Deep Sequential Learning for Adhesive Capsulitis Classification Using Wearable IMU Sensors

In the era of the Industrial Internet of Things (IIoT), the digital transformation of healthcare services is emerging as a pivotal domain for intelligent systems. Musculoskeletal disorders, such as Adhesive Capsulitis (frozen shoulder), require consistent monitoring and objective functional assessment. However, traditional clinical evaluations remain subjective and labor-intensive. This study proposes an IoT-based intelligent rehabilitation assessment system that leverages wearable IMU sensors and deep sequential learning models to automate functional evaluation. We designed a data acquisition framework using wearable Inertial Measurement Units (IMUs) to capture kinematic patterns during rehabilitation exercises. To process the sequential motion data, we employed a Deep Sequential Learning approach using Long Short-Term Memory (LSTM) networks augmented with an Attention mechanism. We compared its performance with a conventional machine learning model, XGBoost, focusing on class imbalance and generalization capability. The experimental results on a dataset of 208 participants demonstrate that the proposed architecture effectively captures temporal dependencies in movement trajectories, achieving a classification accuracy of 98% and an ROC-AUC of 0.99. This study validates the feasibility of deploying intelligent deep learning models within IoT-enabled healthcare environments for automated patient monitoring.