Women who have an abnormal mammogram often have to wait several weeks to find out if they have breast cancer.
Now, researchers at the University of California, San Francisco and the University of California, Berkeley have discovered a way to reduce wait times and anxiety by using AI to quickly identify those most likely to contract the disease. By triaging these patients, AI-guided workflows can complete the diagnostic process from imaging to evaluation and even biopsy in a single day.
“This is a really exciting time,” said Maggie Cheung, MD, lead author of the study published May 19. nature digital medicine. “This brings us closer to personalized care and allows us to tailor plans to ensure each patient receives the right intervention at the right time.”
The researchers used an open-source AI model called Mirai developed by the study’s lead author, Adam Yara, Ph.D., a data scientist at the University of California, Berkeley. After being trained on hundreds of thousands of mammograms associated with patient cancer outcomes, the model now recognizes subtle patterns in screening mammograms and can predict a woman’s cancer risk in a more powerful way than doctors working alone.
Chung and Yala applied the model to more than 4,100 screening mammograms performed at Zuckerberg San Francisco General Hospital and Trauma Center. Mirai determined that 525 women, or approximately 12.7% of screened patients, were at high risk.
These patients can have their mammograms interpreted immediately after they are taken and receive additional imaging the same day if there are any suspicious areas. Some women who needed a biopsy were able to have it done on the same day.
Mirai has reduced wait times for diagnostic evaluations from several weeks to about an hour. And for those ultimately diagnosed with breast cancer, Mirai reduced the average wait time for biopsy from more than two months to less than 10 days.
Mirai does not replace a radiologist and does not perform its own diagnosis. Rather, it is a triage tool that helps doctors identify patients who would most benefit from prompt treatment.
“This is a powerful example of how AI can be a collaborative partner for physicians,” Yala said. Yala, along with Chung, is an assistant professor in the UCSF-UC Berkeley Joint Program in Computational Precision Medicine. “This shows how we can improve care by bringing clinicians and data scientists together to design systems.”
Before launching the program, researchers analyzed more than 114,000 archival mammograms to ensure the model could capture enough high-risk patients without overloading clinics with too many rapid assessments.
Researchers hope that AI will facilitate a more personalized approach to breast cancer screening, tailored to each patient’s breast cancer risk.
Although many women currently follow the same testing schedule, individual risks can vary widely. AI risk assessment gives us the opportunity to identify the women most likely to benefit from prompt care and provide them with what they need. ”
Maggie Cheung, MD, lead author of the study
sauce:
University of California, San Francisco
Reference magazines:
Chung M. others. (2026). Future deployment of AI-based risk stratification will enable faster mammography workflows in safety net environments. npj digital medicine. DOI: 10.1038/s41746-026-02743-x. https://www.nature.com/articles/s41746-026-02743-x
