Marker-controlled watershed for lesion segmentation in mammograms

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Image Segmentation

Background:

• Lesion segmentation in mammograms is crucial for computer-aided diagnosis but challenging due to obscured, irregular, and low-contrast boundaries.
• Existing segmentation methods face difficulties with the inherent complexities of breast lesion imaging.

Purpose of the Study:

• To propose an accurate and robust algorithm for automatic breast lesion segmentation in mammograms.
• To enhance the reliability of lesion quantification in mammographic analysis.

Main Methods:

• A novel algorithm utilizing watershed transformation on a smoothed morphological gradient image.
• Automatic determination of internal and external markers through template matching, thresholding, distance transform, and morphological dilation.
• Quantitative comparison with dynamic programming and plane fitting methods using Area Overlap Metric (AOM), Hausdorff Distance (HD), and Average Minimum Euclidean Distance (AMED).

Main Results:

• The proposed algorithm achieved superior performance compared to two other segmentation methods.
• Quantitative metrics showed a mean AOM of 0.72 ± 0.13, mean HD of 5.69 ± 2.85 mm, and mean AMED of 1.76 ± 1.04 mm.
• Demonstrated reliable segmentation and quantification capabilities for breast lesions.

Conclusions:

• The developed algorithm offers a significant advancement in automatic breast lesion segmentation.
• The findings confirm the algorithm's potential for reliable clinical application in mammography.
• This method can aid in more accurate breast cancer diagnosis and monitoring.