Pyramid graph cut: Integrating intensity and gradient information for grayscale medical image segmentation

Area of Science:

• Medical Image Analysis
• Computer Vision
• Computational Imaging

Background:

• Grayscale medical image segmentation is difficult due to similar pixel intensities and weak gradients.
• Existing methods using only intensity or gradient information often yield inaccurate results.
• Previous attempts at integrating information were complex and not broadly applicable.

Purpose of the Study:

• To develop a novel and effective method for grayscale medical image segmentation.
• To integrate both intensity and gradient information within a unified framework.
• To improve segmentation accuracy and robustness for diverse medical imaging tasks.

Main Methods:

• Introduced a novel technique: pyramid graph cut.
• Combined intensity and gradient information in a pyramid-shaped graph structure.
• Utilized a single source node (intensity) and multiple sink nodes (gradient) with a min-cut approach.

Main Results:

• Pyramid graph cut demonstrated superior performance over intensity-only and gradient-only methods.
• Achieved excellent segmentation results on CT, MRI, and US liver tumor datasets.
• Reported high Dice scores: 90.49±5.23% (CT), 88.86±11.77% (MRI), 90.68±2.45% (US).

Conclusions:

• Pyramid graph cut offers an effective and generic approach for grayscale medical image segmentation.
• The integration of intensity and gradient information in a graph structure enhances segmentation accuracy.
• The method shows significant potential for clinical applications in medical image analysis.