NIMG-73. RADIOMICS OF GLIOBLASTOMA FOR PREDICTING MGMT PROMOTOR METHYLATION STATUS AND PROGNOSIS

Authors:

Abstract
O-6-methylguanine-DNA methyltransferase promotor methylation (pMGMT-met) is identified as a favorable prognostic factor for glioblastoma (GBM) and patients with GBM containing pMGMT-met benefit from temozolomide. On the other hand, radiomic analysis that rely on quantitative texture feature of radiological images has rapidly emerged in the field of neuro-oncology. In this study, the authors attempted to build a radiomic-based prediction model for pMGMT-met status of GB. Building a prognostic model was further attempted by radiomic analysis of conventional MRI. Preoperative MRI (non-enhanced and enhanced T1WI, T2WI, and FLAIR) from 207 newly diagnosed GBM (nGMB) patients were included in this study. Total of 489 texture features including first order feature (histogram of pixels), second order feature (Gray level co-occurrence matrix and Grey level run length matrix) from 166 date sets and location data from 187 data set were collected. Predictive modeling for pMGMT-met status was performed based on LASSO regression with 10-fold cross-validation which was repeated 5 times. Supervised component principal analysis was used for prognosis prediction. One hundred and four cases were pMGMT-met and 103 patients were MGMT promotor unmethylated (pMGMT-unmet). Predictive accuracy of pMGMT-met status was 68% modeled by 28 significant radiomic features. Furthermore, 20 radiomic features correlated with prognosis the analyzed cohort was categorized into high risk and low risk group by use of radiomic score (P value <0.0069). Radiomic score and pMGMT status were independent prognostic factors. Finally, significant survival difference was observed between radiomic high risk and pMGMT-unmet group, intermediated group (which group consists of radiomic low risk and pMGMT-unmet or radiomic high risk and pMGMT-met), and radiomic low risk and pMGMT-met group (P value <0.0001). In conclusion, the current study revealed that biological characteristics of GBM is embedded in radiomic features, recovery of which information could be beneficial for GBM patient care.