NEW YORK, July 12 — A research team at New York University has developed an artificial intelligence tool that matches or outperforms genetic tests in predicting breast cancer recurrence. The work, led by NYU faculty member Krzysztof J. Geras, combines digital images of tumour tissue with basic clinical information to guide treatment decisions.
Geras, whose background sits at the intersection of machine learning and medical imaging, framed the research around a clinical reality that oncologists confront daily. “Breast cancer is not a single disease, and decisions about how aggressively to treat it are often difficult,” he said.
The statement captures the motivation behind the project: to give clinicians a data-driven second opinion at a moment when the stakes are high and the path forward is rarely obvious.
Under the hood
The technical reality of the tool is worth unpacking. Rather than relying on genomic sequencing — the current gold standard for assessing recurrence risk — the system analyses digitised slides of tumour tissue. These are the same pathology slides that hospitals already produce as part of standard care.
By layering in basic clinical data, the model arrives at a risk assessment that the team says matches or beats the genetic tests now on the market. That matters because genetic tests are expensive, not universally available, and require tissue samples to be sent to specialised laboratories.
A tool that works from routine pathology images could, in principle, be deployed far more widely. The detail that matters here is that the system does not ask for anything new from the patient or the hospital; it simply extracts more value from data already being collected.
Part of a broader push
Geras’s work is one strand of a wider effort to apply artificial intelligence to oncology. The field has been moving steadily from proof-of-concept studies toward tools that can actually influence care. This particular project targets a specific inflection point: the moment after surgery when a patient and their doctor must decide whether chemotherapy or hormonal therapy is warranted, and how aggressive that treatment should be.
As ever with AI in medicine, the question is not just whether the model works in a lab but whether it changes outcomes in practice. The NYU team has not yet published data on prospective clinical use, but the performance against genetic tests — which are themselves well-validated — suggests the approach has legs.
What to watch next: whether the tool moves into clinical trials, and whether other institutions can replicate the results on their own patient populations. If they can, the economics of recurrence prediction may shift. A technology that runs on existing pathology images and a spreadsheet of clinical variables is a very different proposition from one that requires a genomic assay costing thousands of dollars.
{“@context”: “https://schema.org”, “@type”: “NewsArticle”, “headline”: “NYU’s Krzysztof Geras Leads Major AI Breast Cancer Study”, “description”: “NEW YORK, July 12 — A research team at New York University has developed an artificial intelligence tool that matches or outperforms genetic tests in…”, “datePublished”: “2026-07-12T09:30:00+00:00”, “author”: {“@type”: “Person”, “name”: “Sebastian Smith”}, “publisher”: {“@type”: “Organization”, “name”: “InfoPulse Today”, “url”: “https://newsfeed.infopulsetoday.com/”}}
