A computational framework for Alzheimer's disease detection using SwinRes Transformer.

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Medical Diagnostics

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder impacting cognitive functions.
• Early and accurate AD detection is crucial for improving patient outcomes.
• Conventional models struggle with capturing complex image features and dependencies efficiently.

Purpose of the Study:

• To implement a novel SwinRes Transformer model for accurate Alzheimer's disease detection.
• To address the limitations of conventional models in capturing structural abnormalities and long-range dependencies.
• To provide a computationally efficient solution for scalable AD detection.

Main Methods:

• Utilized Dilated InceptionV3 for feature extraction with Fused MBConv and Group Convolution to reduce training time and parameters.
• Integrated a Modified Residual Network for local feature extraction and a Convolutional Shifted-Window Spatial-Channel Swin Transformer for global information capture.
• Employed a Feature Fusion Module to refine extracted features, enhancing relevant information and reducing redundancy.

Main Results:

• The SwinRes Transformer model achieved a superior accuracy of 98.93% on four Magnetic Resonance Imaging (MRI) datasets.
• Demonstrated a significantly lower execution time of 1.1 seconds compared to existing methodologies.
• Effectively captured structural abnormalities and long-range inter-regional dependencies from MRI data.

Conclusions:

• The proposed SwinRes Transformer model offers a computationally efficient and highly accurate solution for Alzheimer's disease detection.
• The model's architecture effectively integrates local and global feature extraction for improved diagnostic performance.
• This approach shows promise for scalable and early detection of Alzheimer's disease using neuroimaging data.