AI as good as humans at predicting breast cancer outcomes: Studies

Artificial intelligence (AI) models are as good as human pathologists at a task that helps predict outcomes for patients with breast cancer, two Australian-led studies have found.

The studies, published in The Lancet Oncology, conclude that AI models that count tumor-infiltrating lymphocytes (TILs) in samples of breast tissue should be more widely used "particularly where routine or widespread pathologist assessment is unavailable," according to a recent statement by Australia's Peter MacCallum Cancer Center (Peter Mac).

TILs are immune cells that can be counted, usually by a pathologist examining a microscope slide.

Higher levels indicate a stronger immune response against the tumor and have previously been associated with better outcomes in several types of breast cancer, said researchers who analyzed data from more than 5,600 breast cancer patients.

The first study found that while AI and pathologist scores were not identical, both provided similar information about the patients' prognosis.

The second study showed that AI could also analyze how immune cells were organized and highlight "hotspots," which provided additional prognostic information beyond simply counting TILs.

The researchers said the studies strengthen the case for TILs to be used as a practical biomarker in breast cancer, as well as showing that AI could enable their assessment at scale.

"AI may also allow us to extract information that the human eye cannot readily quantify, such as the spatial organization of immune cells within a tumor.

Ultimately, combining pathologists and computational approaches may give us more information than either approach alone," said Peter MacCallum's Professor Sherene Loi, who led the two new studies.