IMAGING HABITAT CHANGES AFTER PREOPERATIVE RADIOTHERAPY FOR GLIOBLASTOMA: UTILISATION OF A BIOINFORMATICS AND MACHINE LEARNING PIPELINE TO CHARACTERISE REGIONAL TREATMENT RESPONSE

Abstract AIMS Imaging of tumour habitats is a novel technique to assess regional treatment response by defining ‘habitats’, which are regions with shared imaging characteristics. Here, we describe pipeline characterise in glioblastoma using bioinformatics and machine learning. We apply this patients the Preoperative Brain Irradiation Glioblastoma (POBIG) trial that received preoperative radiotherapy for first time. METHOD underwent diffusion/perfusion MRI. A two- stage clustering was developed define relative cerebral blood volume (rCBV) apparent diffusion co- effcient (ADC). 6 were defined occupied 10.6-21.5% volume. Treatment evaluated Ktrans permeability. RESULTS In newly diagnosed cases commonest included those medium high rCBV. contrast, recurrent had significantly greater proportion low rCBV (p = 0.026). Two POBIG 8 Gy radiotherapy. Within 24 hours there global decrease vascularity (-3.7%) increase (+8.2%) irradiated regions. Habitats demonstrated heterogeneous radiotherapy, (mean +83%) versus +40%). CONCLUSIONS have unique delineate applied it trial. can allow monitoring could more targeted future.