Hand Landmark Detection Approach for the Personalized Soft Wearable Robotic Hand on Post-Stroke Hand Rehabilitation

We present an innovative approach for hand landmark detection using a personalized soft wearable robot hand, facilitating task-oriented training in stroke rehabilitation. Our paper introduces the Ring-reinforced Soft Actuator (RSA), accommodating different hand sizes and enabling flexion and extension movements. Individually controlled finger actuators assist stroke patients in various grasping tasks. We also propose an automated method for landmark detection and joint measurement, enabling efficient development of customfit wearable robot hands. Additionally, we developed a wireless task board for seamless interaction with the soft robotic hand and placement of training objects. Evaluation with four stroke subjects demonstrated significant improvements in upper limb functions (FMA), hand motor abilities (ARAT, BBT), and maximum grip/pinch strength after 20 sessions of task-oriented training. These improvements persisted for at least three months post-training, highlighting the system's potential to enhance stroke rehabilitation and promote hand motor recovery. The personalized design allows for frequent hand practice and easy integration into regular rehabilitation routines.