NIMG-41. ASSOCIATION BETWEEN PERFUSION IMAGING METRICS AND RADIO-PATHOMIC MAPS OF TUMOR PROBABILITY WITHIN AND BEYOND THE CONTRAST-ENHANCING VOLUME IN GLIOMA PATIENTS

Abstract

Abstract Perfusion MRI, using both contrast-based techniques such as dynamic susceptibility imaging (DSC) and non-contrast-based techniques like arterial spin labeling (ASL) have been used to monitor blood flow as a marker of angiogenic glioma presence, and showed substantial potential in the differentiation of true progression from pseudoprogression post-treatment. This study uses radio-pathomic maps of cellularity sensitive to areas of non-enhancing tumor to test the hypothesis that perfusion-derived markers of tumor probability will associate with tumor presence within contrast-enhancement but not beyond the enhancing margin. This study included pre-surgical relative cerebral blood volume (rCBV) estimates derived from DSC perfusion data from the PENN-GBM dataset (n=456) and presurgical ASL images from the UCSF-PDGM dataset (n=426) for analyses. The T1, T1C, FLAIR, and ADC images from each patient were used to generate radio-pathomic cell density maps using a previously validated model that excels at accurate detection of non-enhancing tumor presence. Mean cell density, rCBV, and ASL were computed for each corresponding patient within included segmentations for contrast-enhancement (excluding necrotic core) and FLAIR hyperintensity (excluding both necrotic core and enhancement). Pearson’s correlations were then used to quantify the association between perfusion-derived estimates and mean cell density within each region of interest. Both rCBV and ASL showed positive associations with cell density within contrast enhancement (rCBV: R=0.280, p< 0.001; ASL: R=0.117, p=0.016). However, both perfusion metrics also showed no association with cell density within the non-enhancing, FLAIR hyperintense region (rCBV: R=0.0162, p=0.731; ASL: R=-0.213, p=0.634). These results suggest that perfusion imaging successfully identifies angiogenic tumor but may be less efficacious in detecting non-enhancing tumor presence. Future studies should examine how perfusion’s relationship with hypercellular tumor presence evolves in the presence of treatment, as well as in relation to other advanced imaging metrics such as MR spectroscopy and CEST imaging.