How BCRF Researchers Are Using AI to Predict Breast Cancer Before It Starts

| November 19, 2026

BCRF-supported investigators are researching AI’s use in personalized prevention tactics

Mammography has revolutionized early breast cancer detection. But what if doctors could predict who is likely to develop breast cancer years before it happens? Could we actually personalize prevention by tailoring it to each individual? The Breast Cancer Research Foundation (BCRF) is supporting investigators that are building the tools to do just that, and artificial intelligence (AI) is at the center.

Key Takeaways

  • BCRF-funded scientists are working to predict breast cancer early enough to prevent it, an approach called precision prevention.
  • A BCRF-supported study demonstrated that AI-determined risk scores began rising five to six years before diagnosis.
  • Researchers are testing whether AI risk scores could also show when prevention is working.
  • AI could help combine mammograms, genes, and personal history into one comprehensive picture of risk.

From AI-assisted breast cancer detection to prediction

There are several breast cancer screening methods currently in practice, including mammography, ultrasound, and MRI, which have significantly improved early breast cancer detection. While screening helps find breast cancer early, it does not predict someone’s risk of developing the disease.

Risk prediction techniques aim to determine breast cancer risk before the disease can take root. Knowing a person’s risk helps doctors identify prevention strategies that can be tailored to the individual, known as precision prevention. BCRF is supporting research that leverages the use of artificial intelligence to improve risk prediction.

In 2019, BCRF established the Precision Prevention Initiative (PPI), a $5M initiative underwritten by Blizzard Entertainment. Some projects in phase 1 of this initiative were designed to use AI and mathematical tools to build personalized risk models.

BCRF investigators Drs. Regina Barzilay and Adam Yala at the Massachusetts Institute of Technology (MIT), Dr. Constance Lehman at Massachusetts General Hospital (MGH), and their teams developed an AI-based breast cancer risk model called MIRAI, which can be used to analyze mammograms to predict future breast cancer risk. Out of this work, several follow-on studies have been initiated.

Putting AI prediction to the test: Drs. Regina Barzilay and Adam Yala

Continued BCRF-funding enabled Drs. Barzilay, Yala, and their team to conduct international validation studies through the AI Screening Project, supported by the Zeta Tau Alpha Foundation.

Breast MRI is one of the most sensitive tools available for early breast cancer detection and can substantially reduce missed cancers when added to screening mammography. The team seeks to refine methods to determine who should receive supplemental MRI alongside a regular mammogram. Prior studies have shown that more than 75% of patients with breast cancer would not have qualified for MRI screening before diagnosis, representing major missed opportunities for earlier detection. Existing guidelines rely primarily on questionnaire-based risk models, which are not able to help identify a significant number of women who later develop breast cancer.

MIRAI demonstrated substantially improved accuracy compared with traditional questionnaire-based risk models. The team is now conducting the first prospective multi-site clinical studies in the U.S. that evaluate whether AI-guided risk assessment can improve selection of women for supplemental MRI screening. These studies test MIRAI on women’s mammograms going forward, not on past records,and more than 100,000 screening mammograms have now undergone AI-based risk assessment. Early findings demonstrate a high cancer detection rate (20.7/1000) with MIRAI in high-risk women whose screening mammograms had previously been interpreted as negative or benign.

Ongoing studies continue to evaluate the generalizability and clinical utility of AI-guided supplemental MRI screening across diverse healthcare systems and patient populations. This marks an important step toward more personalized and effective breast cancer screening strategies that can improve early detection and patient outcomes.

Reading risk as it changes: Dr. Connie Lehman

Dr. Lehman built on her PPI work to develop Clairity Breast, the first FDA-authorized AI-driven breast cancer risk assessment tool. This authorization was based in part on BCRF-supported studies. Clairity Breast can be used to analyze a woman’s screening mammogram to predict her chances of developing breast cancer within the following five years. On the heels of the authorization, the National Comprehensive Cancer Network® (NCCN) changed their 2026 Clinical Practice Guidelines screening: Women who are 35 and considered high-risk based on an AI risk prediction tool such Clairity Breast are recommended to begin annual screening.

With BCRF support, Dr. Lehman also evaluated longitudinal changes in risk over time using an AI tool that estimates a woman’s future risk of breast cancer by looking only at her mammograms over several years. Researchers developed and tested this tool using sequential screening images from more than 54,000 women. The results showed that AI risk scores rose steadily starting five to six years before diagnosis in the 817 women who developed breast cancer, and stayed flat in women who remained cancer-free.

Group5-6 years before diagnosisFinal screening before diagnosis
Women who developed breast cancer2.16.6
Women who stayed cancer-free1.82.2

The study concluded that AI-based risk scores from screening mammograms evolve over time in a meaningful way, diverging between women who did and did not develop breast cancer. This work supports the potential of AI-based risk scores as dynamic biomarkers (biological indicators) for risk-adaptive screening and prevention strategies.

Beyond the mammogram: combining images, genetics, and family history for risk prediction

While research demonstrates the power of leveraging AI for risk prediction, AI imaging techniques alone do not paint a comprehensive picture of an individual’s risk. Researchers are working to combine information from a mammogram with a person’s family history of breast cancer and genetic risk estimates to round out a more robust, individualized representation of breast cancer risk. BCRF is committed to supporting research in each area; perhaps AI could be the tool that brings them together.

