A Novel 3D-CNN with DAG Framework for Enhanced Alzheimer's Disease Diagnosis Using Structural MRI

Alzheimer's is the most common and neurodegenerative kind of dementia impacting the elderly. It is noteworthy that a limited number of older individuals with Alzheimer's disease are timely and effectively diagnosed in the early stages. This is an essential precondition, since the illness is more controllable and therapy is more successful when diagnosed at the beginning or early mild cognitive impairment stage. In recent years, several researchers have focused on enhancing prediction and diagnostic accuracy using an AI-driven image categorization methodology. This research proposes a novel 3D-CNN with Directed Acyclic Graph architecture aimed at precise and efficient diagnosis of Alzheimer's disease in its early stages, utilizing segmented left and right hippocampus regions, grey matter, cerebrospinal fluid, and white matter extracted from structural MRI images of the brain. The structural MRI images undergo pre-processing using 3D Slicer and SPM12. This study utilizes the ADNI, NACC, and KAGGLE datasets. The findings are persuasive, since the suggested 3D-CNN with Directed Acyclic Graph demonstrates superior performance achieving an accuracy of 96.86% relative to other current studies.