Breast ultrasound image segmentation: a survey

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Biomedical Engineering

Background:

• Breast cancer is a leading global health concern for women.
• Ultrasound imaging is a primary diagnostic tool for breast abnormalities.
• Accurate segmentation of breast ultrasound (BUS) images is vital for effective computer-aided diagnosis (CAD) systems.

Purpose of the Study:

• To comprehensively review and categorize segmentation approaches for BUS images.
• To analyze techniques developed over the last decade.
• To identify challenges and limitations in current BUS image segmentation methods.

Main Methods:

• Literature review focusing on BUS image segmentation techniques.
• Categorization of methods into seven classes: thresholding-based, clustering-based, watershed-based, graph-based, active contour model, Markov random field, and neural network.
• Analysis of representative papers for each technique class.

Main Results:

• A comparative summary of various BUS image segmentation techniques, outlining their respective advantages and disadvantages.
• Identification of persistent challenges in BUS image segmentation, including speckle noise, low contrast, blurry boundaries, and signal inhomogeneity.
• Confirmation that BUS image segmentation remains an open research problem.

Conclusions:

• This paper provides the first comprehensive review of BUS image segmentation approaches.
• The findings are valuable for researchers in ultrasound image segmentation and developers of BUS CAD systems.
• Understanding the strengths and weaknesses of different techniques is essential for advancing BUS image analysis.