Early Mild Cognitive Impairment Detection using a Hybrid Model

Convolutional Neural Networks (CNNs) have been used in image-based applications and have made significant progress toward early detection of hard-to-detect diseases such as Mild Cognitive Impairment (MCI) and its prodromal stage Alzheimer's Disease (AD). Despite this progress, there has been limited research on accurately distinguishing Normal Cognitive (NC) subjects from Early Mild Cognitive Impairment (EMCI) at the subject-level. This paper aims to address this gap by proposing the use of structural MRI (sMRI) images and demographic information, in conjunction with predictive models based on a shallow CNN architecture and a supervised hybrid neural network, to distinguish EMCI from NC at both the slice and subject level. These models have fewer parameters but still maintain a high level of performance in classifying EMCI and NC images and require fewer computational resources. Moreover, the model's performance was trained and tested using only the initial and first-year visit MRI images from the newly released ADNI3 dataset.