Women with abnormal mammograms often have to wait weeks to find out if they have breast cancer.
Now, researchers at UC San Francisco and UC Berkeley have found a way to help reduce the wait and worry by using artificial intelligence to quickly identify those most likely to have the disease. By triaging these patients, AI-driven workflow moves women with abnormal scans through the diagnostic process—from imaging to evaluation and sometimes even biopsy—in one day.
“This is a really exciting time,” said Maggie Chung, MD, first author of the study, which was published May 19 in Nature Digital Medicine. “This brings us closer to personalized care, where we can tailor a plan so that each patient receives the right intervention at the right time.”
The researchers used an open-source artificial intelligence model called Mirai, which was developed by the study’s senior author, UC Berkeley data scientist Adam Yala, PhD. After being trained on hundreds of thousands of mammograms linked to patients’ cancer outcomes, the model can recognize subtle patterns in a screening mammogram and predict a woman’s cancer risk more powerfully than a doctor working alone.
Chung and Yala applied the model to more than 4,100 screening mammograms at San Francisco Zuckerberg General Hospital and Trauma Center. Mirai found that 525 women – about 12.7% of patients screened – were at high risk.
These patients could receive an interpretation of their mammograms immediately after they were performed and undergo additional diagnostic imaging for any suspicious areas on the same day. Some of the women who needed biopsies were able to do so on the same day.
Mirai has reduced the wait time for a diagnostic evaluation from several weeks to about an hour. And for those eventually diagnosed with breast cancer, Mirai reduced the average wait for a biopsy from more than two months to less than 10 days.
Mirai does not replace radiologists – nor does it make diagnoses on its own. Instead, it’s a screening tool that helps doctors identify patients who can benefit most from expedited care.
“This is a powerful example of how artificial intelligence can become a collaborative partner for physicians,” said Yala, who along with Chung is an assistant professor in the Joint UCSF-UC Berkeley Program in Computational Precision Health. “It shows how we can improve care when we bring clinicians and data scientists together to design these systems.”
The researchers analyzed more than 114,000 archival mammograms before starting the program to ensure the model would capture enough high-risk patients without overwhelming the clinic with too many rapid assessments.
The researchers hope that AI will promote a more personalized approach to breast cancer screening that is tailored to each patient’s breast cancer risk.
Right now, many women follow the same screening schedule, but individual risk can be very different. AI risk assessment gives us the opportunity to identify the women most likely to benefit from rapid care and get them what they need.”
Maggie Chung, MD, first author of the study
Source:
Journal Reference:
