Efficient Johnson-S_B Mixture Model for Segmentation of CT Liver Image.

Area of Science:

• Medical Imaging
• Computational Biology
• Image Processing

Background:

• Traditional Gaussian Mixture Models (GMM) struggle with the skewed histogram distributions common in liver CT slices.
• Accurate liver segmentation is crucial for diagnosis and treatment planning in medical imaging.

Purpose of the Study:

• To develop a novel liver CT image segmentation method addressing the limitations of GMM.
• To introduce and validate the Johnson-SB mixture model (J_SBMM) for improved liver CT segmentation.

Main Methods:

• Developed a J_SBMM-based segmentation algorithm using expectation-maximization (EM) and maximum likelihood.
• Proposed a two-part histogram division technique (J_SBMM-TDH) to leverage Johnson-SB's skewness for enhanced accuracy.
• Compared J_SBMM-TDH with GMM on abdominal CT image sequences.

Main Results:

• J_SBMM-TDH demonstrates more stable segmentation thresholds compared to GMM, which are sensitive to cluster numbers.
• The J_SBMM-TDH method achieved preferable segmentation results and superior robustness in liver CT image segmentation.
• Johnson-SB mixture model offers a flexible asymmetrical distribution suitable for complex histogram data.

Conclusions:

• The proposed J_SBMM-TDH method offers a significant improvement over GMM for liver CT image segmentation.
• J_SBMM-TDH provides better accuracy and robustness, making it a valuable tool in medical image analysis.
• The Johnson-SB mixture model is effective for handling skewed distributions in medical imaging applications.