Application of machine learning models to classify Parkinson disease patients using accelerometer

Abstract

Parkinson’s disease (PD) greatly affects mobility, emphasizing the critical need for early detection to optimize treatment outcomes. This study explores the application of advanced machine learning techniques to analyse data collected from specialized motion tracking sensors known as accelerometers. These sensors monitor individual’s movements and symptoms associated with Parkinson’s disease, with a primary focus on comprehending the unique movement patterns and related symptoms prevalent in those affected by this condition. This research leverages the capabilities of machine learning, employing diverse algorithms such as support vector machines, neural networks, and random forest. Through an integrated approach, the study aims to construct a robust ensemble model capable of synthesizing insights from these techniques. The dataset utilized originates from the University of Zaragoza Hospital, encompassing a diverse range of participants, including individuals both diagnosed and undiagnosed with Parkinson’s disease. This diversity ensures a comprehensive exploration of various movement patterns and symptoms among heterogeneous individuals. The primary objective revolves around precise differentiation between individuals affected by Parkinson’s disease and those who are not. To achieve this, the study adopts an ensemble model strategically designed to reconcile conflicting predictions from individual classifiers. This methodological approach seeks consensus by aggregating multiple classifier opinions, thereby minimizing uncertainties arising from divergent predictions. The outcomes of this study are particularly significant, with the ensemble model demonstrating superior performance over traditional machine learning models. The best-performing ensemble model, as identified through comparative analysis, achieved an impressive accuracy of 65.69%, specificity of 73.5%, and precision of 71.17%. These metrics underscore the potential of ensemble classifiers in enhancing diagnostic precision, thereby contributing to the early detection and continuous monitoring of parkinson disease.