ViT-ALZ: Vision Transformer with Deep Neural Network for Alzheimer’s Disease Detection

This study introduces the ViT-ALZ framework, which combines Vision Transformer (ViT) and deep neural networks (DNNs) to accurately detect Alzheimer’s disease (AD) using MRI scans. Our approach involves resizing images and applying augmentation techniques for dataset enhancement. The ViT-ALZ model integrates a ViT-based architecture that captures planar features from axial slices and utilizes weighted fusions to emphasize important visual token characteristics. DNN further enhances feature processing, leading to precise AD classification. Experimental results showcase significant performance metrics: accuracy, sensitivity, specificity, precision, F1-score, and Kappa values are 97.65%, 97.85%, 97.92%, 96.86%, 97.65%, and 95.19%, respectively, on the Kaggle AD dataset. This research underscores the value of deep learning and MRI images in effectively classifying AD, highlighting their potential for early diagnosis and care.