Weighted Random Support Vector Machine Clusters Analysis of Resting-State fMRI in Mild Cognitive Impairment.

摘要: The identification of abnormal cognitive decline at an early stage becomes increasingly significant conundrum to physicians and is major interest in the studies mild impairment (MCI). Support vector machine (SVM) as a high-dimensional pattern classification technique widely employed neuroimaging research. However, application single SVM classifier may be difficult achieve excellent performance because small-sample size noise imaging data. To address this issue, we propose novel method weighted random support cluster (WRSVMC) which multiple SVMs were built different weights given corresponding with performances. We evaluated our algorithm on resting state functional magnetic resonance (RS-fMRI) data 93 MCI patients 105 healthy controls (HC) from Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort. maximum accuracy by WRSVMC 87.67%, demonstrating diagnostic power. Furthermore, most discriminative brain areas have been found out follows: gyrus rectus (REC.L), precentral (PreCG.R), olfactory cortex (OLF.L), middle occipital (MOG.R). These findings paper provide new perspective for clinical diagnosis MCI.