Alzhinet: an explainable self-attention based classification model to detect Alzheimer from 3D volumetric MRI data

Alzheimer’s Disease is a significant global healthcare challenge that requires early and accurate diagnosis for better patient care and a deeper understanding of its pathology. In this study, we introduce “AlzhiNet”, an advanced deep learning model designed to diagnose Alzhimer’s Disease by using 3D Volumetric MRI data for multi-class diagnosis. AlzhiNet uses self-attention mechanisms to distinguish between Alzhimer’s Disease stages like Mild Cognitive Impairment, and Alzheimer’s Disease including subjects who are Cognitively Normal as a control group. It is a pioneering step towards explainability and helps bridge the gap between Artificial Intelligence and clinical expertise by unveiling the slices that are essential to diagnostic decisions. We describe AlzhiNet’s architecture, training methodology, and evaluation results, drawing insights from a dataset of 2098 MRI volumes. AlzhiNet’s impact extends far beyond being just a diagnostic tool, as it signifies a significant stride towards improved patient care and deeper insights into the complex pathology of Alzheimer’s disease.