Left ventricle segmentation via two-layer level sets with circular shape constraint.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Image Segmentation

Background:

• Accurate segmentation of the left ventricle (LV) is crucial for cardiac function assessment.
• Existing methods often struggle with variations in LV shape and image artifacts.

Purpose of the Study:

• To develop a novel, model-free method for segmenting the left ventricle (LV) from short-axis cardiac MRI.
• To improve LV segmentation accuracy by incorporating a circular shape constraint.

Main Methods:

• A two-layer level set method incorporating a novel circle fitting term for endocardial detection.
• The circle fitting term penalizes deviations from a fitted circle, addressing segmentation challenges like outflow tracts.
• The method uses distinct level contours for endocardium and epicardium, evolved by edge-based and region-based active contour models, respectively.

Main Results:

• Quantitative validation on a public dataset (MICCAI 2009 LV segmentation challenge) demonstrated high accuracy.
• The proposed method effectively handles challenges such as basal slice outflow tracts and intensity overlap.
• Experimental results show superior accuracy and robustness compared to state-of-the-art techniques.

Conclusions:

• The proposed circular shape constraint and two-layer level set method offer an accurate and robust approach for left ventricle segmentation.
• This model-free technique provides a valuable tool for cardiac MRI analysis.
• The method demonstrates significant improvements in handling common segmentation difficulties.