Incorporating AI into mammography analysis could reduce the number of breast cancers developed between screenings by 30%, according to a study published April 18 in the Journal of the National Cancer Institute.
Researchers at the University of California, Los Angeles analyzed 185,000 mammograms performed between 2010 and 2019 for the study.
Here are four things to know from the research:
- Of the 185,000 mammograms, researchers identified 148 cases where breast cancer developed between screenings, also called interval breast cancers.
- These mammograms were screened by the AI software Transpara for signs of cancer and scored on a 1 to 10 scale of cancer risk, with scores of 8 or higher being flagged as “potentially concerning,” according to a May 5 news release from the university.
- The AI software flagged 76% of the mammograms later linked to interval breast cancers as potentially concerning and 90% of cases where visible cancer was missed or misinterpreted by the radiologist.
It also found about 89% of cases where there were subtle signs of cancer that could have reasonably been acted on and 72% of cases where the signs of cancer were too subtle to act on.
The software flagged 69% of cases where the cancer was completely invisible on the mammogram and 50% of true interval cancers. - “While AI isn’t perfect and shouldn’t be used on its own, these findings support the idea that AI could help shift interval breast cancers toward mostly true interval cancers,” Tiffany Yu, MD, assistant professor of radiology at the David Geffen School of Medicine at UCLA and lead study author, said in the release. “It shows potential to serve as a valuable second set of eyes, especially for the types of cancers that are the hardest to catch early.”
Read the full study here.
