A variational approach to liver segmentation using statistics from multiple sources.

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Radiology

Background:

• Medical image segmentation is crucial for research and clinical applications.
• Automatic liver segmentation is difficult due to noise and low contrast in CT scans.

Purpose of the Study:

• To develop a robust variational method for semi-automatic liver segmentation in CT volumes.
• To improve segmentation accuracy and efficiency for clinical applications.

Main Methods:

• A slice-by-slice variational approach using a level set framework.
• Incorporation of statistical shape prior, global Gaussian intensity analysis, and local statistical features.
• Utilized an improved Chan-Vese model for refining liver shape, especially in narrow regions.

Main Results:

• Achieved superior performance on 3D-IRCADb and SLIVER07 datasets.
• Obtained best Volumetric Overlap Error (VOE), Root Mean Square Symmetric Surface Distance (RMSD), and Maximum Symmetric Surface Distance (MSD) on 3D-IRCADb.
• Secured best Average Symmetric Surface Distance (ASD) and RMSD on SLIVER07.
• Demonstrated competitive segmentation performance compared to existing state-of-the-art techniques.

Conclusions:

• The proposed variational method offers a significant advancement in liver segmentation accuracy.
• This technique holds promise for enhancing medical image analysis in clinical practice.
• The method's robustness in handling noise and low contrast makes it suitable for real-world applications.