NIMG-69. MRI T2 RELAXIVITY IN THE PREDICTION OF PROGRESSION-FREE SURVIVAL IN NEWLY-DIAGNOSED GLIOBLASTOMA: WHOLE BRAIN MEASUREMENTS AND DEEP LEARNING

Abstract Glioblastoma (GBM) is the most common and aggressive primary brain tumour; we aimed to evaluate T2 relaxivity in their follow. Fifty-one newly diagnosed GBM followed at our department after surgery combined chemo-radiotherapy were included (> 180 days progression-free survival [PFS] as determined by RANO, ≥ 3 MRI examinations PFS interval). All data used prior progression. relaxation rates (1/T2f) calculated voxel-wise assuming mono-exponential decay. Whole-brain (WB) values extracted including skewness, kurtosis, mean, median, 15 85 percentile histogram variance. We additionally segmented 1/T2f volumes using a 5 compartment Gaussian mixture model, producing variance percent per component. To predictive potential with respect PFS, deep long short-term memory (LSTM) model tensorflow (4 timesteps). Twenty subjects inclusion criteria differed (progression, ≤ 120 between all MRIs [mean 44.3 days], 4 MRIs). Two models tested: regression (days progression) classification (±18 months PFS; output layer). Both run 10 times; mean results are presented. found WB median correlated decreased linear slopes also progression vs. pseudo-progression. The LSTM achieved an R2 of 86%. macro precision 85%, recall 84% F1 accuracy classifying progression/no 18 month cutoff. very intriguing measurements distinguishing from pseudo-progression, suggesting despite otherwise unremarkable imaging. A more complex fully-automated segmentation method learning, was promising. Our may find utility monitoring patients.