Unsupervised 4D myocardium segmentation with a Markov Random Field based deformable model

Area of Science:

• Medical Imaging and Image Analysis
• Computational Biology
• Cardiovascular Research

Background:

• Accurate segmentation of the myocardium is crucial for assessing cardiac function.
• Existing segmentation methods may lack robustness or require manual intervention.
• Magnetic Resonance Imaging (MRI) provides detailed anatomical information of the heart.

Purpose of the Study:

• To develop and validate a stochastic deformable model for unsupervised myocardium segmentation in cardiac MRI.
• To improve the accuracy and robustness of cardiac segmentation for functional parameter derivation.
• To evaluate the model's performance in patients with varying degrees of Acute Myocardial Infarction.

Main Methods:

• A probabilistic optimization framework was employed, treating segmentation as a Markov Random Field problem.
• The model integrates temporal-spatial smoothness with intensity and gradient features using Maximum Likelihood estimation.
• Left ventricle detection was performed using established image analysis tools prior to segmentation.

Main Results:

• The proposed stochastic deformable model achieved robust and flexible myocardium segmentation.
• Segmentation results were comparable to manual segmentations for derived cardiac function parameters.
• The method was validated on a cohort of 43 patients with Acute Myocardial Infarction.

Conclusions:

• The developed model offers a promising approach for automated and accurate cardiac MRI segmentation.
• This technique can aid in the quantitative assessment of cardiac function, particularly in the context of myocardial infarction.
• The unsupervised nature of the method enhances its clinical applicability.