Some Insight Into Alzheimer’s Disease Progression

Abstract

Alzheimer’s disease is a neurodegenerative disease that affects a multitude of people globally. It usually affects people of 60 yrs and older causing changes in the anatomy of the brain.Subjects diagnosed with Alzheimer’s disease often gets to live for 5-10 yrs.This makes early and accurate detection of the disease of paramount importance not only for the victim but to better understand disease progression in subjects. Over the years with the advancement of machine learning and deep learning a number of studies have come up to perform disease classification basis different biomarkers which acts as indicators for disease progression. This paper presents a multi-modal study that compares performances of machine learning and deep learning models on 2 sets of inputs namely MRI and cognitive scores. It considers dataset from ADNI (Alzheimer Disease Neuroimaging Initiative) which is a longitudinal multi centre study designed to develop clinical, imaging, genetic and biochemical biomarkers for the early detection and tracking of Alzheimer’s disease (AD) and performs multi-class classification of subjects into 3 groups of CN(Normal Cognition), MCI(Mild Cognitive Impairment) and AD(Alzheimer Disease). For the deep learning model classification using MRI the study proposes use of a modified 2D-CNN that works with MRI scans. Contrary to Deep Convolutional Neural Networks that outputs better accuracy at the cost of higher execution times 2D-CNN performs faster at the cost of accuracy. In addition the study also considers a mix of both forms as input i.e.features extracted from 2D-CNN and cognitive scores to classify subjects basis machine learning models. This hybrid input captures not only brain anatomical changes but also symptoms that manifest.