NI-75 * QUANTITATIVE TEXTURE DESCRIPTORS ON BASELINE-MRI CAN PREDICT PATIENT SURVIVAL IN NEWLY DIAGNOSED GLIOBLASTOMA MULTIFORME PATIENTS

Abstract

OBJECTIVE: Glioblastoma Multiforme (GBM) is the deadliest form of brain tumors with a median survival of 12 months. While methylation of the MGMT gene promoter and IDH1 status have been shown to be associated with prolonged survival in GBM patients, there is controversy about their use as prognostic markers for evaluating response in newly diagnosed patients. The underlying hypothesis of this study is that computer-extracted texture descriptors can capture subtle, multiscale morphologic attributes on baseline-MRI and thus can better predict patient survival as compared to the known molecular markers, IDH1 and MGMT. METHODS: 16 3Tesla MRI studies with Gd-T1C, FLAIR, T2w protocols with known MGMT, IDH1 status were recruited as part of ongoing IRB-approved Ohio brain tumor study. 8 patients had overall survival (OS) > 500 days (prolonged survival) and 8 had OS 0.05) in separating survival outcome. CONCLUSIONS: Laws texture features appeared to capture differences in lesion heterogeneity between prolonged and poor patient population suggesting that there are subtle morphological prognostic cues on baseline-MRI that may better predict patient survival in newly diagnosed GBM patients, that the known IDH1, and MGMT molecular markers.