Anatomical knowledge based level set segmentation of cardiac ventricles from MRI

Area of Science:

• Medical Imaging
• Computational Anatomy
• Biomedical Engineering

Background:

• Accurate segmentation of cardiac ventricles is crucial for diagnosing cardiovascular diseases.
• Existing methods often require manual intervention or extensive training data.

Purpose of the Study:

• To develop a novel level set framework for robust segmentation of cardiac left ventricle (LV) and right ventricle (RV) from MRI.
• To improve segmentation accuracy and stability by incorporating anatomical knowledge and a convexity-preserving mechanism.

Main Methods:

• A bi-layer level set (BILLS) formulation to represent endocardium and epicardium.
• Convexity preserving bi-layer level set (CP-BILLS) evolution for simultaneous contour recovery.
• Automatic selection of the k-level and convex shape decomposition (CSD) for ventricle segmentation.

Main Results:

• The proposed method achieves accurate and convex segmentation of LV and RV endocardial and epicardial contours.
• Demonstrates superior performance in segmentation accuracy and stability compared to traditional methods.
• Achieves comparable accuracy and stability to deep learning methods without the need for training data.

Conclusions:

• The novel level set framework offers an efficient and accurate solution for cardiac ventricle segmentation.
• The method's independence from training data makes it a valuable tool for clinical applications.
• This approach provides a strong alternative to deep learning methods, particularly when training data is limited.