Convolutional Neural Network Based Prediction of Conversion from Mild Cognitive Impairment to Alzheimer's Disease: A Technique using Hippocampus Extracted from MRI

Abstract

Alzheimer's disease (AD) is an irreversible neurodegenerative disorder. Mild Cognitive Impairment (MCI) is a prodromal stage of AD and its identification is very crucial for early treatment. MCI to AD conversion is of imperative concern in current Alzheimer's research. In this study, we have investigated the conversion from MCI to AD using different types of features. The impact of structural changes in entire brain tissues captured through MRI, genetics, neuropsychological assessment scores and their combination are investigated. Computational cost can be significantly reduced by examining only the hippocampi region, atrophy of which is visible in the earliest stages of the disease. We proposed a CNN based deep learning approach for the prediction of conversion from MCI to AD using above mentioned features. Highest accuracy is achieved when left hippocampus is used as a region of interest (ROI). The proposed technique outperforms the other state of the art methods, while maintaining a low computational cost. The main contribution of the research lies in the fact that only a single slice based small region of MRI is used resulting in an outstanding performance. The accuracy, sensitivity and specificity achieved are 94%, 92% and 96% respectively.

Keywords

• artificial neural networks
• computer aided diagnosis
• image analysis
• image classification
• pattern recognition