"INNOVATIVE WEB-BASED GONIOMETRIC ANALYSIS SYSTEM POWERED BY NEURAL NETWORKS"

This article presents the development of an innovative web-based system for real-time measurement of elbow and wrist joint angles using computer vision technologies. Emphasizes the relevance of applying digital solutions in physical rehabilitation, particularly for patients with reduced mobility due to stroke — a condition that has become more frequent and affects younger populations in Ukraine as a result of war-related stress. The paper analyzes traditional goniometry techniques and modern alternatives such as mechanical, electronic, and photogrammetric systems. It also provides an overview of OpenCV and Google’s MediaPipe framework, with a focus on the Holistic module, which combines hand, body, and face tracking based on neural networks. The proposed system integrates Hand Landmarker and Pose Landmarker to precisely extract joint coordinates and calculate angles using the arctan2 method. The web application, developed with JavaScript, HTML, and Canvas, uses CDN integration with MediaPipe to visualize Landmarker in real time, offering immediate feedback during movement exercises. An algorithm is introduced to count physical repetitions (e.g., elbow flexion-extension) by detecting motion phases. The arctangent approach is highlighted for its numerical stability and suitability for realtime applications. Additional features include interface customization, selection of joints for analysis, and integration potential into gamified rehabilitation systems. The presented solution serves as a promising platform for digital transformation in clinical and educational physiotherapy contexts