Design of a CNN–Swin Transformer Model for Alzheimer’s Disease Prediction

This project presents a hybrid deep learning model that combines Convolutional Neural Networks (CNN) and Transformer architectures for classifying Alzheimer’s disease stages using MRI images. The proposed framework integrates the strengths of CNN-based models such as EfficientNet-B3, ResNet-50, , and DenseNet-121 with advanced attention mechanisms from Swin Transforme. Each fusion model is trained to categorize MRI scans into five classes: Irrelevant, Mild Impairment, Moderate Impairment, No Impairment, and Very Mild Impairment. The CNN component captures spatial and hierarchical image features, while the Transformer component enhances contextual learning and long-range dependency extraction. A Flask-based application is developed to provide modules such as Home, Register, Login, Classification, and Logout, ensuring a complete system workflow. The integration of CNNs and Transformers demonstrates a balanced approach that improves classification performance and enhances feature representation.

Keywords: CNN, Swin Transformer, Alzheimer’s disease, MRI, Deep Learning, EfficientNet, ResNet, , DenseNet, Vision Transformer