Automatic delineation of the myocardial wall from CT images via shape segmentation and variational region growing

Area of Science:

• Cardiovascular Imaging and Radiology
• Medical Image Analysis
• Computational Cardiology

Background:

• Accurate quantitative evaluation of cardiac diseases relies on precise measurements of ventricle volume, mass, and ejection fraction.
• Myocardial wall delineation is crucial for these evaluations but is challenging due to anatomical variability and image quality issues in cardiac CT.
• Existing methods often struggle with the complex shapes and image noise inherent in cardiac imaging.

Purpose of the Study:

• To develop and present an automatic method for extracting the myocardial wall of the left and right ventricles from cardiac CT images.
• To improve the accuracy and efficiency of cardiac structure segmentation for diagnostic purposes.
• To provide a robust solution for delineating ventricular walls despite variations in shape and image quality.

Main Methods:

• A sequential localization approach for the left and right ventricles.
• Endocardium localization using online geometric features from CT images and an active contour model on the blood-pool surface (left ventricle).
• Epicardium segmentation using a variational region-growing model, with the active contour model applied to a derived heart surface for the right ventricle.

Main Results:

• Successful automatic extraction of the myocardial wall for both left and right ventricles.
• Demonstrated robustness and accuracy of the proposed method.
• Validation through experimental results on a diverse dataset of 33 human and 12 pig cardiac CT images.

Conclusions:

• The presented automatic method effectively delineates the myocardial walls of the left and right ventricles in cardiac CT images.
• The approach shows promise for enhancing quantitative evaluations in the diagnosis and prognosis of cardiac diseases.
• The technique offers a reliable and accurate solution for a challenging task in cardiovascular imaging.