Early detection and monitoring of axial motor disorders in Parkinson’s disease using an AI-based approach

Early detection and continuous monitoring of neurological disorders remain challenging due to the limited sensitivity and subjectivity of conventional clinical assessments. In this frame, axial motor impairment in Parkinson’s disease plays a crucial role in disease progression, being contemporarily particularly difficult to quantify using only standard rating scales.In this study, we propose an artificial intelligence (AI)-based analytical framework for the objective monitoring and severity discrimination of axial motor symptoms in Parkinson’s disease. The framework is designed to operate on structured datasets of surface electromyography and was validated on a curated database comprising multichannel recordings from bilateral axial muscle groups acquired during a standardized standing task. Signals were pre-processed, including artefact removal through Independent Component Analysis, and a set of time- and frequency-domain features was extracted and combined into bilateral asymmetry indices. Multivariate data fusion was performed using Principal Component Analysis and Canonical Correlation Analysis to derive a compact composite representation of axial neuromuscular impairment.Results showed that univariate feature correlations were insufficient to capture clinically relevant information, whereas the proposed multivariate representation revealed meaningful associations with clinical reference measures. Furthermore, a supervised classification analysis using a Support Vector Machine achieved up to 70% accuracy in discriminating between mild and severe axial impairment for specific thoracic muscle groups.Overall, the proposed AI-based analytical framework demonstrates the potential of multivariate physiological biomarkers to support early detection and continuous monitoring of axial motor dysfunction in PD, providing a scalable and objective tool for neurological assessment.