504 90th Percentile: Long-Term Glioblastoma Survivors Display Favorable Preoperative Radio-Pathomic Maps of Tumor Cell Density

INTRODUCTION: Glioblastoma (GBM) is the most common primary CNS malignancy, with one- and five-year survival rates of 41% and 5%, respectively. Extent of resection correlates with increased survival, but tools to predict long-term survivors (LTS) after gross-total resection (GTR) are lacking. Radio-pathomic maps of cell density, trained using MRI and autopsy tissue, identify tumor features not readily identifiable on traditional MRI. This study evaluates these maps in long-term GBM survivors. METHODS: Using the UCSF-PDGM dataset, we analyzed survival distribution in patients with GBM (n=220) who underwent GTR to identify LTS (>1030 days). Conventional pre-surgical MRI sequences including pre- and post-contrast T1-weighted, FLAIR, and ADC images, were used to compute radio-pathomic maps of cell density and tumor probability using a previously published and validated method. Briefly, 5 by 5 voxels from each acquisition predicted voxelwise cell density using aligned autopsy tissue samples as ground truth. Mean cell density and tumor probability maps were calculated within areas of contrast enhancement, necrotic core, and non-enhancing FLAIR hyperintense regions. Differences between LTS and non-LTS were tested using Student’s T-tests. RESULTS: LTS (n=32) had significantly lower mean cellularity within contrast-enhancing areas and necrotic tumor core (p<0.001 for both) compared to non-LTS (n=188). Tumor probability maps also showed lower intensity values in these areas (p=0.007, p=0.002). Differences in non-enhancing FLAIR hyperintense regions were not statistically significant (p=0.37, p=0.5). CONCLUSIONS: In GBM patients undergoing GTR, those in the 90th percentile for survival had lower cellularity and tumor prediction map values in contrast-enhancing areas and necrotic tumor core. These results suggest radio-pathomic maps reveal inherent favorable characteristics not seen on conventional MRI, potentially aiding neurosurgeons in GTR decision-making. Future research is needed to understand these relationships and their prognostic utility.