Artificial Intelligence Digitalizes the Measurement of Immune Cells in Breast Cancer
Research shows that AI-powered measurements yield meaningful results in predicting patient survival rates and can play a complementary role to expert pathologist evaluations.
The independent CATALINA study, published in The Lancet Oncology, revealed that AI-based digital tools can reliably measure tumor-infiltrating lymphocytes (TILs) in early-stage triple-negative breast cancer and predict patient outcomes.
Traditional Methods and the Need for Artificial Intelligence
Tumor-infiltrating lymphocytes (TILs) are considered one of the most important biomarkers showing the interaction between the immune system and tumor biology, particularly in triple-negative breast cancer. For years, the evaluation of these cells has relied on expert pathologists manually examining tumor sections stained with hematoxylin and eosin.
Although standardized approaches improve accuracy, manual scoring requires expertise, is extremely time-consuming, and makes consistent implementation across different healthcare systems difficult. This situation increases the need for digital and automated solutions.
Details of the CATALINA Study
The independent CATALINA study, published in The Lancet Oncology, evaluated the success of AI algorithms in computer-based TIL (cTIL) measurement. The study examined whether AI could serve as a reliable alternative to traditional methods by utilizing large datasets obtained from randomized clinical trials.
Within the scope of the study, data from 1,759 patients with early-stage triple-negative or HER2-positive breast cancer were analyzed. Of these patients, 1,356 had complete clinicopathological information, stromal TIL values scored by pathologists, and cTIL scores calculated by artificial intelligence.
Clinical Results and the Future of Artificial Intelligence
The findings demonstrated that AI-derived TIL measurements provide statistically significant data in predicting clinical outcomes such as disease-free survival and overall survival rates of patients. Both traditional pathologist assessments and AI scores were found to be independently associated with improved patient outcomes.
However, when AI scores were combined with traditional clinicopathological factors and pathologist evaluations, the additional contribution offered by AI was not statistically significant. This indicates that AI currently serves as a complementary tool rather than replacing expert pathologists.