Technion AI reads routine biopsy slides to predict who benefits from breast cancer chemo

Researchers at the Technion-Israel Institute of Technology announced on 25 March 2026 that they had developed an artificial intelligence model able to predict both the risk that a woman's breast cancer will return and how much she is likely to gain from chemotherapy after surgery, using only the routine tissue slides prepared when the tumour is diagnosed. The work, published in The Lancet Oncology, was led by Dr. Gil Shamai of the Geometric Image Processing Laboratory headed by Professor Ron Kimmel in the Taub Faculty of Computer Science, together with Professor Dvir Aran of the Faculty of Biology.

The deep-learning system examines many regions of a digitised tumour slide and the tissue around it, looking for visual patterns linked to how a cancer behaves, such as the rate of cell division, tissue architecture, immune activity and signs of sensitivity or resistance to treatment. It condenses these into a single score. Because it needs no extra tissue or specialised laboratory tests, the team says it can run within minutes in any pathology lab with a slide scanner and an internet connection.

Crucially, the researchers had access to tissue and clinical data from TAILORx, a landmark trial of more than 10,000 patients who were randomly assigned to receive chemotherapy or not. According to the Technion, this makes it the first AI model of its kind validated in a large randomised trial. It was further tested on thousands of patients from hospitals in Israel, including Carmel, Haemek and Sheba, as well as in the United States and Australia, with consistent results across populations and equipment. Oncologists and pathologists from Dana-Farber Cancer Institute, Mount Sinai, the University of Chicago and Portugal's IPATIMUP took part.

The team is now working toward clinical use in Israel, preparing clinical trials in Brazil and India, and planning to extend the approach to other cancers and treatments. The researchers also intend to found a company to develop faster, more accessible tests.