PURPOSE

Metabolic imaging is possible by employing dynamic ¹⁷O-MRI [1] and is leading to robust and reproducible quantification of the cerebral metabolic rate of oxygen consumption (CMRO₂) in volunteers [2] and patients [3]. The CMRO₂ parameter can be measured via dynamic ¹⁷O-MRI of the stable oxygen isotope ¹⁷O employing an inhalation procedure where enriched ¹⁷O₂-gas is administered during continuous imaging [1]. Recently, the first results of a first application in a brain tumor patient cohort [3] demonstrated the high specificity and metabolic contrast achievable: the whole tumor volume (TV, segmentation based on pure anatomical data) showed a significantly reduced CMRO₂ compared to healthy tissue. Dynamic methods often have the caveat that additional parameters like perfusion or cell density (e.g. in tumors) bias the true metabolic information. The presented work focusses on sub-compartment analysis of dynamic ¹⁷O-MRI data in correlation between the apparent diffusion coefficient (ADC) and the cerebral blood volume (CBV) in high grade glioma patients.

METHODS

Dynamic ¹⁷O-MRI experiments were performed on a 7-Tesla-whole-body-scanner [4] with a nominal spatial resolution of (7.5mm)³ employing a density-adapted radial sequence [5] with a Golden Angle acquisition scheme [6] (TR/TE=20/0.56ms, acquisition time t=30:00min) under administration of 3.8±0.1L of 70%-enriched ¹⁷O₂-gas [7] . The reconstructed temporal resolution was Δt=1:00min. In total datasets of three WHO Grade IV were included. The workflow for partial volume (PV) correction of dynamic data was used as previously described [2,3] to obtain CMRO₂ information in TV and healthy tissue (normal appearing gray and white matter (NGM, NWM), CSF). Contrast-enhanced (CE) MPRAGE data (CE-T_1w-MPRAGE, Fig.1A) were used for segmentation into masks of necrotic region (NE), contrast-enhancing region (CE) and peritumoral edema (PE) in n=3 patients. Additional information of clinical protocols (ADC/ CBV maps; Fig.1B,C) were coregistered to ¹⁷O-data sets of dynamic inhalation experiments and used for a second and third segmentation entity (3D-volume, ~20 slices). This classification based on ADC or CBV maps provided volumes of hyper- and hypo-intense volumes (CBV+/ADC-) for further, separate analysis of CMRO₂. An analysis of variance (ANOVA) with Holm-Sidac pairwise comparisons was performed to test for statistically significant differences.

RESULTS

All patient data in healthy tissue is showing a significantly increased CMRO₂ values in NGM compared to NWM (p<0.001). The quantitative evaluation yielded CMRO_2,NGM=2.00–2.55±0.20^μmol/_g*min compared to white matter tissue (CMRO₂=0.72–0.92±0.09^μmol/_g*min). Quantitative evaluation of healthy tissue and the tumor region of all patients in listed in Tab.1. Here one evaluation focused on CMRO₂ values in NE, CE and PE volumes; a second evaluation on CBV+ areal and a third evaluation on ADC- segmentation (NE-volume is always excluded from evaluation). Tumor metabolization is showing a significant functional decrease in the tumor tissue, with lowest values in NE as well as CE tumor region, compared to NGM or NWM (p<0.01). CBV analysis yields comparable metabolic information in CBV+ (strongly perfused) areas and no increased CMRO₂. The same is observed in ADC- volumes where in both cases no significant change is seen to TV or CE evaluation (p>0.05). A complete overview is given in Fig.2. Maps of relative signal change with correlated anatomical information as well as segmentation outlines for all three segmentation approaches can be found in Fig.1.

DISCUSSION

Dynamic ¹⁷O-MRI is providing a specific metabolic contrast, seen in the significant drop in CMRO₂ values of malignant tissue. To exclude effects originating from cell density or varying blood perfusion an analysis based on CBV and ADC maps are connected to the metabolic information. A significantly reduced (p<0.01) CMRO₂ in CBV+ and ADC- regions to healthy tissue and comparable values to evaluation based on CE-data demonstrates a perfusion and ADC independent information. Due to low spatial resolution and rapid transverse relaxation, the dynamic quantitative ¹⁷O-signal might be underestimated despite the application of a dedicated PVC algorithm. As discussed in previous studies [2] the main sources of error are the uncertainty in the prior ¹⁷O-enrichment and limitations of the PVC. However, quantified metabolic information in NGM and NWM are in good agreement with previous work [1,2]. For better statistical evaluation a larger number of examinations would be desirable, but nevertheless the first results and number of patients provide sufficient information to assess ADC and CBV correlation in a first trial.

CONCLUSION

This study presents data of the first examined patient cohort of dynamic ¹⁷O-MRI in a sub-compartment analyses focusing on ADC and CBV correlation. Results support the hypothesis that this method provides a very specific and pure metabolic information independent of tissue perfusion and cellular density. Increasing of total patient number a correlation to other modalities, e.g. PET could establish ¹⁷O-MRI as a tool to provide further insight into tumor pathophysiology beyond anatomical borders.

Acknowledgements

The authors want to thank NUKEM Isotopes Imaging GmbH for their generous supply of ¹⁷O_2-gas and support of this project.

References

[1] Atkinson IC, Neuroimage 2010 (51):723-733; [2] Niesporek, SC, Magn Reson Med 2018; 79: 2923–2934; [3] Niesporek SC, In Proc. ISMRM 2018, #625; [4] Siemens Healthineers, Erlangen, Germany; [5] Nagel et al., Magn Reson Med 2009 (62):1565-73; [6] Chan, R.W. et al., Magn Reson Med 2009(61): p. 354–363; [7] NUKEM Isotopes Imaging GmbH, Alzenau, Germany.