Neural network-based shoulder instability diagnosis modelling for robot-assisted rehabilitation systems

Abstract

Many researchers are anticipating that robotic systems will contribute to compensating for the shortage of providing therapy for age-related injuries. Therefore, any prospective approach to robot-assisted therapy has to offer systems that can practically and objectively evaluate a patient's physical functions and apply evidence-based rehabilitation protocols. One of the most common disorders, among the general population particularly seniors, 40–60 years of age, is the frozen shoulder. Clinically, frozen shoulder can be diagnosed based on two shoulder functions: instability and cooperativeness of the shoulder joint. The purpose of the present study is to introduce a shoulder instability diagnosis model using artificial neural networks (ANN), which can be applicable using robotic systems. Training, validation, and testing of the neural network were achieved using the force exerted by a subject measured during predesigned clinical examinations. The proposed method has the ability to produce shoulder joint instability evaluations equivalent to those clinically obtained by therapists.