VTherapy: Deep Learning-based Hand Rehabilitation System

In this research, we introduce a virtual hand rehabilitation system, VTherapy, that amalgamates a deep learning model with a Flask web application, aimed at enhancing post-injury hand therapy at home. Our system leverages a deep learning architecture combining ResNeXt50's feature extraction with LSTM's sequential data analysis for precise hand gesture recognition. This methodology is augmented by the Dynamic Time Warping algorithm to accurately assess the similarity between the patient’s exercise video sequences and the reference exercise video. This approach not only democratizes access to effective hand rehabilitation by offering a cost-efficient, home-based solution but also promotes patient autonomy and engagement through interactive feedback mechanisms. The salient features of this system include an exercise performance tracker with goal setting, a real-time interactive rule-based chatbot trained using a neural network, and appointment scheduler to connect patients with physiotherapists. A dedicated SQLite database underpins the system by recording every exercise instance, facilitating feedback-driven progress tracking. The confluence of cutting-edge deep learning with user-friendly web technology presents a paradigm shift in rehabilitative care, potentially transforming the landscape of home-based physiotherapy for patients recovering from hand injuries.