A new computer algorithm assessment was developed that estimates mammographic breast density. Results obtained by the computer assessment were compared using three radiologists’ visual breast density ratings. “The computer ratings (average of four mammograms) yielded a strong correlation (.86) with the reader’s overall ratings suggesting that our algorithm correlated highly with the ratings of our human observers; although we noted that human observers tended to consistently underestimate overall breast density,” said Naomi Saenz, MD, lead author of the study.

“Breast density has been shown to be an important risk factor for breast cancer. It is evaluated visually and subjectively in current practice and we found that assessment of breast density can vary from one radiologist to the next. Our algorithm automatically segments the breast tissue from its background; removes the pectoral muscle; and then uses a region growing method to determine the borders of the areas of the breast that are dense and separates them from other breast tissue and fat. The percentage of tissue over the region of the breast that is identified and categorized as “dense”, excluding the pectoral muscle, is calculated and referred to as mammographic breast density,” she said.

“This method may help physicians give more objective and accurate recommendations on who will need further imaging or who might not; as well as reduce variability from one reader to the next. It could easily be implemented into a PACS or mammographic reading station so that the radiologist would have this information on hand while reading and dictating,” said Dr. Saenz.

This study will be presented at the 2009 ARRS Annual Meeting in Boston, MA, on Wednesday, April 29.

Note: If no author is given, the source is cited instead.

• Breast Cancer
• Women's Health
• Breastfeeding

• Computer Programming
• Computer Science
• Information Technology

• Breast reconstruction
• Mammography
• BRCA1
• Colostrum
• Lavender oil
• Metastasis