AD ALERT Classification of Alzheimer disease with Traditional and Deep Network Models

Abstract

The most common cause of dementia is Alzheimer’s disease(AD). Alzheimer’s disease has a slow rate of advancement, which gives patients the chance to receive early treatment through regular testing. However, due to their high price and restricted availability, current clinical diagnostic imaging techniques do not satisfy the specific needs for screening methods. AD ALERT aims to address the problem by automating the detection of the Alzheimer’s with machine learning techniques. In this paper, the Magnetic Resonance Image (MRI) data is extracted, and feature selection based on random forest is used to select top 30% important and useful features among the total features for the analysis. The deep network models were proposed to classify the patients to AD using these selected features based on random forest. The traditional classification techniques namely K Nearest Neighbour (K-NN), decision tree, Stochastic Gradient Descent (SGD) , and Support Vector Classifier (SVC) were also implemented using the same feature selection model to compare the performance of the deep network models. The deep network models proved to be outperforming the considered model with an accuracy of 0.98%.