Differentiation of glioblastoma from solitary brain metastases using radiomic machine-learning classifiers

This study aimed to identify the optimal radiomic machine-learning classifier for differentiating glioblastoma (GBM) from solitary brain metastases (MET) preoperatively. Four hundred and twelve patients with solitary brain tumors (242 GBM and 170 solitary brain MET) were divided into training (n = 227) and test (n = 185) cohorts. Radiomic features extraction was performed with PyRadiomics software. In the training cohort, twelve feature selection methods and seven classification methods were evaluated to construct favorable radiomic machine-learning classifiers. The performance of the classifiers was evaluated using the mean area under the curve (AUC) and relative standard deviation in percentile (RSD). In the training cohort, thirteen classifiers had favorable predictive performances (AUC≥0.95 and RSD ≤6). In the test cohort, receiver operating characteristic (ROC) curve analysis revealed that support vector machines (SVM) + least absolute shrinkage and selection operator (LASSO) (AUC, 0.90) classifiers had the highest prediction efficacy. Furthermore, the clinical performance of the best classifier was superior to neuroradiologists in accuracy, sensitivity, and specificity. In conclusion, employing radiomic machine-learning technology could help neuroradiologist in differentiating GBM from solitary brain MET preoperatively.