MRI Radiomics-Based Molecular Subgrouping and Survival Prediction in Glioma Patients

Background: Gliomas can be classified into five molecular subgroups based on the status of IDH mutation, 1p/19q codeletion, and TERT promoter mutation, whereas they need to be obtained by biopsy or surgery. We aimed to use MRI-based radiomic approach to noninvasively predict the five molecular subgroups and assess its prognostic value in glioma. Methods: We retrospectively identified 357 patients with grade II-IV gliomas. We extracted radiomic features of tumor and/or edema from preoperative MRI. Features were selected using the least absolute shrinkage and selection operator (LASSO). The predictive model was developed by incorporating preoperative clinical data into the multiparametric MR radiomic signature constructed by machine learning. We compared the prognostic value of predictive molecular subgroups with actual molecular subgroups in predicting progression-free survival (PFS) and overall survival (OS). We developed prognostic nomograms to predict PFS and OS of patients. Findings: Age, tumor location, and LASSO-based multiparametric MR radiomic model yielded the highest performance in predicting IDH mutation status (AUC = 0.855). Decision tree-based multiparametric MR radiomic model performed the best in predicting 1p/19q codeletion (AUC = 0.796) and TERT promoter mutation (AUC = 0.849). The predictive molecular subgroups were comparable to actual molecular subgroups in predicting PFS (C-index: 0.710 vs 0.717, P = 0.890) and OS (C-index: 0.707 vs 0.720, P = 0.630). The prognostic nomograms yielded a C-index of 0.742 and 0.747 in predicting PFS and OS, respectively. Interpretation: Multiparametric MRI-based radiomics with machine learning can be useful for noninvasively detecting molecular subgroups and predicting survival in gliomas. Funding Statement: This work was supported by the National Natural Science Foundation of China (No: 81571664, 81871323, 81801665, 81702465 and U1804172); the National Natural Science Foundation of Guangdong Province (No: 2018B030311024); the Scientific Research General Project of Guangzhou Science Technology and Innovation Commission (No: 201707010328); the China Postdoctoral Science Foundation (No: 2016M600145), the Science and Technology Program of Henan Province (No: 182102310113, 192102310050 and 192102310123), and the Youth Innovation Fund of The First Affiliated Hospital of Zhengzhou University to Zhen-yu Zhang, the Key Research Projects of Henan Higher Education (No: 18A310033). Declaration of Interests: The authors declared no competing interest exists. Ethics Approval Statement: The institutional review board in all participating centers approved this retrospective study and waived the need to obtain written consent.