RVFL Classifier Based Ensemble Deep Learning for Early Diagnosis of Alzheimer's Disease

The prevalence of Alzheimer's Disease (AD) is increasing daily in elderly people, estimated to be 15 million by the year 2050. AD is an irreversible neurodegenerative disorder that may lead to the death of the affected person. Only early AD diagnosis at the Mild Cognitive Impairment (MCI) stage can help clinicians to convert MCI patients back to Cognitive Normal (CN) or slow down the progression of the disease. The advent of neuroimaging techniques like Magnetic Resonance Imaging (MRI) helps in observing the anatomical changes in the brain of MCI and AD patients with improved resolution. AD mainly affects the temporal lobe structure, hippocampus volume, and cerebral cortex, which are visible in MRI scans. In this paper, an ensemble of three planes of MRI is proposed using a deep learning model, and the extracted features are classified using Random Vector Functional Link (RVFL) neural networks. The experiments are done on the publicly available dataset, Alzheimer's Disease Neuroimaging Initiative (ADNI), to classify AD vs CN vs MCI. The performance of the proposed model is compared in terms of accuracy, specificity, sensitivity, and precision.