Several BCRF-funded projects are in the arena of genetic and risk-based research:

  • Dr. Fergus Couch is conducting the CARRIERS (Cancer Risk Estimates Related to Susceptibility) study, a large multi-institution study to investigate gene mutations beyond BRCA1 and BRCA2 that predispose individuals to breast cancer. His work can inform and improve risk prediction.
  • Dr. Mark Robson developed a multidisciplinary approach in his PPI project aimed to improve risk prediction for particularly aggressive breast cancers, such as triple-negative breast cancer (TNBC). He is developing polygenic risk scores—an estimate of a person’s inherited risk of breast cancer based on the combined effects of many common genetic variations, rather than looking at a single high-risk gene like BRCA1 or BRCA2—to give women more precise risk estimates. His work has laid the groundwork for developing personalized risk models.
  • Dr. James Ford is developing blood-based early detection tools that use polygenic risk scores in combination with multigene panels to estimate breast cancer risk.
  • Dr. Gad Rennert is conducting studies in more than 30,000 participants to find causes of breast cancer in younger women beyond known inherited mutations. To do this, his team is digitizing breast cancer pathology slides to identify patterns that distinguish cancers in younger women from those in older women. The team is also incorporating data from roughly 10,000 women to build a risk score based on the genetic and behavioral patterns specific to each population studied. Their goal is to improve prevention and early detection efforts for younger women at risk.
  • Dr. Graham Colditz is constructing a unique risk score, HERS (H&E Risk Score), that can predict the likelihood of benign lesions or ductal carcinoma in situ (DCIS) progressing to invasive breast cancer. This work involves integrating classical pathology, deep-learning whole-slide-imaging (using AI to analyze large scans), and molecular markers of benign breast disease (any noncancerous changes) to improve predictions of subsequent breast cancer in younger and racially diverse women.
  • Dr. Laura Esserman is conducting the WISDOM trial which is recruiting 100,000 women aged 30-74 to compare annual mammograms from age 40 with screening based on each woman’s risk factors, including breast density, genes and family history.

What comes next?

Dr. Lehman’s work on changes in risk estimates over time is expanding to ask a different question: Can a mammogram show whether prevention interventions are working? Her team will determine if changes in risk scores predict short-term risk—and if they decrease after prevention interventions, such as weight loss or chemoprevention. The results of these studies may one day inform clinical decision-making and widen the power of AI in precision prevention.

On September 22, 2026, BCRF announced its newest endeavor, BCRF Ventures, a venture philanthropy program, to invest in early-stage companies that seek to move breast cancer research from the lab toward the clinic. BCRF Ventures was launched with $23 million in seed donations, including an $11 million anchor gift from Karen and Rob Hale. The goal is to raise and deploy $100 million over the next five years to quicken the process of bench to bedside research. BCRF Ventures’ first investment was in Clairity, Inc. to help make Clarity Breast more widely available across medical centers in the U.S, and will consider more funding opportunities beginning in 2027.

What we know, and what researchers are still testing

AI cannot prevent breast cancer on its own, nor is a risk score a diagnosis. While research has demonstrated the power of AI in breast cancer, there is still much to do. Researchers continue to test AI tools in prospective studies; determining what changing risk scores can tell us; making these tools more accessible and equitable; and more.

“We need to have a screening regime which is personalized to your individual risk rather than just [based on] your age and your genetic status,” Dr. Barzilay said in an interview with BCRF. “So I think it’s key for these tools to be really equitable because this is the only way they can make a difference at the population level.”

BCRF is building the future of precision prevention

By funding the imaging, genetics, and prevention research mentioned above, studies that may one day boost AI’s ability to pull it all together, investing in companies advancing the science through BCRF Ventures, the Foundation is focused on personalizing prevention and moving discoveries toward patients as quickly as possible.

Researchers are working toward prevention tactics matched to every individual’s risk, and donor support moves that work forward. Help fund the research that could make prevention personal for every woman. Donate to BCRF here.

Frequently asked questions

Can AI predict breast cancer before it develops?

AI can estimate breast cancer risk over a set period; it cannot determine for certain who will develop cancer (Lehman, 2026).

What is precision prevention?

Precision prevention refers to the matching of prevention strategies to each woman’s individual risk and biology.

Can genetic testing predict breast cancer risk?

Genetic testing can indicate inherited risk for some women, but it doesn’t determine who and who will not develop breast cancer. BCRF-funded genetics research is deepening our understanding of susceptibility genes, which can increase a person’s risk. A conversation with your doctor regarding genetic testing is particularly important if you have a family history of breast cancer.

What AI research does BCRF fund?

BCRF is funding ongoing AI research of Drs. Barzilay and Adam Yala. The Foundation also invests in AI-driven digital pathology research and projects that aim to integrate AI into clinical care and planning. Investing in new innovations through BCRF Ventures is another way BCRF supports AI research in breast cancer.

Selected references icon-downward-arrow

Lehman, C. D., Mercaldo, S. F., Azam, S., Tamimi, R. M., Carney, A., Friedewald, S. M., & Lamb, L. R. (2026). Longitudinal Analysis of Changes in Deep Learning Image-based Breast Cancer Risk Scores over Time. Radiology, 319(3), e253023. https://doi.org/10.1148/radiol.253023

Knowledge is power.

Stay empowered with the latest news, insights, and resources delivered to your inbox.

legacy-Society

Support research with a legacy gift. Sample, non-binding bequest language:

I give to the Breast Cancer Research Foundation, located in New York, NY, federal tax identification number 13-3727250, ________% of my total estate (or $_____).

Learn More
Donate