A Novel Multi-Scale Deep Learning Approach for the Early Detection of Alzheimer’s Disease Using fMRI

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that remains challenging to diagnose in its early stages. This study presents a novel 3D multi-scale Convolutional Neural Network (CNN) model that utilizes resting-state functional MRI (rs-fMRI) data to classify different stages of Alzheimer’s disease. The proposed model integrates multi-scale feature extraction from the hippocampal region, optimizing the detection of fine-grained and global patterns indicative of AD progression. The model was evaluated on a balanced dataset of 300 subjects, applying cross-validation to ensure robustness across multiple binary classification tasks. Our experimental results demonstrate a classification accuracy of up to 93% for distinguishing Alzheimer’s Disease from Cognitively Normal subjects. High sensitivity, specificity, and precision were observed across all classification tasks, indicating the model’s effectiveness in early-stage Alzheimer’s detection. These findings suggest the proposed approach could enhance clinical diagnosis and facilitate timely interventions for AD patients.