Early Prediction of Parkinson's Disease through Computerized Analysis of Handwriting

Abstract

Parkinson disease (PD) is a neurological disorder that influences movement of muscles causing disabled posture, rigidity and tremors. People suffering from PD face trouble in sleeping, walking and sitting. Generally, potential PD subjects are examined by an expert medical practitioner who employs the clinical symptoms like stiffness and slowing movements to detect the presence or absence of the disease, only after the disease had progressed considerably. Research on PD has revealed that analysis of handwriting and speech can serve as an effective early warning for Parkinson. With the advancements in image analysis and pattern classification, the manual analysis of these handwritten samples is being replaced with computerized analysis and automated prediction systems. This project presents a system that exploits online features of handwriting to predict PD in subjects. Features considered in our study include writing speed, pen pressure and pen-up/pen-down times. The features from PD patients and control subjects are used to train a support vector machine classifier. Evaluations on a benchmark database of online handwriting samples realized promising classification rates.