Intra-tumoural perfusion habitats showed prognostic value in glioblastoma patients

Abstract

Abstract The perfusion within glioblastoma is associated with tumour microenvironment and may create invasive tumor habitats that could potentially be revealed by perfusion imaging. The purpose of this study is to characterize the peritumoural habitats of glioblastoma for treatment target. Dynamic susceptibility contrast-enhancement (DSC) MRI was acquired pre-operatively on 115 newly-diagnosed glioblastoma patients. All images were co-registered to post-contrast T1-weighted images. The relative cerebral blood volume (rCBV), mean transit time (MTT) and relative cerebral blood flow (rCBF) maps were generated from the DSC images. The contrast-enhanced and peritumoural tumor regions were semi-automatically segmented from the post-contrast T1-weighted and FLAIR images. To delineate the habitats of different perfusion levels, a two clusters mixture model with Gaussian distribution was fitted to the rCBV, rCBF, and MTT within both contrast-enhanced and peritumoural regions. Perfusion parameters of the identified habitats were compared, and the prognostic values of habitats were investigated using survival analysis. The results showed that although non-enhanced, the peritumoral high perfusion (PHP) habitat demonstrated similar perfusion level with the contrast high perfusion (CHP) habitat, with similar rCBV (PHP: 1.13 ± 0.18, 95% CI [1.10, 1.15]; CHP: 1.21 ± 0.25, 95% CI [1.16, 1.21]) and rCBF (PHP: 1.08 ± 0.23, 95% CI [1.05, 1.08]; CHP: 1.08 ± 0.19, 95% CI [1.05, 1.08]). Multivariate Cox regression showed that the volumes of both habitats were associated with worse patient overall survival (PHP: P = 0.032; HR= 7.09; CHP: P = 0.008; HR= 12.01). Our results suggest that the intra-tumoural perfusion habitats may potentially offer treatment targets.