Two studies conducted by Australian scientists have concluded that artificial intelligence (AI) models demonstrate effectiveness comparable to human pathologists in predicting outcomes for patients with breast cancer.
According to a recent statement from the Peter McCallum Cancer Centre in Australia, AI models that count tumor-infiltrating lymphocytes (TILs) in breast tissue samples should be applied more widely, especially in cases where manual or mass assessment by pathologists is unavailable.
TILs are immune cells that are typically counted by a pathologist when examining a microscopic slide. Higher concentrations of these cells indicate a stronger immune response against the tumor and have previously been associated with better outcomes in various types of breast cancer. The researchers analyzed data from over 5600 patients with this disease.
The first study showed that although AI and pathologist assessments did not fully align, they provided similar information regarding patient prognosis. The second study demonstrated that AI is also capable of analyzing the structure of immune cells and identifying so-called 'hotspots,' which provides additional prognostic information beyond simple TIL counting.
The scientists emphasized that these works strengthen the arguments for using TILs as a practical biomarker in breast cancer, while also confirming the possibility of large-scale assessment using AI. Professor Sheren Loi from Peter McCallum, who led both studies, noted that AI can allow for the extraction of information that the human eye cannot easily quantify, such as the spatial organization of immune cells within the tumor. She added that combining the work of pathologists and computational methods can provide more data than either approach alone.