Correlation of Vessel Abnormality Quantification from TOF MRA with Histologic and Gene Markers
Maddalena Strumia1,2,3, Wilfried Reichardt1,2,3, Ori Staszewski4, Dieter Henrik Heiland5, Astrid Weyerbrock5, Irina Mader6, and Michael Bock2

1German Cancer Research Center (DKFZ), Heidelberg, Germany, 2Medical Physics, University Medical Center Freiburg, Freiburg, Germany, 3Deutsches Konsortium für Translationale Krebsforschung (DKTK), Heidelberg, Germany, 4Neuropathology, University Medical Center Freiburg, Freiburg, Germany, 5Neurosurgery, University Medical Center Freiburg, Freiburg, Germany, 6Neuroradiology, University Medical Center Freiburg, Freiburg, Germany

Synopsis

The glioblastoma is one of the most malignant and aggressive tumors of the brain-white-matter, and it expressing vascular-endothelial-grow-factors to initiate and maintain angiogenesis. To assess whether a novel therapeutic drug inhibits neo-angiogenesis, non-invasive imaging methods to study changes of neo-angiogenic vasculature are needed. In this work we show a correlation between a quantification measure of the pathological vessels structure, the dot product, and histology and genetic markers, in glioblastomas using TOF-MRA data. The correlation has been applied to 11 patients and shows that morphologic markers can be correlated with histologic and genetic markers to assess tumor vasculature.

Introduction

The glioblastoma is one of the most malignant and aggressive tumors of the brain. Glioblastomas are very heterogeneous, they are growing diffusely and they are expressing vascular endothelial grow factors (VEGF) to initiate and maintain angiogenesis. The suppression of the VEGF signal cascade is currently a target for therapeutic approaches. To assess whether a novel therapeutic drug inhibits neo-angiogenesis, non-invasive imaging methods to study changes of the neo-angiogenic vasculature are urgently needed. In this work we show a correlation between a quantification measure of the structure of the pathological vessels and histology and genetic markers, in glioblastomas using Time-Of-Flight MR Angiography (TOF-MRA) data.

Materials and Methods

TOF-MRA data were acquired in 11 glioblastoma patients as part of a routine imaging protocol. For MR imaging, a 3T MR system (Siemens Tim Trio) and a 24-channel head coil were used. The 3D TOF-MRA protocol was defined with the following parameters: TE = 4.12 ms, TR = 23 ms, α = 18°, 84 partitions, slab thickness: 0.55cm, in plane resolution of 0.39×0.39 mm2 and acquisition time of about TA = 3 min.

For every patient the tumor and a normal-appearing contralateral (nac) region were manually annotated by an expert physician on T2w images (Fig. 1a), resulting in two binary masks. Subsequently, the T2w image was rigidly registered with 3D Slicer [1] to the TOF-MRA data to delineate the tumor and the nac boundaries on the TOF-MRA image (Fig. 1b).

To quantify the abnormality of the tumor vessels, initially, a segmentation of the vasculature in the TOF-MRA image was performed with the EMILOVE method [2]. On the vessel mask a dot product $$$\overline{D}$$$ measure was calculated as described earlier [3], which quantifies the local (in)-coherency of the vessels orientations. The dot product was computed in both the tumor and nac regions of each patient data, resulting in $$$\overline{D_{tumor}}$$$ and $$$\overline{D_{nac}}$$$.

For histologic correlation also the proliferation marker Ki-67[MIB1] was determined in tissue samples retrieved during tumor surgery. Ki-67[MIB1] describes the percentage of all tumor cells that are in a mitotic state. $$$\overline{D_{tumor }}$$$ was plotted as a function of MIB percentage for the tumor, and for comparison also the nac values are included (note, that no samples were taken from the nac region). A linear regression was calculated for the values of the tumor. In 9 out of 11 patients genome-wide expression analysis was performed (RNAeasy kit, Qiagen with 1.5 µg RNA). Arrays were analyzed by human genome 2.0 chip (Affymetrix), and expression data were normalized and log2-transformed. $$$\overline{D_{tumor}}$$$ values and gene expression were Spearmann-ranked correlated. Positive and negative correlated genes (>0.8) were extracted and unsupervised clustered to identify expression-based subgroups. Survival analysis was done by Kaplan-Meier statistics. Progression-free survival (PFS) was available for 10 patients: 4 patients had a tumor progression (range: 115 – 392 days), and 6 patients did not (observation interval: 84 – 511 days).

Results and Discussion

The comparison with histology shows that a slight negative linear correlation exists between MIB and $$$\overline{D_{tumor}}$$$, i.e. $$$\overline{D_{tumor}}$$$=(0.38±0.04)+(-0.0015±0.0011)∙MIB (Fig. 2). In the genetic analysis, two subgroups could be identified by unsupervised Spearmann-ranked clustering of pre-ranked genes. Both subgroups showed significant (p=0.003) difference $$$\overline{D_{cluster I}}$$$ (0.37±0.01) and $$$\overline{D_{cluster II}}$$$ (0.31±0.02) (Fig. 3a). In addition, both cluster I and cluster II had different progression-free survival outcome (cluster I: mean 9.93 CL95% 7.07-12; cluster II: mean 5.54 CL95% 3.10-9.98) (Fig. 3b). These differences are statistically not significant in Cox-regression analysis (p=0.08). This preliminary study in a very limited patient cohort demonstrates that morphologic markers such as the dot product can be correlated with histologic and genetic markers to assess tumor vasculature. Our results show that the combination of the dot product with gene analysis and histologic correlation can help to connect macroscopic vessel appearance with cellular abnormalities in blood vessels and might therefore be suited to identify prognostic subgroups in glioblastoma.

Acknowledgements

No acknowledgement found.

References

1. Pieper S, et al. 3D Slicer, http://www.slicer.org.

2. Skibbe H, Reisert M., Maeda S., Koyama M., Oba S., Ito K., Ishii S. Efficient Monte Carlo Image Analysis for the Location of Vascular Entity. IEEE TMI. 2015;34(2):628-643.

3. Strumia M, Reichardt W., Mader I, Bock M. Vessels Abnormality Quantification in TOF-MRA. ISMRM. 2014.

Figures

Figure 1: T2w (a) and TOF-MRA (b) with tumor boundaries (red) and contralateral side boundaries (blue). Below vessels orientation in the two regions.

Figure 2: Dot product as a function of MIB for tumor and normal-appearing contralateral side (nac). A weak negative correlation (solid line) is seen between $$$\overline{D_{tumor}}$$$ and MIB, whereas no correlation was calculated for nac (dotted line represents mean value).

Figure 3: (a) Unsupervised cluster of gene expression which were positively (green) and negatively (red) associated to $$$\overline{D_{tumor}}$$$. (b) Kaplan-Meier survival curves of subgroup I (blue) and II (red).



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
2522