Abstract 4299: Evaluation of a new multimodality voxel-based imaging biomarker for therapeutic response assessment in GBM

Abstract

Abstract The purpose of this study was to evaluate the diagnostic performance of MRI based T1 and relative cerebral blood volume (rCBV) maps analyzed a multi-modality voxel-based analysis referred to as multi-parametric response mapping (mPRM) for early chemoradiation response prediction in patients diagnosed with high-grade gliomas. Patients (n=23) with Grade III/IV glioma were recruited in a prospective imaging trial. Patients underwent MRI before RT, 1 and 3 weeks after RT. The MRI protocol included fluid-attenuated inversion recovery imaging (FLAIR), quantitative T1 mapping, dynamic contrast-susceptibility T2*-weighted imaging (rCBV) and contrast-enhanced T1-weighted imaging (CE-T1w). All images were co-registered to CE-T1w images acquired before RT. Following co-registration, the tumor VOIs were manually contoured either on the CE-T1w or on the FLAIR images and applied to the rCBV and T1maps. For each patient, mid-treatment time point and VOI the percentage change, PRM and mPRM techniques were used to analyze data. Briefly, PRM was performed by calculating the difference in the rCBV values of each voxel within the tumor at mid-treatment values with respective pre-treatment values. A threshold was then applied to the absolute difference of the rCBV in a voxel and all like voxels were summed to obtain tumor volume fractions that showed significantly increasing (PRMrCBV+), decreasing (PRMrCBV-), and unchanged (PRMrCBV0) rCBV values following therapy. We used the same procedure for determining the PRM of T1 maps. mPRMs maps were computed by combining both PRMs (T1 and rCBV) in a single color map. Receiver operator characteristic analysis (ROC), assessed for 12 month survival, was used to determine the optimal cutoff for each parameter. The patient population was then stratified based on the optimal cutoffs obtained from the ROC analysis. Overall survival for each parameter was determined using Kaplan-Meier curves and the log-rank test. Standard PRM, using T1or rCBV maps (specifically PRMT1+ and PRMrCBV- metrics) and percentage change methods were found to be predictive of survival only for specific time of acquisition and VOI used. It was determined that mPRMT1+/rCBV- significantly identified patients resistant to therapy irrespective of tumor volume delineation on CE-T1w or FLAIR images and the time point mid-treatment used. Our results show that mPRM improves the sensitivity of quantitative T1 and rCBV maps to predict overall patient survival. This study introduces a new approach for analyzing and combining multi-parametric MR images into a single multi-parametric response map. The mPRM metric showed promise as an early and robust imaging biomarker of treatment response in patients diagnosed with high grade gliomas. This novel approach is a very sensitive tool for analysis of multiparametric and/or multimodal data with enhanced sensitivity for early detection of therapy response. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 4299. doi:1538-7445.AM2012-4299