A Novel Attention-guided 3D-CNN Framework for Alzheimer's Disease Classification

Abstract Alzheimer's disease is a devastating illness that frequently leads to dementia and is a major cause of death among the elderly, posing a significant global health challenge. The alarming statistic that one in three elderly individuals will develop dementia or Alzheimer's underscores the growing urgency of this problem. This study presents a novel 3D-CNN model enhanced with an attention mechanism for more accurate detection and classification of Alzheimer's disease from MRI scans. The model combines the power of CNNs and MLPs to improve both feature extraction and classification performance. To automatically identify and prioritize crucial features in Magnetic Resonance Imaging (MRI) scans, this study employed spatial attention mechanisms, eliminating the need for manual feature selection. Evaluated on the ADNI dataset, the model achieved cutting-edge classification accuracies of 97.86% (Alzheimer's disease (AD) vs. Normal Control (NC), 90.65% (Mild Cognitive Impairment (MCI) vs. Normal Control (NC), and 92.18% (Mild Cognitive Impairment (MCI) vs. Alzheimer's disease (AD), surpassing previous approaches.