Resizer Swin Transformer-Based Classification Using sMRI for Alzheimer’s Disease

Structural magnetic resonance imaging (sMRI) is widely used in the clinical diagnosis of diseases due to its advantages: high-definition and noninvasive visualization. Therefore, computer-aided based on sMRI images broadly applied classifying Alzheimer’s disease (AD). Due excellent performance Transformer computer vision, Vision (ViT) has been employed for AD classification recent years. The ViT relies access large datasets, while sample size brain datasets relatively insufficient. Moreover, preprocessing procedures are complex labor-intensive. To overcome limitations mentioned above, we propose Resizer Swin (RST), a deep-learning model that can extract information from only briefly processed achieve multi-scale cross-channel features. In addition, pre-trained our RST natural image dataset obtained better performance. We achieved 99.59% 94.01% average accuracy ADNI AIBL respectively. Importantly, sensitivity 99.59%, specificity 99.58%, precision 99.83% dataset, which than or comparable state-of-the-art approaches. experimental results prove prediction compared with CNN-based models.