Synopsis

Direct ¹⁷O-MRI enables quantification of cerebral metabolic rate of oxygen consumption (CMRO₂). The low MR sensitivity of the ¹⁷O nucleus prevents pixel-wise CMRO₂ quantification and the fast T₂^*≈2ms decay leads to partial volume artifacts. In this work the DIESIS method is proposed, which performs a direct least squares estimation of the ¹⁷O‑MR signal in image regions (parcels) obtained from coregistered ¹H data. CMRO₂ values of 1.38-1.87µmol/gtissue/min and 0.51-0.77µmol/gtissue/min were found with DIESIS in gray and white matter consistent with rates from ¹⁵O-PET studies. With DIESIS, CMRO₂ maps of high resolution can be reconstructed and partial volume artifacts can be reduced.

Introduction

Cerebrovascular, neurodegenerative diseases and tumors are coupled to abnormalities in brain oxygen metabolism, which manifest in alternated cerebral metabolic rate of oxygen consumption (CMRO₂). CMRO₂ can be quantified with radioactive ¹⁵O‑PET [1], and direct or indirect ¹⁷O-MRI methods. In direct ¹⁷O-MRI [2–5], CMRO₂ rate can be reliably quantified by fitting a pharmacokinetic model to the ¹⁷O‑MR signal dynamic during and after the administration of ¹⁷O-enriched gas [5].

Besides the intrinsically low MR sensitivity of ¹⁷O (1.1·10^-5 compared to ¹H), ¹⁷O‑MRI is limited by partial volume artefacts mainly caused by the fast T₂^* decay. Partial volume correction (PVC) can be achieved with the ‘geometric transfer matrix’ (GTM) algorithm [4]. In this study we present an alternative correction scheme: DIrect EStimation of the MR ImageS (DIESIS) without Fourier transformation (FT) of k‑space signal into image space, which was motivated by direct imaging of functional networks in functional MRI (fMRI) [6].

Material and Methods

¹⁷O data sets of one volunteer were acquired in two dynamic¹⁷O-MR experiments [5], denoted by Exp1/Exp2, on clinical 3T MR system (Siemens Magnetom TIM Trio) with a custom-built Tx/Rx ¹⁷O head coil and a 3D ultrashort-TE density-adapted projection sequence [7]. 2.5 liters of 70%‑enriched ¹⁷O gas (NUKEM Isotopes Imaging) were inhaled, and the nominal spatial resolution was Δx=10/8mm at a temporal resolution of 1min.

At clinical field strengths, the low SNR of the ¹⁷O data hampers pixel-wise CMRO₂ quantification, so that the signal average over larger regions-of-interest (ROIs) has to used [3-5]. Thus, CMRO₂ values can only be computed from a small subset of large regions with mixing of tissues such as white (WM) and gray (GM) matter.

Radially acquired images are typically reconstructed by Kaiser-Bessel (KB) gridding followed by FT (Fig.1, upper part). We propose a new method (DIESIS; Fig.1, bottom part) that avoids full reconstruction of MR images (Fourier encoding) and directly estimates the ¹⁷O signal based on parcellation (segmentation of the 3D image to a predefined number of ROIs on the basis of a coregistered ¹H image). A similar method was originally discussed for application in fMRI direct imaging of functional networks [6], where a few hundred of predefined parcels were used. Here we define 3 parcels for WM, GM and cerebrospinal fluid (CSF) in the brain and 180 outside the brain (including regions outside the head) to minimize the influence of susceptibility artifacts, volunteer motion under radial sampling and coil inhomogeneities.

DIESIS models the MR image as the sum over parcel’s masks $$$\mathit{M_{i}(r)}$$$ weighted by the intensity values $$$\mathit{\lambda_{i}}$$$:

$$I(\textbf{r})=\sum_i^{\sharp\space{parcels}}M_i(\textbf{r})\cdot\lambda_i.$$

The k-space signal for one spoke ($$$\mathit{sp}$$$) in radial acquisition of i_th parcel is then:

$$S^{sp}_i(t)=\int{}dr\cdot{}M_i(\textbf{r})\cdot{}e^{i\cdot\overline{\textbf{r}}\cdot\overline{\textbf{sp}}\cdot\gamma\cdot{}t}e^{\frac{-t}{T_2^*}},$$

where the last term with experimentally determined $$$T_{2}^{*}=2ms$$$ [8] is optionally included for explicit PVC.

