Multiparametric radiomics-integrated models for predicting glioma molecular genotypes using 18F-FET PET/MRI

Motivation: There are few studies on the prediction of multiple genotype status in glioma by method of multiparameter 18F-FET PET/MR radiomics. Goal(s): To explore the value of radiomics features based on multiparametric O-(2-18F-fluoroethyl)-L-tyrosine (18F-FET) PET/MRI in predicting the molecular biomarker status of glioma. Approach: Retrospective radiomics analysis of 97 glioma patients' multiparametric PET/MRI data, employing Bayesian modeling and Receiver Operating Characteristic (ROC) curves to assess predictive performance for IDH, TERTp, and MGMT status. Results: Multi-modal radiomics models demonstrated superior performance in the testing cohorts (IDH AUC=0.97, TERTp AUC=0.86, MGMT AUC=0.90) compared to single-modal models across all types of molecular biomarkers. Impact: The radiomics models, which incorporate structural, proliferative, and metabolic information based on multiparameter 18F-FET PET/MRI, demonstrated effective predictive performance for the molecular biomarker status of glioma and had the potential to optimize diagnostic processes.