2905

Comparison of quantitative blood oxygenation imaging methods for the assessment of brain tumor oxygen consumption in native MRI.
Jan Sedlacik1, Julia Götz1, Patrick Borchert1, Divya Bolar2, Lasse Dührsen3, Nils Ole Schmidt3, Jan-Hendrik Buhk1, and Jens Fiehler1

1Neuroradiology, UKE, Hamburg, Germany, 2Radiology, UCSD, San Diego, CA, United States, 3Neurosurgery, UKE, Hamburg, Germany

Synopsis

The comparison of the blood oxygen extract fraction (OEF) of solid tumor regions between the QUantitative Imaging of eXtraction of Oxygen and TIssue Consumption (QUIXOTIC) and quantitative Blood Oxygenation Level Dependent (qBOLD) methods showed an opposite trend of qBOLD-OEF and QUIXOTIC-OEF between tumor and cortical GM. Further analysis suggests that QUIXOTIC-OEF may be compromised by the strong and long lasting magnetization of the increased interstitial water due to tumor edema and, therefore, no reliable parameter to assess tumor OEF. On the other hand, qBOLD-OEF may have correctly detected a higher OEF of solid tumor regions.

INTRODUCTION

QUantitative Imaging of eXtraction of Oxygen and TIssue Consumption (QUIXOTIC) is a velocity-selective spin labeling technique, which is able to assess the magnetization of the post capillar venular blood at different spin-echo times on a voxel-by-voxel basis [1]. Since R2 of blood depends on the blood's oxygen saturation and hematocrit, an oxygen extract fraction (OEF) map can be determined from the R2 map of the venular blood. On the other hand, the quantitative Blood Oxygenation Level Dependent (qBOLD) method assesses the blood oxygenation by modeling the effect of the blood capillary network on the tissue magnetization measured by gradient echo sampled spin-echo (GESSE) [2]. Beside OEF, the qBOLD method also obtains the deoxygenated blood volume (DBV) of the underlying blood capillary network. However, the parameters of the qBOLD method show quite an amount of interdependency between each other leading to instabilities in the fitting of the qBOLD model especially for measurements with low signal-to-noise ratio [3]. Also, tissue R2 values can be determined by both methods. The aim of this work was to compare the OEF values obtained by both blood oxygenation imaging methods in tumor patients.

METHODS

29 patients with preoperative gliomas (WHO grades 2-4) were imaged at 3T. The subjects gave written informed consent prior to inclusion into the study, which was approved by the local ethics committee. Fluid-attenuated inversion recovery (FLAIR) and diffusion-weighted (DWI) images were used to locate the tumor and the solid tumor regions (Fig.1). The QUIXOTIC sequence was scanned three times with three different T2prep times 20, 40 and 60ms, 20 labeled and 20 control images, 4x4mm2 in-plane resolution, 64x64 matrix, 5.5mm slice thickness, TE=12ms, TR=3000ms and 2min acquisition duration for each T2prep time. GESSE was acquired with gradient echo times TEs=20-80ms, DTE=4ms, spin echo occurring at 30ms, TR=3000ms, 3x3mm2 in-plane resolution, 64x64 matrix, 4mm slice thickness, averages=2, acquisition time 5:42 min. ROIs were transferred to QUIXOTIC and qBOLD images after co-registering to FLAIR and DWI using NiftyReg [4]. To improve the signal to noise ratio and therefore robustness of the quantitative parameter calculation [3], quantitative parameters used for statistical analysis were calculated based on median ROI values of raw image data and were not determined from precalculated voxel-by-voxel parameter maps (Fig.2). Data reconstruction was done using in house developed Matlab code followed by statistical analysis using SPSS (Wilcoxon-Mann-Whitney Test and Spearman's rank correlation coefficient).

