The optimal support vector machines model, integrating intratumoral and peritumoral radiomics features, showed promise in predicting brain metastasis in patients with newly diagnosed lung cancer, according to a study recently published in Clinical Radiology. Researchers investigated the potential of using radiomics analysis and machine-learning models to predict brain metastasis among 183 newly diagnosed lung cancer patients. With a training cohort of 146 and a validation cohort of 37 patients, the study extracted radiomics features from plain computed tomography images of primary lesions. The study team developed predictive models and assessed diagnostic performance using four machinelearning algorithms. A total of 22 highly predictive radiomics features, including nine from the peritumor region, following a feature screening. All four machine-learning models exhibited predictive capability, with SVM demonstrating superior efficiency and robustness. The SVM model achieved an area under the ROC curve of 0.918 in the training cohort and 0.901 in the validation cohort. Decision curve analysis indicated the highest net benefit. Moreover, the time-dependent ROC curve dem-onstrated efficacy in predicting metachronous brain metastasis occurrence over 1-, 2-, and 3-year follow-up periods, with all AUC values surpassing 0.7. According to the researchers, the study’s results may influence clinical diagnosis and treatment strategies
