Split Bregman method based level set formulations for segmentation and correction with application to MR images and...

Area of Science:

• Medical Imaging
• Computer Vision
• Image Processing

Background:

• Magnetic Resonance (MR) imaging is crucial for clinical diagnosis, but image segmentation is challenging due to intensity inhomogeneity.
• Accurate and fast segmentation of MR images is vital for medical research.

Purpose of the Study:

• To develop an improved active contour model for accurate MR image segmentation and bias field correction.
• To enhance segmentation accuracy and efficiency by combining level set evolution (LSE) with the split Bregman method.

Main Methods:

• An improved active contour model integrating the level set evolution (LSE) model and the split Bregman method.
• Incorporation of a bias field correction into the energy functional to address image inhomogeneity.
• Application of two-phase, multi-phase, and vector-valued formulations to segment synthetic and real MR images.

Main Results:

• The proposed model achieved satisfactory segmentation and correction for both grayscale and color MR images.
• Demonstrated higher accuracy and superiority compared to the standard LSE model.
• Showcased robustness to noise and insensitivity to initial contour placement.

Conclusions:

• The improved active contour model effectively segments inhomogeneous MR images and provides corrected, homogeneous images.
• The split Bregman method accelerates the segmentation process, reducing computation time and iterations.
• The model offers a robust and accurate solution for MR image analysis in clinical medicine and research.