PVTAD: ALZHEIMER'S DISEASE DIAGNOSIS USING PYRAMID VISION TRANSFORMER APPLIED TO WHITE MATTER OF T1-WEIGHTED...

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Biomarker Discovery

Background:

• Alzheimer's disease (AD) diagnosis requires early intervention, but current methods struggle to capture complex neural connection patterns.
• Existing machine learning models like CNNs and ViTs may not fully extract multidimensional local and global features indicative of AD.

Approach:

• Proposed PVTAD, a novel approach using a pretrained pyramid vision transformer (PVT) on T1-weighted structural MRI (sMRI) white matter data.
• Combines CNN and ViT advantages to extract local and global AD-specific features from WM coronal middle slices.
• Utilized the ADNI dataset for experiments on subjects with T1-weighted MPRAGE sMRI scans.

Key Points:

• PVTAD achieved high accuracy (97.7%) and F1-score (97.6%) in classifying Alzheimer's disease (AD) vs. cognitively normal (CN) cases.
• The novel approach outperforms traditional CNN and standard ViT architectures for AD detection using sMRI data.
• Focuses on extracting biomarkers from white matter patterns in MRI scans for improved diagnostic precision.

Conclusions:

• PVTAD demonstrates significant potential as an advanced tool for accurate and early Alzheimer's disease diagnosis.
• The findings highlight the efficacy of combining PVT with sMRI white matter analysis for neurodegenerative disease detection.
• This research contributes to developing more effective AI-driven diagnostic solutions in neurology.