Application of Video-Analysis Program Based on MediaPipe Pose for Diagnosis of Gait Disorders in Outpatient Setting: Two Clinical Cases

BACKGROUND: In recent years, alternative methods based on video recording and machine learning (ML) have been actively developed. These methods allow for objective assessment of movement parameters using a single device, significantly reducing time and cost. However, for the differential diagnosis of normal and pathological conditions, clinical validation of such approaches is required. AIM: To demonstrate the potentials of the developed gait video analysis program based on the MediaPipe Pose platform for quantitative assessment and comparison of spatiotemporal gait parameters in a healthy volunteer and a female patient with osteoarthritis (OA) of the knee joints. METHODS: The study was conducted on a stand (a treadmill, a GoPro 4K/60fps camera). The program based on MediaPipe Pose identified 33 key body points with calculation of knee joint angles, step timing characteristics, symmetry coefficients, and variability. Two clinical cases are presented. CONCLUSION: A video analysis program based on MediaPipe Pose allows for quantitative differentiation between pathological and normal walking pattern. Changes identified in a patient with gonarthrosis are consistent with the clinical picture and can serve as objective markers of disorders. The method holds promise for clinical application as an accessible screening tool. Further work to improve the program will focus on studying flexion angles in the lower extremity joints relative to gait events, examining their interrelationships to obtain more accurate results.