Knowledge-driven interpretative conditional diffusion model for contrast-free myocardial infarction enhancement...

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Medical Image Synthesis

Background:

• Contrast agents (CAs) in myocardial infarction (MI) imaging pose risks and complicate workflows.
• Current methods for synthesizing MI enhancement (MIE) images without CAs struggle with interpretability and information loss.
• Existing approaches lack robust integration of kinematic and morphological data.

Purpose of the Study:

• To develop a knowledge-driven model for synthesizing contrast-free MIE images with improved interpretability and quality.
• To address limitations in kinematic inference, morphological knowledge integration, and fusion in current MIE synthesis methods.
• To enhance the accuracy of detecting myocardial scars and their boundaries.

Main Methods:

• Proposed a knowledge-driven interpretative conditional diffusion model (K-ICDM).
• Introduced cardiac causal intervention for interpretable kinematic inference.
• Implemented a knowledge-driven cognitive strategy for morphological feature capture.
• Utilized an information-specific adaptive fusion strategy for kinematic and morphological data integration.

Main Results:

• K-ICDM achieved state-of-the-art performance in contrast-free MIE image synthesis.
• Improved Structural Similarity Index Measure (SSIM) by at least 2.1% over existing methods.
• Demonstrated effective capture of the relationship between myocardial motion and scar distribution.
• Enabled accurate synthesis of subtle scar boundaries.

Conclusions:

• The K-ICDM model successfully synthesizes high-quality, interpretable MIE images without contrast agents.
• The innovations address key limitations in existing AI-driven medical image synthesis.
• This approach offers a safer and more efficient alternative for MI diagnosis and treatment planning.