NIMG-68. TEMPORAL CHANGES IN TUMOR COMPOSITION DIFFER ACROSS TUMOR GRADE IN GLIOMA

Abstract

Abstract Gliomas are the most common primary adult brain tumor in the United States and their biological heterogeneity makes it difficult to assess and treat tumor progression over time. This study used conventional MRI tumor segmentations to determine longitudinal growth differences between low grade (≤ G2), G3, and G4 gliomas. A total of 60 patients with at least three scans collected before death were organized into low-grade (n = 9), G3 (n = 13), and G4 (n = 38) gliomas at initial diagnosis. T1, T1C, FLAIR, and T2 images from every timepoint across all patients were used as input for tumor segmentation, using BraTS, to segment contrast enhancement (CE), FLAIR hyperintensity (FH), and the necrotic core (NC). Volumes within each segmentation and the time between each subsequent MRI was calculated, where the first post-surgical MRI was considered time 0. Linear mixed models were fitted for each segmentation volume, where Time, Group (i.e., G2-4) and their interaction (Group*Time) were fixed effects and subject was a random effect. We found that Time and Group*Time interaction were significant predictors of FH growth (p < 0.001 and 0.04, respectively); however, no group-level differences were observed in FH volumes over time. Group was the only significant predictor of CE growth (p = 0.006). Specifically, G4 growth was greater than G2 and had a trending increase than G3 (p = 0.003 and 0.08, respectively). Group, Time, and Group*Time were all significant predictors of NC growth (p = 0.04, 0.001, and 0.01, respectively). G4 NC growth was significantly greater than both G2 and G3 (both p < 0.05). No difference any of these tumor segmentations growth between G2 and G3 was detected. These results further show the heterogeneity in glioma growth over time and the necessity to assess how treatment effects longitudinal tumor composition progression and overall survival.