A Novel Segmentation Approach Combining Region- and Edge-Based Information for Ultrasound Images

Area of Science:

• Medical Imaging
• Image Processing
• Computational Biology

Background:

• Ultrasound (US) imaging is a widely used diagnostic tool.
• Accurate segmentation of US images is crucial for analysis but challenging due to low image quality.

Purpose of the Study:

• To develop a novel segmentation scheme for ultrasound images.
• To enhance segmentation accuracy by integrating region and edge-based information using a robust graph-based (RGB) method.

Main Methods:

• The proposed method utilizes a region of interest (ROI) defined by two diagonal points.
• Image preprocessing includes bilateral filtering, histogram equalization, and pyramid mean shift filtering.
• Segmentation is performed using the RGB method optimized with particle swarm optimization (PSO).

Main Results:

• The RGB segmentation method demonstrated superior performance compared to three existing approaches.
• Achieved the lowest Average Region Error (ARE) at 10.77%.
• Obtained the second-highest True Positive Volume Fraction (TPVF) at 85.34% and the second-lowest False Positive Volume Fraction (FPVF) at 4.48%.

Conclusions:

• The proposed RGB segmentation method effectively addresses the challenges of ultrasound image segmentation.
• The integration of region and edge information, coupled with optimization techniques, leads to improved segmentation accuracy and reliability.