Automated pose estimation for cerebral palsy gait metrics: a deep learning approach to clinical rehabilitation

Gait analysis is widely used for diagnosing and treating mobility restrictions in cerebral palsy (CP) patients. Gait analysis cycle still relies on costly motion capture systems or human observation. These approaches require a clinical setting and manual labor, making it hard to access and inflexible. The case study focuses on the application of computer vision techniques for lower body gait analysis in diplegia using deep learning and machine learning models. In this study, BlazePose is used to record video and capture image frames of children with diplegia cerebral palsy, using real-time pose approximation. The system outlines primary pose measurements of knee flexion, hip flexion, ankle dorsiflexion, and R1 and R2 hamstring responses from extracted body landmarks using the BlazePose model. Reporting of progress is captured through a web interface built into the platform, which makes it easy to monitor patient progress and record maintenance. In brief, this research study emphasizes the significance of simple and quick gait analysis and its potential to standardise the use of computer vision techniques over heavy equipment used in traditional gait analysis cycles.