Detection of Alzheimer Disease in Neuroimages Using Vision Transformers: Systematic Review and Meta-Analysis

Area of Science:

• Artificial Intelligence in Medicine
• Neuroimaging Analysis
• Deep Learning for Disease Diagnosis

Background:

• Alzheimer disease (AD) is a progressive neurodegenerative disorder causing cognitive decline.
• Vision transformers (ViTs) are advanced deep learning models with potential in medical image analysis.
• ViTs show promise for early detection and diagnosis of AD.

Purpose of the Study:

• Systematically review studies applying ViTs to AD detection.
• Evaluate the diagnostic accuracy of ViTs in AD.
• Assess the impact of network architecture on ViT performance for AD diagnosis.

Main Methods:

• Systematic literature search across major databases (PubMed, Web of Science, etc.) from 2020-2024.
• Included studies used ViT models with neuroimaging data (MRI, PET) for AD detection.
• Meta-analysis of diagnostic accuracy metrics (sensitivity, specificity, likelihood ratios) using random-effects models.

Main Results:

• Meta-analysis of 11 studies showed high pooled diagnostic accuracy for ViTs in AD detection.
• Sensitivity: 0.925, Specificity: 0.957, AUC: 0.924.
• ViTs demonstrate significant potential for accurate and early AD diagnosis.

Conclusions:

• ViT models are effective in distinguishing AD patients from healthy controls.
• This review supports ViTs as valuable tools for neuroimaging-based AD diagnostic methodologies.
• Findings offer insights for future advancements in AI-driven AD diagnostics.