Hybrid Feature Vector Generation for Alzheimer’s Disease Diagnosis Using MRI Images

In the advent of computer-aided diagnosis (CAD), analysis of medical images with the help of various feature extraction techniques is gaining a lot of interest. Nowadays, a lot of research is done in the analysis of neuro-images, particularly the MRI modality to study the structural changes or atrophy of the brain due to various neurological disorders. Alzheimer's disease (AD) being one such illness, affects a vast proportion of elderly populations worldwide. The current diagnosis of AD which is done manually by a clinician using neuro-images can sometimes be inaccurate. Although image processing and machine learning have facilitated feature extraction and diagnosis of AD respectively, there are still many challenges that exist in the current systems. This work proposes a system that will diagnose AD using MRI images by generating a hybrid feature vector (HFV). The HFV is created from fusing features obtained from different techniques such as volumetric analysis of grey matter (GM) and white matter (WM), surface-based cortical thickness and tissue volumetric study of regions of interest (ROI) acquired from Voxel Based Morphometry (VBM) analysis. Machine learning techniques are later used to classify patients (AD) from normal controls (NC). The experimentation of the proposed system was carried out using the T1-weighted MRI scans of 100 subjects (50 AD and 50 NC) obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. The maximum accuracy achieved by the proposed system is 93% using the SVM classifier which validates the promising performance of the HFV in AD diagnosis.