Brain imaging-to-graph generation using adversarial hierarchical diffusion models for MCI causality analysis.

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Effective connectivity analysis reveals causal brain patterns crucial for understanding cognitive diseases.
• Current methods for estimating effective connectivity from brain imaging are labor-intensive and prone to errors due to manual parameter tuning.
• Accurate effective connectivity estimation is vital for early diagnosis and therapeutic development in cognitive disorders.

Purpose of the Study:

• To propose a novel Brain Imaging-to-Graph Generation (BIGG) framework for mapping functional magnetic resonance imaging (fMRI) data to effective brain connectivity.
• To enhance the analysis of mild cognitive impairment (MCI) through improved estimation of causal patterns in brain networks.
• To overcome limitations of existing methods by automating parameter settings and reducing estimation errors.

Main Methods:

• The BIGG framework utilizes diffusion denoising probabilistic models (DDPM).
• Each denoising step within DDPM is modeled as a generative adversarial network (GAN) for progressive translation of fMRI data to effective connectivity.
• A diffusive factor is introduced to improve the efficiency and quality of denoising inference, allowing for larger sampling step sizes.

Main Results:

• The BIGG framework demonstrated feasibility and efficacy on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset.
• The proposed model achieved superior prediction performance compared to existing competing methods.
• The model successfully identified MCI-related causal connections that align with findings from clinical studies.

Conclusions:

• The BIGG framework offers a robust and automated approach for estimating effective brain connectivity from fMRI data.
• This novel method shows significant potential for improving the early diagnosis and understanding of mild cognitive impairment.
• The findings suggest BIGG can contribute to advancing research in cognitive disease mechanisms and drug development.