An Approach Using Deep Neural Networks for the Binary Classification of Mild Cognitive Impairment in Alzheimer's Disease: Early and Late Stages

Mild Cognitive Impairment (MCI) refers to the intermediate stage between normal brain function and the more severe decline caused by dementia. Alzheimer's Disease (AD) is a significant cause of dementia. Early detection of MCI can be beneficial for detecting individuals who exhibit the initial signs of AD. The AD Neuroimaging Initiative (ADNI) utilized Positron Emission Tomography (PET) images to diagnose MCI. MCI can be classified into two distinct categories: Early MCI (EMCI) and Late MCI (LMCI). In this study, a Deep Learning (DL) method was deployed with a limited number of PET images for binary classification of the two subcategories of MCI in AD patients. A total of 377 patients participated in this study, including 222 patients diagnosed with EMCI and 155 patients diagnosed with LMCI. The results indicated that the proposed Convolutional Neural Network (CNN) model reached a validation accuracy of 99.26%, a training accuracy of 98.68%, and a testing accuracy of 99.26%. The model has exceptional classification capabilities, highlighting its ability to accurately distinguish between the EMCI and LMCI classes with minimal misclassifications.