Spatial Overlapping between the Subvolumes with Elevated CBV and with Hypercellularity in Glioblastoma
Hemant Parmar1, Daniel Wahl2, Priyanka Pramanik2, Michelle Kim2, Theodore S Lawrence2, and Yue Cao1,2

1Radiology, University of Michigan, Ann Arbor, MI, United States, 2Radiation Oncology, University of Michigan, Ann Arbor, MI, United States


Standard imaging for glioblastoma relies on post-contrast T1 and T2 FLAIR MRI sequences, which do not accurately reflect tumor biology. Advanced MRI techniques can define biologically relevant and prognostic features of glioblastoma including perfusion-based volumes with high cerebral blood volume (VhCBV) and high b-value diffusion-based hypercellular subvolume (HCV). We defined VhCBV and HCV in 24 patients with glioblastoma prior to undergoing chemoradiation. Surprisingly, there was little overlap between VhCBV and HCV within individual patients, which suggests that these volumes represent distinct aspects of tumor biology and may be independently prognostic. Analysis of failure patterns and prognostic relevance is ongoing.


Tumor volumes of glioblastoma (GBM) defined on post-contrast T1 weighted and T2 FLAIR images underestimate and overestimate gross tumor volume, respectively, resulting in inadequate target definition for surgery and radiation therapy. However, elevated CBV in GBM is an established biomarker for prognosis.1-3 Recent work also shows that the hypercellularity abnormality subvolume (HCV) and non-enhanced HCV in GBM identified by high b-value diffusion weighted imaging are significant prognostic indicators.4 To what extent elevated CBV and HCV are complementary or redundant is unknown. This study aimed to investigate the spatial relationship between elevated CBV and HCV and whether combining them leads to better radiation target definition and prediction for outcomes.


The study was approved by IRB. Twenty-four patients (age: 24-77 years) with newly diagnosed GBM had MRI scans prior to concurrent chemoradiation therapy (RT). All MRI scans were performed on a 3T scanner (Skyra, Siemens). Scans included T1 weighted pre- and post-contrast, T2 FLAIR, T1 weighted dynamic contrast enhanced (DCE), and diffusion weighted (DW) images. CBV maps were computed by fitting the generalized Toft model to the DCE images that were acquired with temporal resolution of ~3 s at ~2mm isotropic voxel size. DW images were acquired in 3 orthogonal directions with b-values of 0, 1000, and 3000 s/mm2. Using the high b-value (3000) DW images, in which fluid, edema, grey matter, and white matter for an extent are suppressed, the HCV of each patient was determined by a threshold (mean intensity + 2SD) obtained from a VOI in normal appearing tissue and contralateral to the tumor. Similarly, the high CBV volume in GBM (VhCBV) was determined by a threshold (mean + 1 SD) obtained in contralateral grey matter. After co-registration of CBV, DW, FLAIR, post-contrast T1-weighted images pre RT, planned dose volumes and MRI at recurrence, spatial overlapping between the VhCBV and HCV in GBM as well as with the gross tumor volume (GTV_Gd) on Gd-enhanced T1 weighted images, abnormality volume on FLAIR images, 95% radiation dose volume and pattern of failure were compared. A Jaccard similarity index was used to quantify the spatial overlap between the VhCBV and HCV.


Of the 24 patients, the VhCBV varied from 0.1 to 66 cc with a median of 5.5 cc, and the HCV from 0.1 to 66 cc with a median of 7.6 cc. Both were much smaller than tumor volumes obtained from FLAIR images (a median of 63 cc with a range from 17.8 to 180 cc). There was little overlap between VhCBV and HCV, with overlapping volumes ranging from 0 to 4.4 cc with a median of 0.9 cc, resulting in a median Jaccard similarity index of 7.9% (range: 0 to 50%). Only five patients had the Jaccard index greater than 20% (Fig. 1), suggesting the VhCBV and HCV represent different aspects of tumor biology or phenotype. Combining the VhCBV and HCV into a single volume led to a median of 13.4 cc (range: 2.6 to 92 cc), which was much smaller than the FLAIR tumor volume and had a similar volume to the GTV-Gd (a median of 27.5 cc and a range from 2.3 to 93.9 cc).


Our data suggest that elevated CBV and hypercellularity in GBM represent two different tumor properties. Both should be considered for tumor response assessment. Combining the two should be considered as a target for intensification radiation therapy. The evaluation of the predictive power of combining these two imaging biomarkers on clinical outcomes is ongoing.


This work is supported in part by NIH/NINDS/NCI RO1 NS064973 (Cao).


1. Cao et al, Int J Rad Onc Biol Phys, 64(3):876-885, 2006.

2. Law et al Radiology, 247(2):490-498, 2008.

3. Jain et al Radiology, 267(1): 212-220, 2013.

4. Pramanik et al, Int J Rad Onc Biol Phys, 92(4):811-819, 2015.


Fig. 1. Four left panels: Diffusion weighted images with b=3000 s/mm2 (grey) and CBV (color) of two cases with Jaccard indices of 20% and 29%. Yellow contour: HCV; Red contour: VhCBV. Red arrows point to the regions without the overlap of the two volumes. Right panel: Jaccard indices for all patients.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)