AI test predicts if breast cancer will return

In recent years, notable advances have been made in diagnosing and treating breast cancer. However, its recurrence continues to plague thousands—deepening the need to find ways to better predict the likelihood of return.

In a paper in the journal Nature Communications, a team of researchers reports the creation of an AI test that predicts the risk of breast cancer recurrence—one that does so more quickly and more inexpensively than current methods, which involve genomic testing. 

“Breast cancer is not a single disease, and decisions about how aggressively to treat it are often difficult,” explains Krzysztof J. Geras, a visiting scholar at NYU’s Center for Data Science and an adjunct assistant professor at NYU Grossman School of Medicine, who led the work. “This research shows that an AI test can read the same tumor slides pathologists already examine and, combined with basic clinical details, accurately estimate how likely a patient’s cancer is to return.”

“The model’s accuracy doesn’t come from hand-labeled data alone,” adds Yann LeCun, Jacob T. Schwartz Chaired Professor of Computer Science and Data Science at New York University and one of the paper’s authors. “It comes from self-supervised pretraining that lets it learn rich representations first, which then translate into strong downstream performance—a recipe that should generalize far beyond breast cancer and, more broadly, is the kind of new AI science these hard problems demand.”

Genomic tests used today assess how likely a patient’s hormone-receptor-positive breast cancer is to recur and whether that patient is likely to benefit from chemotherapy. However, this costly method can take weeks to generate results. Moreover, this testing requires examining, and then discarding, the tissue samples extracted as part of a lumpectomy or mastectomy—thereby preventing them from being used for future testing.

In seeking an alternative predictive tool, the authors developed and evaluated a multi-modal AI test by drawing from 15 patient populations across seven countries. 

They built the AI test by considering pathology slides—microscopic tissue samples on glass used to spot diseases—combined with routine clinical data such as tumor stage, patient age, and hormone-receptor status. The researchers then evaluated the test’s efficacy using data from more than 3,500 patients. They used standard statistical methods to gauge its accuracy: the C-Index, which assesses how well a predictive model discriminates between patients, and a Hazard Ratio, which compares the risk of an event (in this case, breast cancer) occurring in one group compared to another over time.

Overall, the AI test reliably separated higher- from lower-risk patients. It also performed well in evaluating the probability of recurrence in two types of breast cancer—triple-negative and HER2-positive—that currently have no reliable genomic test.

The researchers emphasize the need for evaluation in completed randomized clinical trials to build confidence in using the AI test to assess future breast-cancer risk and to guide treatment. However, they see the work as a meaningful progress toward using AI to help combat an affliction that plagues millions. 

“In testing on thousands of patients, our AI test matched or outperformed a widely used genomic test,” says Geras, who is also co-founder and chief scientific officer of Ataraxis AI, a company that uses AI to develop cancer treatments and diagnoses. “Because it relies on existing slides, it could deliver answers in hours instead of weeks, at lower cost, while sparing tissue for future testing.”

The paper’s other authors were Professor Carlos Fernandez-Granda of NYU’s Courant Institute School of Mathematics, Computing, and Data Science and Center for Data Science, along with Kangning Liu, a doctoral student at the Center for Data Science at the time of the study, as well as the following NYU Grossman School of Medicine researchers: Nancy Chan, Freya Schnabel, Ugur Ozerdem, Nitya Thakore, Mohammad Sadic, Frank Yeung, Elisa Liu, Theodore Hill, Benjamin Swett, Danielle Rigau, Andrew J. Clayburn, Linda M. Pak, and Natalie Klar.

# # #

Editor’s Note: 

Some of the paper’s authors are equity holders of Ataraxis AI, and NYU maintains financial and intellectual property interests in Ataraxis AI.

In November 2025, NYU announced the establishment of the Courant Institute School of Mathematics, Computing, and Data Science. The newly established school recognizes the storied history of the Courant Institute of Mathematical Sciences—and its strengths in both applied and pure mathematics—while encompassing NYU’s Center for Data Science and linking the computer science departments at Courant and the Tandon School of Engineering.