The measured ¹⁷O-MRI k-space data ($$$\mathit{S_{meas}}$$$) is expressed as a linear combination of the calculated k-space signal of the parcels $$$\mathit{S_{i}(t)}$$$ for the particular k-space trajectories:

$$S_{meas}(t)=\sum_i^{\sharp\space{parcels}}S_i(t)\cdot\lambda_i.$$

The intensity values $$$\mathit{\lambda_{i}}$$$, which are to be mapped back to image space, are then obtained by direct least squares estimation without Fourier encoding by minimization of:

$$LL(\lambda)=\mid\mid{}S_{meas}(t)-\sum_i^{\sharp\space{parcels}}S_i(t)\cdot\lambda_i\mid\mid{}_{2}.$$

DIESIS was applied to each raw data set of the ¹⁷O MR time-series, and CMRO₂ values were extracted from the signal dynamics (Fig.1).

Results and Discussion

Based on DIESIS, the ¹⁷O-MR signal dynamics in WM and GM could be successfully fitted to the pharmacokinetic model [2,5] (Fig.2), and although the ¹⁷O-MRI is corrupted with noise (Fig.3b,f), CMRO₂ values (Fig.3c,d,g,h) for both ¹⁷O‑MR experiments were in good agreement with the results of ¹⁵O‑PET studies (Tab.1).

CMRO₂ values obtained from conventional KB gridding are 45-70% higher in WM and 8-15% lower in GM as compared to ¹⁵O-PET studies (Tab.1). The DIESIS method without explicit PVCs provided better values for WM (3-31% overestimation). DIESIS including fast T₂^* relaxation yielded CMRO₂ values within 0-18% of the ¹⁵O-PET values [1]. Moreover, DIESIS allows suppressing the signal increase in CSF, which leads to substantially lower CMRO₂ values in CSF, where no oxygen metabolism is expected. PVCs of the CMRO₂ values with DIESIS method are in better correspondence with ¹⁵O-PET studies, compared to the results of GMT method [4].

Conclusions and Outlook

Acknowledgements

References

[1] Leenders KL, Perani D, Lammertsma AA, Heather JD, Buckingham P, Healy MJ, Gibbs JM, Wise RJ, et al. Cerebral blood flow, blood volume and oxygen utilization. Normal values and effect of age. Brain 1990;113:27–47.

[2] Atkinson IC, Thulborn KR. Feasibility of mapping the tissue mass corrected bioscale of cerebral metabolic rate of oxygen consumption using 17-oxygen and 23-sodium MR imaging in a human brain at 9.4 T. Neuroimage 2010;51:723–733.

[3] Hoffmann SH, Begovatz P, Nagel AM, Umathum R, Schommer K, Bachert P, Bock M. A measurement setup for direct ¹⁷O MRI at 7 T. Magn. Reson. Med. 2011;66:1109–1115.

[4] Hoffmann SH, Radbruch A, Bock M, Semmler W, Nagel AM. Direct ¹⁷O MRI with partial volume correction: first experiences in a glioblastoma patient. MAGMA 2014;27:579–87.

[5] Kurzhunov D, Borowiak R, Hass H, Wagner P, Krafft AJ, Timmer J, Bock M. Quantification of oxygen metabolic rates in Human brain with dynamic ¹⁷O MRI: Profile likelihood analysis. Magn. Reson. Med. 2016;00:1–11.

[6] Wong EC. Direct imaging of functional networks. Brain Connect. 2014;4:481–6.

[7] Nagel AM, Laun FB, Weber M-A, Matthies C, Semmler W, Schad LR. Sodium MRI using a density-adapted 3D radial acquisition technique. Magn. Reson. Med. 2009;62:1565–1573.

[8] Borowiak R, Groebner J, Haas M, Hennig J, Bock M. Direct cerebral and cardiac ¹⁷O-MRI at 3 Tesla: initial results at natural abundance. MAGMA 2014;27:95–99.