Israeli researchers develop AI tool to predict chemotherapy benefit
Technion model could replace costly genomic tests with faster pathology analysis.
Researchers at Technion - Israel Institute of Technology have developed an artificial intelligence model that could significantly change how doctors in Israel, and potentially worldwide, decide whether breast cancer patients should receive chemotherapy.
The system, trained to analyze routine pathology samples, predicts both the likelihood of cancer recurrence and whether a patient is likely to benefit from chemotherapy. The research, published in The Lancet Oncology and presented at a major European oncology conference, marks the first time such a model has been validated using data from a large randomized clinical trial.
For Israel, where around 5,000 people are diagnosed with breast cancer each year, the implications are immediate: a faster, more accessible decision-making tool that could reduce unnecessary treatment while ensuring those who need chemotherapy receive it.
Determining whether to administer chemotherapy after surgery remains one of the most difficult decisions in early-stage breast cancer care. While the treatment can reduce the risk of recurrence, many patients derive little or no benefit, yet still face significant side effects.
Today, this decision often relies on genomic tests such as Oncotype DX. These tests are costly, can take weeks to deliver results, and are not universally accessible. Even where available, their predictive accuracy is limited.
The Technion model seeks to bypass these constraints by using data already collected during standard diagnosis.
Instead of analyzing genes, the system examines high-resolution digital images of tumor tissue. Using deep learning, it identifies patterns in both the cancer cells and their surrounding environment, patterns linked to how aggressive the tumor is and how it might respond to treatment.
The study was led by Dr. Gil Shamai, Prof. Ron Kimmel, and Prof. Dvir Aran, in collaboration with oncologists and pathologists from institutions including Dana-Farber Cancer Institute, Mount Sinai Medical Center, the University of Chicago Medical Center, and IPATIMUP Medical Center in Portugal.
“These are biological signals that the human eye cannot reliably quantify,” said Gil Shamai. “The model integrates many subtle cues into a score that reflects both recurrence risk and expected benefit from chemotherapy.”
According to Ron Kimmel, the approach represents a conceptual shift. “Instead of testing genes, we look directly at the tissue,” he said, comparing the method to identifying eye color by observation rather than DNA analysis.
Researchers were granted access to data from the TAILORx trial, one of the largest breast cancer studies conducted, involving more than 10,000 patients.
“Using randomized trial data allowed us to test whether the model predicts actual benefit from chemotherapy, not just risk,” Shamai said.
The system was also tested on thousands of additional patients across multiple countries, including hospitals in Israel such as Carmel, Haemek and Sheba Medical Centers. The results showed consistent performance across different healthcare systems and imaging conditions.
“This is the first AI model shown to predict treatment benefit in breast cancer directly from pathology samples,” said Dvir Aran.
