Changes in AI risk scores over time from screening mammograms can be used to stratify women who will develop breast cancer and those who will not. The change in scores was considered by the authors of a multisite retrospective study published in Radiology, to be a dynamic biomarker for risk-adaptive screening approaches and prevention strategies.

“With the power of AI, computer vision, and the ability to extract predictive data, we are able to apply the power of imaging to risk assessment and preventing disease from developing,” said lead researcher Constance D. Lehman, MD, PhD, Professor of Radiology at Harvard Medical School in Boston and CEO and Founder of Clairity, Inc. The company’s Clairity Breast deep learning platform was recognized as the first FDA-authorized AI platform for predicting a woman’s 5-year risk of breast cancer and is now commercially available at select institutions. “Having a dynamic risk score opens up a whole new domain of more effective preventive therapies for breast cancer, similar to how we screen for and treat patients with high cholesterol and hypertension.”

Lehman et al conducted a longitudinal analysis of 158,807 screening mammograms from 54,014 women who were screened between January 2009 and December 2019 across six imaging sites. All exams were standard 2D bilateral full-field digital mammography screening exams acquired with or without digital breast tomosynthesis. They compared successive scans from women who were diagnosed with breast cancer within a year of the index examination (n = 817), defined as the final screening mammogram within 1 year prior to a breast cancer diagnosis or the final mammogram in the 5-year study period for cancer-free controls, with cancer-free controls (n = 53,197). The researchers gathered a median of three screening mammograms per woman.

An image-only deep learning model, Mirai, generated continuous 5-year breast cancer risk scores. Image pixels were the only model inputs; no demographic, clinical, or historical imaging data were used. Score trajectories were tracked over time.

The median risk score prior to diagnosis was 2.1 in women who developed breast cancer, which increased to 6.6 at the index examination, while cancer-free controls had stable median scores between 1.8 and 2.2.

“We observed clinically relevant differences in risk trajectories between women who did and did not develop cancer,” Dr. Lehman said. “The increase in scores among cancer patients was detectable as early as 6 years prior to diagnosis and became more pronounced over time.”

Scores increased over time in longitudinal models for women who developed cancer (slope = 1.13 per year; 95% confidence interval [CI] = 1.07–1.18; P < .001). In the cancer-free control group, scores changed minimally (slope = 0.09 per year; 95% CI = 0.08–0.10; P < .001). The difference between slopes was 1.04 (95% CI = 0.99–1.09; P < .001).

Additionally, the slope of score changes increased more steeply in the years immediately preceding the cancer diagnosis, while cancer-free trajectories remained essentially flat across the study period.

“These trends remained robust across subgroups defined by age and breast density, further supporting the generalizability of our findings,” Dr. Lehman said. “This is particularly relevant given persistent disparities in screening performance across patient populations. A dynamic biomarker approach grounded in the imaging data could mitigate some of these disparities by enabling risk-based personalization that does not rely on self-reported or inconsistent clinical data.”

“AI-derived risk scores can identify patients who are otherwise predisposed to the disease, and our findings demonstrate that image-based AI risk scores evolve over time and that changes in those scores may provide additional information about future breast cancer risk,” she concluded.

DISCLOSURES: For full disclosures of the study authors, visit pubs.rsna.org.

ASCO AI in Oncology is published by Conexiant under a license arrangement with the American Society of Clinical Oncology, Inc. (ASCO^®). The ideas and opinions expressed in ASCO AI in Oncology do not necessarily reflect those of Conexiant or ASCO. For more information, see Policies.