AI matches pathologists in breast cancer trials

A breast-tissue slide can reveal an immune response as well as a tumor. Usually, a pathologist counts the tumor-infiltrating lymphocytes — TILs, immune cells inside and around the tumor — through a microscope. In two Peter Mac-led studies, AI models provided prognostic information similar to pathologists' assessments across more than 5,600 patients.

The first analysis, called CATALINA, combined seven clinical trials involving triple-negative breast cancer, a form that accounts for about 15% of breast cancers in Australia. More than 1,300 patients had their TILs assessed by both a pathologist and an AI model. The scores did not match exactly, but the signal did: patients with more TILs had better outcomes whichever method was used.

The second study examined more than 4,300 tumor samples from the Phase III APHINITY trial in early HER2-positive breast cancer. Manual pathologist scoring, digital image analysis and AI-based methods all linked higher TIL levels with better outcomes. The AI analysis went beyond counting, identifying the spatial organization of immune cells and hotspots that supplied additional prognostic information.

That matters because TILs could become a practical biomarker — a measurable feature used to help assess disease and treatment — across different breast cancer types. In APHINITY, patients with the highest TIL levels, particularly those with node-positive breast cancer, appeared to gain greater benefit from adding pertuzumab to standard care. The studies were published simultaneously in The Lancet Oncology.

So what changes in practice? AI could help hospitals assess this immune signal more objectively and at larger scale, including places where routine expert pathology is unavailable. It is not presented as a replacement for pathologists: the studies show comparable prognostic information in trial analyses, while the models' use in routine care has not been established here. Australia records more than 20,000 cases a year, making breast cancer its most common cancer in women.