Advancements in Developing an Automated Breast Density Detection Technique for Breast Cancer Risk Prediction:...

Area of Science:

• Radiology
• Medical Imaging
• Biostatistics

Background:

• Breast density is a key breast cancer risk factor.
• Accurate estimation of breast density from mammograms is crucial for risk prediction.
• Previous automated methods faced performance degradation due to signal-dependent noise (SDN) variations.

Purpose of the Study:

• To enhance an automated percentage of breast density detection method.
• To address performance degradation caused by SDN variations across different image data representations.
• To improve the accuracy and applicability of breast density estimation for breast cancer risk prediction.

Main Methods:

• Utilized signal-dependent noise (SDN) analysis, relating expected variance and mean.
• Developed capabilities to transform noise to an optimized quadric SDN form.
• Implemented ensemble averaging over images to boost signal.
• Applied probability density transformation for combining measurements from different mammographic technologies.

Main Results:

• The enhanced algorithm produced significant odds ratios across all tested image data representations.
• Demonstrated improved performance on full field digital mammography (raw and clinical) and digital breast tomosynthesis images.
• Showcased the ability to combine measurements from different imaging technologies effectively.

Conclusions:

• The enhanced automated breast density detection method is robust across diverse mammographic technologies.
• The advancements allow for accurate breast cancer risk prediction using mammograms.
• The technique is suitable for both research and clinical applications with minimal adjustments.