Parkinsonian Assessment Tool

Description

This project develops an objective assessment for determining whether a person has Parkinson's disease. We propose a method which analyzes the correlation between joint movements, since Parkinsonian patients often have trouble coordinating different joints in a movement. Two features are extracted from a simple pointing movement: auto-correlation coefficients of single joint movement and cross-correlation coefficients of multiple joint movement. Using Least Squares Support Vector Machine, we achieve a 96% classification rate. Experiments verify the proposed method with 395 pointing movements by 14 PD patients and 476 pointing movements by controls.