Functional MR habitat imaging feature for preoperative prediction of TERT promoter and EGFR gene phenotype in IDH-wildtype glioblastoma

Motivation: The 2021 WHO classification has identified the important role of molecular phenotypes in the outcomes of newly diagnosed glioblastoma. The habitat imaging plays a vital role in predicting molecular status of glioblastoma. Goal(s): To explore the value of the MR habitat imaging model in predicting the TERT and EGFR phenotype in glioblastoma. Approach: Glioblastoma was split automatically into several habitat subregions. TERT and EGFR phenotype prediction models were constructed using the RF. Results: The AUC of the training, test and validation sets of the TERT and EGFR prediction model were 0.877, 0.783, 0.796, and 0.877, 0.784, 0.878, respectively. Impact: The habitat imaging model based on functional MR were helpful for accurately predicting TERT and EGFR molecular status of tumors.