Non-hand-worn, load-free VR hand rehabilitation system assisted by deep learning based on ionic hydrogel

Many individuals suffer from stroke, osteoarthritis, or accidental hand injuries, making hand rehabilitation greatly significant. The current hand rehabilitation therapy requires repetitive task-oriented hand exercises, relying on exoskeleton mechanical gloves integrated with different sensors and actuators. However, these conventional mechanical gloves require wearing heavy mechanical components that need weight-bearing and increase hand burden. Additionally, these devices are usually structurally complex, complicated to operate, and require specialized medical institutions. Here, a Virtual Reality (VR) hand rehabilitation system is developed by integrating deep-learning-assisted electromyography (EMG) recognition and VR human-machine interfaces (HMIs). By applying a wet-adhesive, self-healable, and conductive ionic hydrogel electrode array assisted by deep learning, the system can realize 14 Jebsen hand rehabilitation gestures recognition with an accuracy of 97.9%. The recognized gestures further communicate with the VR platform for real-time interaction in a virtual scenario to accomplish VR hand rehabilitation. Compared with present hand rehabilitation devices, the proposed system enables patients to perform immersive hand exercises in real-life scenarios without the need for hand-worn weights, and offers rehabilitation training without time and location limitations. This system could bring great breakthroughs for the development of a load-free hand rehabilitation system available in home-based therapy.