Researchers at Karolinska Institutet have examined the ability of various AI models to predict the prognosis of triple-negative breast cancer by analyzing specific immune cells within the tumor. Published in eClinicalMedicine, the study marks a significant step toward integrating AI into cancer care to enhance patient outcomes.
Image Credit: Guschenkova/Shutterstock.com
Tumor-infiltrating lymphocytes are immune cells crucial in combating cancer. Their presence in a tumor indicates an immune response aimed at attacking and eliminating cancer cells.
These immune cells can provide valuable insights into how a patient with triple-negative breast cancer will respond to treatment and how the disease is likely to progress. However, pathologists' assessments of these cells can be inconsistent. AI offers a way to standardize and automate this evaluation, though proving its reliability for healthcare use has been challenging.
Compared Ten AI Models
The researchers evaluated ten AI models to compare their performance in analyzing tumor-infiltrating lymphocytes from triple-negative breast cancer tissue samples.
The results revealed variations in the analytical performance of the AI models. However, eight out of ten demonstrated strong prognostic capabilities, predicting patients' future health outcomes with similar accuracy.
Even models trained on fewer samples showed good prognostic ability, suggesting that tumor-infiltrating lymphocytes are a robust biomarker.
Balazs Acs, Researcher, Department of Oncology-Pathology, Karolinska Institutet
Independent Studies Needed
The study highlights the necessity of large datasets to compare various AI tools and verify their effectiveness before implementation in healthcare. Although the findings are encouraging, further validation is required.
Our research highlights the importance of independent studies that mimic real clinical practice. Only through such testing can we ensure that AI tools are reliable and effective for clinical use.
Balazs Acs, Researcher, Department of Oncology-Pathology, Karolinska Institutet
Journal Reference:
Vidal, J. M., et al. (2024) The analytical and clinical validity of AI algorithms to score TILs in TNBC: can we use different machine learning models interchangeably? EClinicalMedicine. doi.org/10.1016/j.eclinm.2024.102928.