A Novel Feature Selection and Classification Method of Alzheimer's Disease based on Multi-features in MRI

In this paper, we describe a novel machine learning method for classifying Alzheimer's disease (AD), Mild cognitive impairment (MCI) and Normal Control (NC) subjects based on structural MRI. We first extracted features from MRI scans, including cortical volumes, cortical thicknesses, subcortical volumes, and hippocampal subfields volumes. Then a new feature selection method combining the support vector machine-recursive feature elimination (SVM-RFE), maximal-relevance-minimal-redundancy (mRMR) and random forest (RF) was proposed to select the optimal subsets among all these features. Finally, the SVM classifier was used for AD/MCI/NC classification by 10-fold cross-validation. We applied the proposed method to the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, and the experimental results show a high degree of accuracy, sensitivity and specificity, which are superior to some other state-of-the-art approaches.