Morphological Area Gradient: System-independent Dense Tissue Segmentation in Mammography Images.

Area of Science:

• Radiology
• Medical Imaging
• Computational Pathology

Background:

• Breast density is a significant breast cancer risk factor.
• Accurate segmentation of dense tissue in mammograms is challenging.
• Current methods often require system-specific calibration or raw image data.

Purpose of the Study:

• To introduce a generic and robust method for automatic dense tissue segmentation in mammograms.
• To develop a measure that does not require calibration or access to raw mammograms.
• To improve the reliability of breast density assessment for breast cancer risk evaluation.

Main Methods:

• The Morphological Area Gradient (MAG) was developed as a novel measure for mammography images.
• MAG is based on the derivative of segmented tissue area with respect to pixel intensity.
• High-density regions were segmented by minimizing the MAG of mammograms.

Main Results:

• The MAG method demonstrated superior performance compared to state-of-the-art algorithms.
• A median absolute error of 7.6% was achieved.
• A Dice similarity coefficient of 0.83 was obtained using 566 full-field digital mammograms.

Conclusions:

• The Morphological Area Gradient (MAG) provides a generic and effective approach for dense breast tissue segmentation.
• This method overcomes limitations of existing techniques, offering improved accuracy and reproducibility.
• MAG has the potential to enhance breast cancer risk assessment through more reliable mammogram analysis.