Naoki Kamiya^1,, Norimitsu Shinohara²

Data in Japan shows that the risk of developing breast cancer increases after the age of 40 and peaks in the late 40s. The most common method of breast cancer screening in Japan is through mammograms, and in recent years, experts have considered combining mammograms with an ultrasound to increase the detection rate of breast cancer. Meanwhile, in the United States, physicians alert the patients of their mammary gland density after the mammogram. The physician offers the possibility of tumors being covered up by the mammary tissue, and use this data to determine the appropriate interval between check-ups. However, this advice based on mammary gland density relies on the physician's visual assessment, and reproducibility remains a challenge. Software that quantitatively evaluates mammary gland density is already commercially available, but is optimized for the Western population. This study aims for the automatic evaluation of mammary gland density of Japanese subjects. We define an evaluation index of the mammary gland amount based on the breast thickness obtained from the DICOM header, and the characteristic amount of breast tissue measured through image analysis. We verified the accuracy of the proposed indicator in its ability to correctly classify mammary gland density across 458 cases, and found it consistent with the physician's classification in 98.5% of the cases. In the future, we look to create an index to calculate average glandular dose (AGD) based on this index.

mammary gland density, mammary gland, breast thickness, architectural evaluation, computer-aided diagnosis