Nonrigid active shape model-based registration framework for motion correction of cardiac T₁ mapping.

Area of Science:

• Medical imaging
• Cardiovascular MRI
• Image processing

Background:

• Accurate myocardial T1 mapping is crucial for assessing myocardial tissue characteristics.
• Cardiac motion from breathing and diaphragmatic drifts introduces significant artifacts in T1-weighted images, complicating accurate T1 map reconstruction.
• Existing methods struggle to fully correct for these motion-induced errors, limiting the diagnostic potential of T1 mapping.

Purpose of the Study:

• To propose and evaluate a novel framework utilizing active shape models for motion correction in myocardial T1 maps.
• To enhance the accuracy of T1 measurements by mitigating cardiac motion artifacts.
• To enable more comprehensive analysis of myocardial segments by reducing T1 estimation errors.

Main Methods:

• Developed multiple appearance models at varying inversion times to capture blood-myocardium contrast and brightness changes.
• Automatically segmented myocardial borders using these models to extract contours for registering T1-weighted images.
• Trained and validated the framework on data from 210 patients using a free-breathing acquisition protocol, with quantitative validation using mean absolute distance and Dice index.

Main Results:

• Demonstrated a significant reduction in mean absolute distance (3.3±1.6 to 2.3±0.8 mm) and an increase in Dice index (0.89±0.08 to 0.94±0.04%) in a testing set of 180 patients.
• Improved T1 map quality in 70% of motion-affected maps post-correction.
• Significantly reduced motion-corrupted myocardial segments from 21.8% to 8.5%.

Conclusions:

• The proposed nonrigid registration framework effectively corrects for motion-induced artifacts in T1-weighted images.
• This method enhances the accuracy of T1 measurements in the myocardium.
• The framework allows for T1 measurements in a greater number of myocardial segments, improving diagnostic utility.