MRI-Based Deep Learning Models for Precise Detection of Alzheimer’s Disease

As a progressive neurodegenerative disease, Alzheimer’s poses a significant challenge in determining an accurate diagnosis and classification. The medical interpretation of its symptoms primarily relies on the expertise of trained professionals, which can lead to variations in diagnosis and potential errors. A promising solution is to leverage machine learning and deep learning techniques. This study evaluates various deep-learning approaches for classifying Alzheimer’s disease using MRI images. We explored traditional CNNs, ResNet50, Inception v3, Swin Transformer, and Vision Transformer (ViT), comparing their performance with and without data augmentation. Our results demonstrate that the deep learning models, particularly ViT and Swin Transformer, significantly improved classification metrics when data augmentation was applied. Specifically, ViT achieved an accuracy of 95.67%, precision of 96.21%, recall of 95.67%, and an AUC of 99.75% after augmentation. This study highlights the effectiveness of deep learning in enhancing diagnostic accuracy for Alzheimer’s disease, underscoring the potential of these techniques in clinical settings to reduce diagnostic variability and improve patient outcomes.