Hierarchical Feature Extraction for Early Alzheimer’s Disease Diagnosis

Abstract

Mild cognitive impairment (MCI) is the early stage of Alzheimer’s disease (AD). In this article, we propose a novel voxel-based hierarchical feature extraction (VHFE) method for the early AD diagnosis. First, we parcellate the whole brain into 90 regions of interests (ROIs) based on an Automated Anatomical Labeling (AAL) template. To split the uninformative data, we select the informative voxels in each ROI with a baseline of their values and arrange them into a vector. Then, the first stage features are selected based on the correlation of the voxels between different groups. Next, the brain feature maps of each subjects made up of the fetched voxels is fed into a convolutional neural network (CNN) to learn the deep hidden features. Finally, to validate the effectiveness of the proposed method, we test it with the subset of the Alzheimer’s Disease Neuroimaging (ADNI) database. The testing results demonstrate that the proposed method is robust with promising performance in comparison with the state-of-the-art methods.