The Potential of Mammography Image-Based AI Models for Assessing Risk: An Interview with Constance Lehman, MD

Could an image-based artificial intelligence (AI) model offer better accuracy than breast density in risk stratification for breast cancer?

In new multicenter research recently presented at the Radiological Society of North America (RSNA) conference, researchers compared breast density assessment versus an image-only AI model (Clairity Breast, Clairity) for the evaluation of five-year risk for breast cancer in a cohort that included 245,344 bilateral 2D screening mammograms. Employing National Comprehensive Cancer Network (NCCN) thresholds for AI-predicted, the study authors defined average risk as less than 1.7 percent, intermediate risk between 1.7 and 3 percent, and high risk as greater than 3 percent.

For the AI model, the researchers demonstrated a significantly elevated incidence of breast cancer incidence in the high-risk group in comparison to women with average risk (5.9 percent vs. 1.3 percent). However, the study authors pointed out a 3.2 percent incidence for those with dense breasts versus 2.7 percent for women with non-dense breasts.

In a recent interview at the RSNA conference, study co-author Constance Lehman, M.D., Ph.D., emphasized that breast density alone is insufficient for assessing long-term risk stratification for breast cancer.

“It's (a) very minimal increased risk, not one that you'd want to use to change recommendations for an individual patient. But on the other hand, when we looked at our high-risk patients (with) greater than 3 percent five-year risk by our AI risk model, those women were 4.5 times more likely to develop breast cancer in (the) subsequent five years than their counterparts with an average AI risk score,” pointed out Dr. Lehman, a professor of radiology at Harvard Medical School and co-director of the Breast Imaging Research Center at Massachusetts General Hospital in Boston.

Should image-based AI models be included in guidelines for assessing breast cancer risk?

“ … While it's important for women to know their density, it's a very weak predictor of risk and density is in the guidelines. We have something that we and others have demonstrated is significantly more performant than what is now in the guidelines, and we are very hopeful this will be included in the guidelines,” noted Dr. Lehman, the co-founder of Clairity, Inc. “We submitted our proposal to suggest that the (NCCN Breast Cancer Screening and Diagnostic Panel) consider this, and we're really eager to work closely with them so that we can have this available to more women.”

(Editor’s note: For related content, see “Mammography AI Platform for Five-Year Breast Cancer Risk Prediction Gets FDA De Novo Authorization,” “RSNA: Study Suggests Significant Incidence of Invasive Breast Cancer in Young Women” and “Mammography Study: Multi-Stage Use of AI for DBT Exams Yields Over 21 Percent Increase in Breast Cancer Detection.”)

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