Convexity preserving level set for left ventricle segmentation.

Area of Science:

• Medical Imaging
• Computational Biology
• Image Segmentation

Background:

• Accurate cardiac left ventricle (LV) segmentation is crucial for clinical applications.
• Existing segmentation algorithms struggle to differentiate trabeculae and papillary muscles from myocardium due to similar intensities, leading to inaccurate LV segmentation.
• The desired LV segmentation should encompass the cavity, trabeculae, and papillary muscles, forming a convex shape.

Purpose of the Study:

• To propose a novel level set method for cardiac left ventricle (LV) segmentation that ensures the convexity of the segmented region.
• To improve the accuracy of LV segmentation by preventing misclassification of trabeculae and papillary muscles.

Main Methods:

• A level set method incorporating a convexity-preserving mechanism was developed.
• The curvature of level set contours is utilized to control and enforce the convexity of the segmented LV shape.
• The method was evaluated experimentally and compared against other level set and deep learning techniques.

Main Results:

• The proposed level set method demonstrated superior segmentation accuracy compared to traditional level set methods.
• It achieved comparable accuracy to state-of-the-art deep learning methods.
• The method successfully preserved the convex shape of the left ventricle during segmentation.

Conclusions:

• The proposed convexity-preserving level set method offers an effective solution for accurate cardiac left ventricle (LV) segmentation.
• It provides a viable alternative to deep learning approaches, particularly in scenarios lacking large training datasets and manual segmentations.
• This method is advantageous for clinical applications requiring precise and robust LV segmentation.