Classification of Alzheimer and MCI Phenotypes on MRI Data Using SVM

Abstract

Alzheimer disease (AD) is a common form of dementia affecting people older than the age of 65. Moreover, AD is commonly diagnosed by behavioural paradormants, cognitive tests, and is followed by brain scans. Computer Aided Diagnosis (CAD), applies medical imaging and machine learning algorithms, to aid in the early diagnosis of Alzheimer’s severity and advancement from prodromal stages i.e. Mild Cognitive Impairment (MCI) to diagnosed Alzheimer’s disease. In this work, SVM (support vector machine) is used for dementia stage classification. Anatomical structures of the brain were obtained from FreeSurfer’s processing of structural Magnetic Resonance Imaging (MRI) data and is utilized for as features for SVM. To be more precise, the system is processed using T1-weighted brain MRI datasets consisting of: 150 mild cognitive impairment (MCI) patients, 80 AD patients and 130 normal controls (NC) obtained from Alzheimer Disease Neuroimaging Initiative (ADNI) database. The volumes of brain structures (hippocampus, medial temporal lobe, whole brain, ventricular, cortical grey matter, entorhinal cortex and fusiform) are employed as biomarkers for multi-class classification of AD, MCI, and NC.