Glioblastoma: Pre-treatment geometry and texture of postcontrast T1 MRI matter

Abstract

The potential of tumor's volumetric and texture measures obtained from pretreatment magnetic resonance imaging (MRI) sequences of glioblastoma (GBM) patients as predictors of clinical outcome has been controversial. Mathematical models of GBM growth have suggested a relation between tumor's geometry and its aggressiveness. Multicenter retrospective clinical studies were designed to study geometrical and 3D textural measures on pretreatment postcontrast T1 MRIs of 117 and 79 GBM patients respectively. Clinical variables were collected, tumors segmented and measures computed. Kaplan-Meier and univariate Cox survival analysis showed that spherical contrast enhancing width (p=0.007, HR=1.749) and geometric heterogeneity of the contrast enhancing rim (p=0.015, HR=1.646) were the outstanding parameters in terms of overall survival. Patients with tumors having small geometric heterogeneity and/or spherical rim widths had significantly better prognosis. Regarding 3D textural measures, the co-occurrence matrix feature Entropy showed a gain in median survival for the favorable subgroup of 8.22 months respectively (p=0.013). Patients with tumors having small entropy values had significantly worst prognosis. These geometrical and texture imaging biomarkers have a strong individual and combined prognostic value for GBM patients. “Gliomator” is suggested as a GBM prognostic classifier based on the combined use of geometrical and texture information.