IoMT and explainable AI-enabled wearable system for classifying tremor and motor patterns in Parkinson’s disease

The prevalence of Parkinson’s disease (PD) represents a significant global health concern due to its debilitating motor symptoms, particularly tremors. This pilot study presents an interpretable Internet of Medical Things (IoMT)-based monitoring system for analyzing hand tremors in individuals with PD. The proposed framework integrates a 6-axis inertial sensor for motion tracking, a low-power wireless microcontroller for data acquisition and connectivity, and a gradient boosting-based machine learning model (LightGBM) for tremor classification. During the data acquisition phase, signals from a triaxial accelerometer and gyroscope are sampled at 66.67 Hz, covering the characteristic 4–6 Hz Parkinsonian tremor frequency band. A RESTful API transmits the acquired data to a server over Wi-Fi, where it is stored in a relational database. Signal processing includes noise reduction, temporal segmentation, and frequency-domain feature extraction using the fast Fourier transform (FFT). The LightGBM classifier categorizes motor activity into Parkinsonian tremor, voluntary movement, or absence of tremor. Additionally, a web-based user interface supports real-time signal visualization and structured clinical data entry. As a feasibility study, the system demonstrates strong classification performance, achieving 95.64% accuracy and an F1-score of 0.95 on a dataset comprising 10,314 samples from nine participants (four with clinically diagnosed PD and five healthy controls). Although evaluated under controlled laboratory conditions, these results support the potential of the proposed system as a low-cost, interpretable tool for objective tremor assessment and continuous monitoring of Parkinson’s disease symptoms.