Parkinson's disease (PD) and scans without evidence of dopaminergic deficit (SWEDD) are two distinct neurological disorders that require different therapeutic approaches; therefore it's critical to classify the two disorders. The neuroimaging technology based on dMRI provided connectivity information and voxel features that can make it possible for researchers to analyze SWEDD and PD differences. In this work, a novel method of ReliefF-SVM-based dMRI analysis was presented to study the potential relations between PD and SWEDD. Some sensorimotor connections were found group-wise differences, and SVM was suggested to successfully classify PD and SWEDD. These results indicate that our method using connectivity information and voxel features may provide a new strategy for disease analysis with small sample data.
Keywords: Machine learning; Parkinson’s disease; SWEDD; Tractography; dMRI.
Copyright © 2018. Published by Elsevier B.V.