Classification of Alzheimer's disease based on several features extracted from MRI T1-weighted brain images

Abstract

Alzheimer’s disease (AD) diagnosis at an early stage plays a significant role in reducing its symptoms and decelerating cognitive deterioration. Hence the use of computer-aided systems for early and accurate AD diagnosis is critical. The proposed diagnostic tool depends on classifying features extracted from brain Magnetic Resonance Imaging (MRI). These Features must accurately capture main AD-related variations of the anatomical brain structures, such as hippocampi region atrophy, lateral ventricle enlargement, cortical thickness, brain volume, etc. In this work, T1-weighted structural MRIs were employed for extracting these AD-related features. The images resulting from MRI scans are interpreted to high-intensity visible features, making preprocessing and segmentation less complex. This work has proposed a software pipeline consisting of several preprocessing steps, a segmentation method for segmenting brain tissues, and Convolutional Neural Networks (CNN) for Regions of Interest (ROIs) Parcellation that is AD-related. Features extracted from these segmented tissues and ROIs are utilized for the final AD classification using a Support Vector Machine (SVM) classifier. The results show that the proposed approach has reached 89.1% accuracy in the binary classification of AD vs. CN (Cognitively Normal), Demonstrating promising classification performance.