AI to Predict Breast Cancer in 5 Years — FDA Declares Technology a Breakthrough

A new technology that uses artificial intelligence (AI) to analyze mammograms and improve the accuracy of predicting a woman’s individual five-year risk of breast cancer has received Breakthrough Device designation from the U.S. Food and Drug Administration (FDA). The technology, developed by researchers at Washington University School of Medicine in St. Louis, has been licensed to Prognosia Inc., a University of Washington startup.

The system analyzes mammograms and produces a risk score that determines the likelihood that a woman will develop breast cancer within the next five years. The technology is compatible with both types of mammographic images: four 2D images of the breast obtained by full-field digital mammography and a synthetic 3D image of the breast obtained by digital breast tomosynthesis.

Importantly, the system provides an absolute five-year risk that compares a woman’s risk to the average risk based on national breast cancer rates. This provides a meaningful estimate that is consistent with U.S. national risk reduction guidelines, so clinicians know what steps to take if a woman has an elevated risk.

The FDA's Breakthrough Device designation provides an expedited review process for full market approval to provide patients and physicians with faster access to new medical devices. Products that receive this designation have already undergone rigorous testing and have demonstrated high potential to improve the treatment or diagnosis of debilitating or life-threatening conditions.

The software package, called Prognosia Breast, was developed by Graham A. Colditz, MD, PhD, the Naiss-Hein Professor of Surgery and associate director of prevention and control at the Siteman Cancer Center at Barnes-Jewish Hospital and the University of Washington School of Medicine, and Shu (Joy) Jiang, PhD, associate professor of surgery in the Department of Public Health at the University of Washington School of Medicine.

Kolditz and Jiang co-founded Prognosia in 2024 in collaboration with the University of Washington's Office of Technology Management (OTM) and BioGenerator Ventures, the latter of which provided both financial backing and business strategy expertise from Entrepreneur-in-Residence David Smoller, Ph.D.

The software is a pre-trained machine learning system that analyzes mammograms and produces an estimate of the likelihood of developing breast cancer over the next five years, based only on the images and the woman’s age. According to the developers, Prognosia Breast estimates the five-year risk of developing breast cancer 2.2 times more accurately than the standard method, which relies on a questionnaire that takes into account factors such as age, race, and family history.

The system was trained on previous mammograms of tens of thousands of women who had been screened for breast cancer at Siteman Cancer Center. Some of them later developed cancer, allowing the system to “learn” to recognize early signs of tumor development — signs that even a highly experienced doctor would not notice.

“We are excited about the potential of this technology to improve breast cancer risk prediction and prevention on a large scale – no matter where a woman is screened,” Colditz said. “The long-term goal is to make this technology available to every woman who is screened mammogram-wise, anywhere in the world.”

"Regardless of the type of image obtained, our data show the potential of the software to identify women at increased risk of developing breast cancer over the next five years, giving them the opportunity to take targeted steps to reduce that risk."

The new device could have a significant impact on risk prediction because the infrastructure already exists to immediately deploy the software anywhere mammography is performed. In addition, many women already get regular mammograms. According to a 2023 survey from the Centers for Disease Control and Prevention (CDC), more than 75% of women ages 50 to 74 reported having had a mammogram in the past two years.

Even with widespread screening, about 34% of women diagnosed with breast cancer in the United States are diagnosed at late stages of the disease. The ability to estimate risk five years before disease onset is likely to improve early detection, reducing the number of late-stage diagnoses, according to the researchers. Early detection has been shown to improve treatment effectiveness and reduce breast cancer mortality.

“Receiving the Breakthrough Device designation is a powerful recognition of this research team’s exceptional dedication and vision to improve the diagnosis and treatment of breast cancer,” said Doug E. Franz, Ph.D., vice chancellor for innovation and commercialization at the University of Washington.

“Creating software that can be quickly integrated into the workflow of any mammography center takes years of dedicated work. It significantly improves the clinical value of routine mammograms, no matter where they are performed. It is a prime example of the important role entrepreneurship and commercialization play at the University of Washington in translating cutting-edge research into real-world technologies that improve patient care.”

The device provides a five-year risk score that is intended to complement, not replace, the analysis provided by radiologists, who will continue to review mammograms according to standard protocols. A five-year risk of 3% or higher is considered elevated, according to the American Society of Clinical Oncology and the U.S. Preventive Services Task Force. Women with elevated risk scores should be referred to specialists who can further counsel them about additional screening and prevention strategies, the groups recommend.

About 1 in 8 women in the United States will be diagnosed with breast cancer in their lifetime. Women at high risk may receive more frequent screening, which may include other imaging tests, such as MRI, and in some cases, chemotherapy drugs, such as tamoxifen, or endocrine therapy as a preventive measure. When these options are available, it is important to identify women at high risk so they can access specialists who can help them make important choices.

The team is planning a clinical trial at Siteman Cancer Center that will use the Prognosia Breast risk assessment in conjunction with standard mammography screening protocols. Standard protocols include mammogram reviews and breast density assessments, which are already provided to all patients. Women at increased risk will be referred to breast health specialists who help patients navigate their options for managing high breast cancer risk.

“Despite the fact that modern breast imaging is high-tech and widely used to detect existing tumors, today’s breast cancer risk prediction is still questionnaire-based and does not estimate future risk very well,” Jiang said. “Our work focused on addressing this shortcoming. Moving to image-based risk prediction, which our research shows is much more accurate, could revolutionize patient care.”

The FDA's current designation applies to the analysis of mammogram images taken at a single point in time. In the future, the researchers plan to update Prognosia Breast so that the system can analyze mammograms from the same patient over multiple years, which could further improve the accuracy of predictions.