Electronic Thesis/Dissertation
 

Detecting Alzheimer’s Disease Using Vision Transformer With Quadrangles On Brain MRI

Open Access Deposited

Neural Attention To Neurodegeneration

Alzheimer’s disease presents many challenges in disease detection and diagnosis. Early detection is important for treatment planning and therapeutic interventions. Yet, conventional diagnostic methods tend to be invasive for patients and rely heavily on the subjective judgment of neurologists.This praxis addresses these challenges by using a novel Vision Transformer with Quadrangle Attention (ViT-QA) model to identify Alzheimer’s disease in brain MRI scans. The ViT-QA design combines global understanding from pre-trained Vision Transformers with focused local spatial attention through Quadrangle Attention. This approach detects widespread brain shrinkage patterns and localized anatomical brain shifts that mark Alzheimer’s disease progression. We evaluated our proposed ViT-QA model alongside a vanilla Vision Transformer and a ResNet-50 model. Our findings reveal that vision trans- former architectures outperform traditional CNNs. The vanilla ViT achieved 93.96% accuracy on the ADNI-1 test data, compared with ResNet-50 at 86.25%. ViT-QA achieved results of 85.73% for ADNI-1 and 99.15% for OASIS-1, with an AUC of 0.98, outperforming results for global attention mechanisms to date for Alzheimer’s disease detection. Augmenting the images with a feature optimization pipeline netted performance gains across all models, with ViT-QA demonstrating a 15.26% improvement in accuracy and recall, an 11.37% increase in precision, and a 17.51% increase in the F1-score. Attention map visualizations confirmed that the model focuses on relevant brain areas, namely the hippocampus, entorhinal cortex, lateral ventricles, and temporal lobes, all of which are known early indicators of Alzheimer’s disease pathology. Our work contributes to the growing body of research on vision trans- formers in medical imaging. It illustrates how the Quadrangle Attention architecture can assist in the detection of Alzheimer’s disease, laying the groundwork for future research into diagnosing neurodegenerative conditions.

Author Language Date created Type of Work License
  • All rights reserved
Rights statement GW Unit Degree Advisor Persistent URL

Notice to Authors

If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.

Thumbnail Title Date Uploaded Visibility Actions
Preview of Brown_gwu_0075A_17714.pdf Brown_gwu_0075A_17714.pdf 2025-12-15 Open Access