° and 16.5° for sit-to-stand, seated flexion and reclined flexion, respectively. CONCLUSION Validity of PhysioROM to accurately measure knee flexion was high, less so in extension, during sit-to stand movements in a seated posture. Measurement in a reclined position were less accurate. PhysioROM may provide an effective method of accurately and objectively quantifying knee movement during a telerehabilitation session using 2D video.