Joint Brain Structure-Function Analysis with Correlation-Consistent Learning for Alzheimer's Disease Diagnosis.

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Computational Neuroscience

Background:

• Integrating structural MRI (sMRI) and resting-state functional MRI (rs-fMRI) for Alzheimer's Disease (AD) diagnosis presents challenges in data fusion and clinical interpretability.
• Existing methods struggle to effectively combine heterogeneous neuroimaging data while maintaining clinical relevance.

Purpose of the Study:

• To develop a novel transformer-based framework for unified multi-modal neuroimaging analysis in AD diagnosis.
• To enhance diagnostic accuracy and clinical interpretability by integrating sMRI and rs-fMRI data.

Main Methods:

• A transformer-based framework was proposed, leveraging anatomical priors to guide attention to AD-relevant regions.
• A learnable mapping transformed sMRI features into regional representations, with a consistency constraint for structural-functional coupling alignment.
• Bayesian fusion integrated aligned multi-modal features for diagnosis.

Main Results:

• The proposed method achieved superior diagnostic accuracy on the ADNI dataset compared to state-of-the-art approaches.
• The framework provided clinically interpretable insights into AD-related brain connectivity patterns.
• Demonstrated effective unification of heterogeneous imaging features into coherent region-level representations.

Conclusions:

• The novel framework successfully integrates sMRI and rs-fMRI data for enhanced AD diagnosis.
• The approach offers a significant advancement in multi-modal neuroimaging by balancing diagnostic performance with clinical interpretability.
• This work paves the way for more effective AI-driven diagnostic tools in neurodegenerative diseases.