Multi-Class Classifier in Parkinson’s Disease Using an Evolutionary Multi-Objective Optimization Algorithm

In this contribution, a novel methodology for multi-class classification in the field of Parkinson’s disease is proposed. The structured two phases. first phase, most relevant volumes interest (VOI) brain are selected by means an evolutionary multi-objective optimization (MOE) algorithm. Each these VOIs subjected to volumetric feature extraction using Three-Dimensional Discrete Wavelet Transform (3D-DWT). When applying 3D-DWT, high number coefficients obtained, requiring use selection/reduction algorithms find features. method used contribution based on Mutual Redundancy (MI) and Minimum Maximum Relevance (mRMR) PCA. To optimize VOI selection, group 550 MRI was 5 classes: PD, SWEDD, Prodromal, GeneCohort Normal. Once Pareto Front solutions obtained (with varying degrees complexity, reflected VOIs), tested second phase. order analyze SVM classifier accuracy, test set 367 used. obtains results classification, presenting several with different levels complexity precision (Pareto solutions), reaching result 97% as highest data.