A massive new breast cancer study is about to put artificial intelligence to the test in real clinics, with radiologists and algorithms sharing the spotlight. In a nationwide trial, screening mammograms will be read either by a radiologist alone or by a radiologist using an AI support tool to determine whether the software helps catch more cancers while reducing unnecessary callbacks. The project, known as PRISM (Pragmatic Randomized Trial of Artificial Intelligence for Screening Mammography), is co-led by UCLA and UC Davis and is expected to cover hundreds of thousands of exams at academic centers across several states. Organizers put the effort at roughly $16 million and emphasize that patients' views are central, with human radiologists still making every final call.
According to PCORI, the award for PRISM carries a budget of approximately $15.8 million. It includes a formal, multi-year project period to support enrollment across multiple centers, as well as detailed outcome analyses. The agency lists the study under its full title and notes that the funding was approved earlier this year.
How PRISM will run
Each screening exam in PRISM will be randomized so that about half of the mammograms are interpreted with an FDA-cleared AI decision-support tool activated, and the other half are read without it. In all cases, a radiologist will still issue the official interpretation. The trial will track cancer detection rates and recall rates, and it will also utilize surveys and focus groups to capture the actual thoughts of patients and radiologists regarding AI in the reading room. By pairing hard clinical outcomes with patient-centered measures, investigators hope to inform how AI is adopted, covered by payers, and explained to the public, according to The ASCO Post.
Sites and scale
The PRISM network pulls together seven academic medical centers and breast imaging sites across California, Florida, Massachusetts, Washington, and Wisconsin. The University of Wisconsin–Madison estimates that about 50,000 participants will be involved in its state alone and notes that the first phase is expected to run for roughly two years, with the overall study set to interpret hundreds of thousands of screening exams nationwide, according to UW–Madison Radiology.
Which AI tool will be tested
ScreenPoint Medical has announced that its Transpara system has been selected as the AI support tool for PRISM and will be integrated into clinical workflows through Aidoc's aiOS platform, according to the company. The firm notes that Transpara has already been used on millions of mammograms internationally, while study organizers say PRISM will serve as the first large-scale, randomized trial of an AI system in U.S. breast cancer screening practice, according to the same press release.
What researchers want to know
"The goal is not to replace radiologists with AI but to see how effective AI could be as a co-pilot in reading mammography," said principal investigator Diana Miglioretti. Investigators say they want randomized evidence on whether AI helps radiologists detect more cancers without increasing false positives, and they aim to understand better how both patients and clinicians perceive AI tools in everyday care.
Why this matters locally
For California patients, the heavy presence of in-state sites means PRISM's findings could quickly shape what happens at local hospitals and imaging centers, including whether AI becomes a routine reading aid and how doctors discuss it in the exam room. UCLA researchers say that if AI proves clearly beneficial or clearly problematic in real-world settings, the results will have immediate implications for clinical practice, payer coverage decisions, and patient communication, according to UCLA Health.
PCORI notes that funding for PRISM was structured over multiple years to support ongoing outcome tracking and patient-reported data collection. Local investigators add that the initial phase alone will span about two years and involve tens of thousands of participants at some sites, including an estimated 50,000 in Wisconsin, according to UW–Madison Radiology.
