An Efficient Deep Neural Network Binary Classifier for Alzheimer’s Disease Classification

Abstract

In recent research, deep neural networks have better classification results in the medical research fields. In this paper, a deep neural network with fully connected layers is designed to perform binary classification. Three different types of activation functions are used for the hidden layers. After performing k-folds validation with different activation function combinations, a model with the best performance is used. We used feature features extracted from the ADNI image for classification. To determine the best model, an experiment is performed for the classification of two groups: Alzheimer’s Disease (AD) and Cognitively Normal (CN). The proposed DNN with the best validation accuracy score obtained 85.19%, 76.93%, and 72.73% accuracy on the test data for AD vs. CN, Mild Cognitive Impairment (MCI) vs. CN, and AD vs. MCI classifications, respectively. This accuracy score is higher in comparison with other traditional machine learning algorithms.