Can an Emerging AI Software for DBT Help Reduce Disparities in Breast Cancer Screening?

One of the key challenges with the development of artificial intelligence (AI) tools in radiology has been a lack of representative diversity in test sets used to assess the performance of AI algorithms. However, new research with an emerging software for digital breast tomosynthesis (DBT) demonstrated nearly equivalent sensitivity for breast cancer detection across different races and ethnicities.

For the study, presented at the Radiological Society of North America (RSNA) conference, researchers evaluated the Genius AI Detection 2.0 software (Hologic) in a review of DBT data for over 7,500 women. The cohort included 5,395 White women, 956 Black women, 774 Hispanic women and 279 Asian women, according to the study.

The researchers found that the AI software offered an overall sensitivity rate of 90.1 percent, including similar sensitivity rates for White (90.3 percent), Black (88.6 percent), Hispanic (90.1 percent) and Asian women (91.5 percent).

“I was really happy to see that the Genius AI (Detection) 2.0 (software) was able to perform similarly in all of the populations studied albeit with a (slightly higher) improvement in cancer detection in the Asian population,” noted lead study author Sarah Friedewald, M.D., the vice chair for women’s imaging and chief of breast imaging in the Department of Radiology at the Northwestern Medicine Feinberg School of Medicine.

(Editor’s note: For additional coverage of RSNA, click here.)

In a recent interview, Dr. Friedewald shared her perspective on the study findings and the potential of AI in effectively triaging patients who need supplemental imaging. While she emphasized the need for continued research in real-world settings, Dr. Friedewald said AI can play a key role in mitigating disparities in breast cancer detection and care.

“As long as we are cognizant of our test sets and that we continually monitor the performance of the artificial intelligence, we will be confident that we are serving our patients equally and that everybody will benefit from the technology,” emphasized Dr. Friedewald.

(Editor’s note: For related content, see “Mammography Study: Paying for Adjunctive AI Screening Led to 43 Percent Higher Rate of Breast Cancer Detection,” “FDA Clears Updated AI Platform for Digital Breast Tomosynthesis” and “AI Mammography Platform Shows Promising Results for Detecting Subclinical Breast Cancer.”)

For more insights from Dr. Friedewald, watch the video below.

Reference

1. Friedewald SM, Kshirsagar A, Smith AP, Pohlman S. Performance of a digital breast tomosynthesis AI detection algorithm in common US racial/ethnic groups. Poster presented at the Radiological Society of North America (RSNA) 2024 110^th Scientific Assembly and Annual Meeting Dec. 1-5, 2024. Available at: https://www.rsna.org/annual-meeting .