Classification of Alzheimer Disease Using Feature Segmentation and 3D CNN

Alzheimer Disease is a type of neurodegeneration that is irreversible and leads to memory loss, generally seen in older people. Several machine learning and imaging techniques have been proposed to identify such type of neurodegeneration which uses structural MRI. In this study, we investigate a deep learning model for AD diagnosis by using morphological features, specifically, gray matter volume and Jacobian determinant of MRI to train the 3D convolutional neural network model. The focus of this study is to make a less complex and computationally efficient solution by training of 3D CNN using extracted image features. The experiments were carried on ADNI1 dataset. The obtained result shows the accuracy of 96% for AD classification. Moreover, the sensitivity of the proposed method reveals that the model is less error prone in normal conditions.