A directional distance aided method for medical image segmentation.

Area of Science:

• Medical image analysis
• Computer vision
• Computational geometry

Background:

• Image segmentation faces challenges with boundary leakage due to poor edge resolution and insufficient local information.
• Level set methods are widely used but can struggle with accuracy in complex structures.

Purpose of the Study:

• To propose a novel directional distance aided (DDA) image segmentation method within the level set framework.
• To prevent boundary leakage in image segmentation by introducing a new regularization term.
• To enhance the accuracy and stability of segmenting complex structures, including medical images.

Main Methods:

• Formulated a new directional distance (DD) term to measure the "degree of protrusion" of points on the zero level set.
• Incorporated DD term with curvature, gradient terms, and user constraints to guide level set evolution.
• Augmented Gaussian smoothing with an antishrinkage operation and integrated vertex-based smoothing.

Main Results:

• The DDA method effectively prevents leakage into adjacent structures by penalizing protruding points.
• The DD term's independence from intensity or gradient boundaries enhances regional shape regulation.
• Demonstrated promising results and stability in segmenting simulated objects and abdominal aortic aneurysms in CT angiograms (2D and 3D).

Conclusions:

• The proposed DDA method offers a robust solution for image segmentation, particularly in medical imaging.
• The novel DD term significantly improves boundary adherence and detail preservation.
• The method shows potential for clinical applications requiring precise segmentation of vascular structures.