RESULTS

Box plots of all analyzed parameters are shown in Fig.3. Only R2ex of the qBOLD method and tissue R2 of the QUIXOTIC method showed a statistically significant difference (p<0.001) between tumor and cortical GM with lower R2 for tumor and higher R2 for cortical GM. Both, qBOLD-OEF and QUIXOTIC-OEF, show no statistically significant difference between tumor and cortical GM, however, qBOLD-OEF seems to be higher for tumor than for cortical GM. Contrarily, QUIXOTIC-OEF seems to be lower for tumor than for cortical GM. Correlation coefficients between all analyzed parameters and exemplary scatter plots are shown in Fig.4 and Fig.5. Here, qBOLD-R2ex and QUIXOTIC-R2 showed a strong and statistically significant correlation (rho=0.57). Interestingly, QUIXOTIC-OEF also showed a statistically significant correlation (rho=0.34) with QUIXOTIC-R2 and a little weaker correlation (rho=0.21) with qBOLD-R2ex. Furthermore, qBOLD-OEF and QUIXOTIC-OEF (rho=-0.24) showed a weak negative correlation as well as qBOLD-R2ex and qBOLD-DBV (rho=-0.24).

DISCUSSION and CONCLUSION

The lower tissue R2 values observed with both methods (qBOLD-R2ex and QUIXOTIC-R2) for the tumor are most likely caused by tumor edema due to increased amount of interstitial water. The strong correlation between both R2 measurements confirms their similar sensitivity. Interestingly, qBOLD-OEF and QUIXOTIC-OEF show an opposite trend between tumor and cortical GM, which is confirmed by their negative correlation with each other. However, the statistically significant correlation between QUIXOTIC-OEF and QUIXOTIC-R2 suggests that the magnetization of the post capillar venular blood may be compromised by the strong and long lasting magnetization of the increased interstitial water due to tumor edema. Therefore, the employed T2prep QUIXOTIC may not allow to reliably assess tumor OEF. On the other hand, the negative correlation between qBOLD-R2ex and qBOLD-DBV demonstrates their not fully solved interdependency even by using high signal-to-noise data from median raw image ROI values. However, the absent correlations of qBOLD-OEF with qBOLD-DBV and qBOLD-R2ex suggest a successful isolation of qBOLD-OEF from the interdependency with the other qBOLD parameters. Therefore, the higher solid tumor OEF may have been correctly detected by the qBOLD method.

Acknowledgements

We thank the German Research Foundation (DFG) for financial support: grant no. SE 2052/1-1.

References

1. Bolar DS, Rosen BR, Sorensen AG, Adalsteinsson E. QUantitative Imaging of eXtraction of oxygen and TIssue consumption (QUIXOTIC) using venular-targeted velocity-selective spin labeling. Magn Reson Med. 2011 Dec;66(6):1550-62.

2. He X, Yablonskiy DA. Quantitative BOLD: mapping of human cerebral deoxygenated blood volume and oxygen extraction fraction: default state. Magn Reson Med. 2007 Jan;57(1):115-26.

3. Sohlin MC, Schad LR. Theoretical prediction of parameter stability in quantitative BOLD MRI: dependence on SNR and sequence parameters. In: Proceedings of 18th Annual Meeting of ISMRM, Honolulu, Hawaii, 2009. (abstract 1623).

4. Modat M, Ridgway GR, Taylor ZA, Lehmann M, Barnes J, Hawkes DJ, Fox NC, Ourselin S. Fast free-form deformation using graphics processing units. Comput Methods Programs Biomed. 2010 Jun;98(3):278-84.

Figures

Figure 1: Exemplary ROI placement using FLAIR and DWI images. Solid tumor was defined as hyperintense on DWI.

Figure 2: Upper row: Tissue R2 and OEF maps resulting from the QUIXOTIC method. Lower row: R2ex, OEF and DBV maps resulting from the qBOLD method. Please note: quantitative parameters used for statistical analysis were calculated from median ROI values of raw image data and not from precalculated parameter maps.

Figure 3: Box-plots of the analyzed parameters for solid tumor and cortical GM ROI.

Figure 4: Exemplary scatter-plots of statistically significant correlations (p<0.05) and weaker correlations of interest. Please see Fig.5 for all tested parameters.

Figure 5: Table of Spearman's rank correlation coefficients and p-values. P-values below 0.05 were highlighted by bold font style.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)
2905