NIMG-47. A HISTOGRAM-BASED, BACK-PROJECTION METHOD FOR TREATMENT RESPONSE ASSESSMENT IN GLIOBLASTOMA USING MULTI B-VALUE ADVANCED DIFFUSION MRI

Abstract

INTRODUCTION: Early and accurate assessment of therapeutic response in glioblastoma is important for clinical patient management, and as platform for clinical trials of novel therapies. Current response criteria such as RANO (Response Assessment in Neuro-Oncology) rely on semiquantitative measurements with limited sensitivity and specificity, especially early during treatment. Functional diffusion maps based on voxel-to-voxel comparison of quantitative diffusion measures are confounded by change in tumour size. Towards improving response assessment, we propose a histogram- based, voxel back projection method using advanced quantitative diffusion MRI. METHOD(S): We used least-square fitting to model four dif- fusion parameters from mono-, bi-, and stretch-exponential models, and performed a histogram analysis of all voxels located within ROIs dictated by the radiotherapy-planned clinical target volume. Histograms were generated for 10 patients with GBM at 2 time-points, before, and at 6 weeks of treatment with standard-of-care regimen (RT with concomitant and adjuvant temozolomide). For each parameter, percentile ranges corresponding to diffusion values with the greatest cumulative difference between the time points were identified, and associated voxels were back-projected onto their respective maps for spatial correspondence and comparison with standard radiological imaging. RESULTS & DISCUSSION(S): The greatest cumulative difference in diffusion parameters measured at baseline and at 6 weeks of treatment, spatially corresponded to voxels lying within regions of high FLAIR signal intensity and regions radiologically-defined as healthy brain. The magnitudes of diffusion imaging parameters within these regions were heterogeneously distributed and did not show a direct spatial correspondence to radiological signal intensities. This heterogeneous spatial distribution may be influenced by microcellular changes, for example, in areas of infiltrative tumor which are not visible on standard imaging. Further assessment of this dataset, augmented by on-going patient recruitment across multiple centres, will provide insight to the prognostic value of advanced diffusion MRI as a method for response assessment in glioblastoma.