AI-generated risk scores help predict future breast cancer cases
Breast cancer risk scores change over time, with those who go on to develop the disease showing unique patterns that could provide valuable guidance into their diagnostic needs.
That’s according to new research published in RSNA’s flagship journal Radiology. The study compared longitudinal data from more than 54,000 women to their artificial intelligence-generated risk scores to analyze how long-term changes related to cancer. Researchers determined the risk scores of the women who went on to develop cancer steadily rose throughout several years before increasing significantly the year prior to diagnosis.
“Deep learning models have been primarily used to assess cancer risk scores at a static point in time,” lead researcher Constance D. Lehman, MD, PhD, professor of radiology at Harvard Medical School and CEO of vendor Clairity Inc., said in a June 23 news announcement from the Radiological Society of North America. “In this study, we evaluated longitudinal changes in the image-only deep learning breast cancer risk score using serial mammograms from a large screening cohort.”
For their study, researchers applied a validated, open-source, deep learning model that produced five-year risk scores based on the mammograms of more than 158,000 women. The exams were completed between 2009 and 2019, with each woman having up to six serial mammograms plus an index exam (defined as the final screening mammogram within the year prior to their cancer diagnosis) available for AI evaluation. The team analyzed the risk scores of each exam to assess how they changed over time in those who were diagnosed with cancer.
Just 817 of the 54,014 women included in the analysis were diagnosed with cancer. Of those cases, 55% were invasive, 14% were ductal carcinoma in situ (DCIS) and 30% were unknown. About 83% were screen-detected, while 17% were interval cancers.
Those who were with the disease exhibited consistent changes in their risk scores throughout the six years that preceded their diagnosis. Median scores increased from 2.1 in the first five to six years of the study period to 6.6 by the time the index exam was completed. Scores increased the most during the years closest to the index exam. In contrast, those who did not receive a cancer diagnosis showed stable scores throughout the entire study period.
What’s more, these findings were consistent across all subgroups, including different breast density categories.
“This is particularly relevant given persistent disparities in screening performance across patient populations,” Lehman said. “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.”
Read more from the study here.
In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She began covering the medical imaging industry for Innovate Healthcare in 2021.