A dual-stage method for lesion segmentation on digital mammograms.

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Biomedical Engineering

Background:

• Accurate mass lesion segmentation on mammograms is crucial for early breast cancer detection.
• Lesions are often obscured by dense breast tissue, posing a significant challenge for automated analysis.

Purpose of the Study:

• To develop and evaluate an automated method for precise delineation of mass lesion boundaries on digital mammograms.
• To improve the accuracy and reliability of computer-aided segmentation in mammography.

Main Methods:

• A geometric active contour model minimizing an energy function based on regional homogeneities was employed.
• An initial contour was generated using a radial gradient index (RGI)-based method for computational efficiency.
• Automatic background estimation and a dynamic stopping criterion were integrated to refine contour evolution.

Main Results:

• The proposed algorithm achieved an 85% correct segmentation rate at an overlap threshold of 0.4 on a database of 739 mammograms.
• This represents a statistically significant improvement over conventional region-growing (69%) and RGI-based (73%) methods.
• The method demonstrated enhanced accuracy in segmenting lesions embedded within varying parenchymal densities.

Conclusions:

• The developed active contour model with integrated RGI segmentation offers a robust and accurate solution for automatic mass lesion segmentation in digital mammography.
• This advancement holds potential for improving the efficacy of computer-aided detection systems in breast cancer screening.
• The proposed method provides a significant leap in segmentation performance, aiding radiologists in diagnosis.