EFFECTS OF A DEEP LEARNING-BASED SMARTPHONE APPLICATION ON SHOULDER ABDUCTION KINEMATICS AND BRAIN ACTIVATION IN ADHESIVE CAPSULITIS: A RANDOMIZED CONTROLLED TRIAL

Abstract

Patients with adhesive capsulitis (AC) demonstrate limited shoulder movement, often accompanied by pain. Common treatment methods include pain medication, and continuous passive movement (CPM). However, it is sometimes difficult to improve the reduction of pain and movement using a CPM intervention because the patient’s interest is diminished. In this study, we developed an innovative deep learning-based smartphone application (Funrehab exercise game (FEG)) to provide accurate kinematics movement and motivation as well as high-intensity and repetitive movements using deep learning. We compared the effects of CPM and FEG on brain activity and shoulder range of motion in patients with AC. Sixteen patients (males, [Formula: see text]; females, [Formula: see text]; mean age, [Formula: see text] years) with acute AC were randomized into either CPM group or FEG group 4 days/week for 2 weeks. The outcome measures were shoulder abduction kinematics movement and electroencephalography (EEG) brain activity (bilateral prefrontal, bilateral sensorimotor cortex, and somatosensory association cortex) during the intervention. The analysis of variance (ANOVA) test was performed at [Formula: see text], and the analysis demonstrated that FEG showed superior effects on shoulder abduction kinematics and brain [Formula: see text] and [Formula: see text]-wave activations compared to CPM. Our results provide a novel and promising clinical evidence that FEG can more effectively improve neurophysiological EEG data and shoulder abduction movements than CPM in patients with AC.