Advancing Early Diagnosis: Predicting Mild Cognitive Impairment Progression in Normal Individuals Using Deep Learning on MRI Features

Abstract

One of the most challenging tasks for neurologists is the early diagnosis of Alzheimer’s disease (AD). Early and accurate diagnosis of the mild cognitive impairment (MCI) stage can enhance efforts to slow down the major consequences linked to this condition. Deep learning systems provide a promising performance in diagnosing the disease through neuroimaging analysis. This research aims to develop a deep learning-based system that efficiently identifies and analyzes the progression from Cognitively Normal (CN) to MCI, addressing the growing need for more accessible, accurate diagnostic tools. The proposed model comprises two distinct feature extraction paths to capture local and global image features. Each path includes advanced modules for feature refinement associated with the channel attention mechanism. The resultant output features are produced using a learned fusion technique from the two paths’ features and applied to the CN vs. MCI binary classifier. Furthermore, the proposed Suspected Subject Classifier (SSC) system applies various machine-learning methods to identify the suspected MCI subjects. The results showed a comparative performance for the binary diagnosis of CN individuals and those with MCI, achieving an accuracy of 91.6% and 88.4% for multi-class diagnoses, including the prediction of progression from normal to confirmed MCI. This study represents an exceptional stride toward predicting early MCI in normal individuals. By enhancing prediction efficiency for early disease progression in normal individuals, our method can potentially advance intervention strategies and improve patient care outcomes.