Alzheimer’s disease (AD), a progressive neurodegenerative disorder, poses significant challenges for early diagnosis due to subtle symptom onset and overlap with normal aging. This study aims to develop an effective deep learning model for classifying four AD stages (Non-Demented, Very Mild Demented, Mild Demented, Moderate Demented) using brain MRI scans. We propose a Multi-Axis Vision Transformer (MaxViT)-based framework, leveraging transfer learning and robust data augmentation on the Kaggle Alzheimer’s MRI Dataset to address class imbalance and enhance generalization. The model employs MaxViT’s multi-axis attention mechanisms to capture both local and global patterns in MRI images. Our approach achieved a classification accuracy of 99.60%, with precision of 99.0%, recall of 98.1%, and F1-score of 98.51%. These results highlight MaxViT’s superior ability to differentiate AD stages, particularly in distinguishing challenging early stages. The proposed model offers a reliable tool for early AD diagnosis, laying a strong foundation for future clinical applications and interdisciplinary research in neurodegenerative disease detection. Future work should explore larger, more diverse datasets and additional biomarkers to further validate and enhance model performance.
Aslan et al. (2025) studied this question.