Active contours driven by local and global intensity fitting energy with application to brain MR image segmentation

Area of Science:

• Medical Image Analysis
• Computer Vision
• Computational Imaging

Background:

• Active contour models are widely used for image segmentation.
• Traditional models often struggle with initialization sensitivity and boundary leakage.
• Variational level set methods offer a robust framework but require careful energy functional design.

Purpose of the Study:

• To propose an improved region-based active contour model within a variational level set framework.
• To enhance segmentation accuracy and robustness through a novel energy functional design.
• To enable flexible contour initialization for complex image segmentation tasks.

Main Methods:

• Developed a novel energy functional combining local and global intensity fitting terms.
• Integrated the energy functional into a variational level set formulation.
• Incorporated a level set regularization term for accurate computations.
• Extended the model from a two-phase to a multi-phase formulation.

Main Results:

• The proposed model demonstrates superior accuracy and robustness compared to existing methods.
• Flexible initialization of contours is achieved, overcoming limitations of traditional approaches.
• Successful application to brain MR image segmentation yielded desirable results.

Conclusions:

• The improved region-based active contour model offers significant advantages for image segmentation.
• The novel energy functional design effectively guides contour evolution and boundary adherence.
• This method shows great potential for medical image analysis, especially in brain MRI segmentation.