Gazi University

Machine Learning Model Achieves 96% Accuracy in Predicting Breast Cancer Treatment-Related Cardiac Events

- A novel machine learning model combining clinical, radiomic, and dosiomic parameters achieved 96% accuracy in predicting treatment-related cardiac events in breast cancer patients, significantly outperforming traditional clinical and dosimetric approaches (67% accuracy). - The study analyzed 42 breast cancer patients using high-sensitivity troponin T levels above 14 ng/L as the threshold for cardiac events, with the combined model demonstrating superior predictive performance through gradient-boosted classification algorithms. - This represents the first study to utilize heart-segmented dosiomic data in adult breast cancer patients, potentially enabling early detection of subclinical cardiotoxicity before irreversible cardiac damage occurs. - The findings suggest that imaging biomarkers extracted from CT scans and radiation dose distributions could revolutionize cardiovascular risk assessment in cancer treatment, though validation in larger patient cohorts is needed.