Diffusion boost: Leveraging diffusion model for groupwise registration in myocardial T1 mapping

Area of Science:

• Medical Imaging
• Cardiovascular Disease Diagnosis
• Quantitative MRI

Background:

• Myocardial T1 mapping is vital for cardiovascular disease diagnosis using quantitative MRI.
• Cardiac and respiratory motion degrade T1 estimation accuracy.
• Groupwise registration aligns all images simultaneously, but struggles with T1 mapping's varying contrast.

Purpose of the Study:

• To develop a diffusion model-based framework for groupwise registration of myocardial T1 mapping.
• To address the challenge of extracting structural information from T1 mapping data with drastic contrast changes.

Main Methods:

• A novel template-free groupwise registration framework utilizing a diffusion process.
• A Hybrid Attention Feature Fusion (HAFF) module for multi-scale feature fusion.
• Evaluation on a public myocardial T1 mapping dataset of 210 patients.

Main Results:

• The proposed method significantly outperforms state-of-the-art approaches in myocardial T1 mapping registration.
• Achieved a Dice score of 0.839, groupwise Dice score of 0.601, Hausdorff distance of 10.389 mm, and T1 mapping error of 11.372 ms.
• Demonstrated superiority in quantitative metrics compared to existing methods.

Conclusions:

• The framework enables robust groupwise registration for myocardial T1 mapping.
• Leverages diffusion models for strong feature extraction in image registration.
• Highlights the potential of diffusion models beyond image generation for medical applications.