Structural MRI classification for Alzheimer's disease detection using deep belief network

Abstract

Early detection of Alzheimer's disease (AD) is the key of preventing, slowing, and stopping the disease. An early detection of AD can be performed by analyzing the neuro-imaging data. The magnetic resonance image (MRI) can be used as a modality of neuro-imaging data in order to detect AD. The MRI also have several advantages such as high-quality of spatial resolution, widely availability, adequate contrast and without requiring radioactive pharmaceutical injection during acquisition process. However, the main challenge of structural MRI data classification is the high dimensionality of the data. Therefore, this study proposes a classification method of AD based on structural modalities using Deep Belief Network (DBN) which is has power in term of predictive models. Support vector machine (SVM) has been used as a comparative classification model againts DBN. The result shows that this approach outperforms SVM and current method in previous study. The DBN achieves 0.9176, 0.9059 and 0.9296 in accuracy, sensitivity and specificity.