AI-Driven Generation of Post-Contrast T1 and ECV Maps from Native T1 Map in Cardiac MRI

Area of Science:

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

Background:

• Cardiac magnetic resonance (CMR) imaging often requires contrast agents for myocardial tissue characterization.
• Gadolinium-based contrast agents carry risks and increase examination costs.
• Developing non-invasive methods for myocardial tissue assessment is crucial.

Purpose of the Study:

• To develop and evaluate an AI-based method for generating virtual post-contrast T1 and extracellular volume (ECV) maps from native T1 maps.
• To assess the performance of the AI-generated maps compared to reference methods.
• To explore a contrast-free approach for myocardial tissue characterization using CMR.

Main Methods:

• A modified self-consistent recursive diffusion bridge framework was employed to generate virtual post-contrast T1 maps from native T1 maps.
• Cardiac MRI data from 813 patient slices with suspected myocardial disease were used for training and validation.
• Virtual ECV maps were computed by combining virtual post-contrast T1 maps with native T1 maps.

Main Results:

• Myocardial T1 values from virtual and reference post-contrast maps showed similar distributions with a reduced offset after ECV transformation.
• Virtual myocardial ECV maps demonstrated acceptable agreement with reference ECV maps (RMSE: 3.05%, R²: 0.585).
• The method achieved Bland-Altman 95% limits of agreement of -5.98% to +6.06% for ECV, despite slice-to-slice variability.

Conclusions:

• The AI method successfully generated post-contrast T1 and ECV maps from native T1 maps without contrast agents.
• This contrast-free approach offers a promising non-invasive alternative for myocardial tissue characterization.
• The method has the potential to reduce costs, mitigate contrast-related risks, and enhance patient safety in CMR.