AI continues to charge into healthcare with gusto, and now two West Coast institutes of higher education are collaborating on a $16 million trial to evaluate the effectiveness of the technology in detecting the presence of breast cancer.
“This is the first large-scale randomized trial of AI in breast cancer screening in the United States,” said Dr. Joann G. Elmore, dual principal investigator and professor of medicine at UCLA. “We’re looking carefully and objectively at whether AI helps or hinders — and for whom. Expert radiologists remain in the driver’s seat for all interpretations.”
UCLA and UC Davis will co-lead the study, which is focused on evaluating whether AI can help support radiologists in interpreting mammograms more accurately, “with the goal of improving breast cancer screenings and reducing unnecessary callbacks and anxiety for patients.” The study is called PRISM, for Pragmatic Randomized Trial of Artificial Intelligence for Screening Mammography.
The trial is straightforward. Participating medical centers will read hundreds of thousands of mammograms — in California, Florida, Massachusetts, Washington and Wisconsin — as they normally would. Behind the scenes, each exam will be randomly assigned to one of two paths: a radiologist working alone, or a radiologist aided by Transpara, an FDA-cleared AI support tool integrated into clinical workflow with Aidoc’s platform. Either way, humans make the final call.
For patients, nothing changes. No added steps, no new machines, no awkward conversations about algorithms. But the implications could reshape how mammograms are read nationwide.
“AI has great promise, but it also raises real questions,” Dr. Elmore said. “We want to know whether AI helps radiologists find more cancers, or just flags more exams that ultimately turn out to be normal.”
That tension echoes through AI in medicine. On one end are researchers teaching algorithms to discover hidden red flags in routine scans, such as the Massachusetts General Brigham team that trained an AI system to pick up signs of coronary artery calcium from ordinary chest CTs. On the other end, some studies are raising alarms about what happens when humans begin trusting machines a little too much. In Poland, where endoscopists adopted a polyp-detecting AI system in 2021, researchers observed a troubling pattern: doctors actually found fewer abnormalities on their own after working with AI. “We were quite surprised,” said Dr. Marcin Romańczyk, who led the study. A 6% absolute decrease in detection suggested a subtle but real “deskilling” effect — a risk that becomes especially relevant as AI spreads across specialties.
The PRISM trial, expected to run for two years, is a fact-finding mission. “There’s never been a trial of this scope looking at AI in breast cancer screening in the U.S.,” said Dr. Hannah Milch, co-principal investigator and assistant professor of radiology at UCLA. “The results will help inform not just clinical practice, but also insurance coverage, technology adoption, and patient communication.”
UC Davis will lead the study’s Data Coordinating Center. “We’re rigorously evaluating whether AI-assisted interpretation improves screening outcomes,” said Dr. Diana Miglioretti, dual principal investigator and chief of biostatistics at UC Davis. “The goal is not to replace radiologists with AI but to see how effective AI could be as a co-pilot in reading mammography.”
Breast cancer remains the second leading cause of cancer death among women in the U.S. Screening works, but it is imperfect. False positives create spirals of stress; missed cancers can be deadly. The messaging during Breast Cancer Awareness Month, October, highlights early detection, and for women over 40, annual mammograms remain the standard.
“This is our opportunity to generate trustworthy evidence — with the patient perspective front and center,” Miglioretti said.
