Synopsis

Keywords: Quantitative Imaging, Oxygenation

The brain’s oxygen extraction fraction (OEF) is a key parameter in assessing cerebrovascular function and brain oxygen consumption. We investigated the feasibility of measuring cerebral venous oxygenation (SvO₂), the key determinant of OEF, based on R2* estimated from a clinical susceptibility weighted imaging protocol and vessel segmentation aided by gadolinium administration, in a cohort of patients with Multiple Sclerosis. We were able to estimate venous oxygenation in the superior sagittal sinus, straight sinus, transverse sinuses and internal cerebral veins. The global median SvO₂ ranged between 0.60 and 0.68, within the range of values expected in the human brain.

Introduction

The measurement of cerebral venous oxygen saturation (SvO₂) is necessary for quantifying the brain’s oxygen extraction fraction and, therefore, the assessment of the cerebral metabolic rate of oxygen consumption. Particularly, three-dimensional (3D) mapping of cerebral SvO₂ would allow investigation of regional oxygenation in health and its alteration in disease. Recently, it was shown that a map of SvO₂ in large veins could be calculated based on voxel-wise R2* values of venous blood¹. This method relied on a multi-echo gradient-echo, incorporating flow compensation and velocity encoding (phase contrast) gradients. However, this sequence is not currently readily available in a clinical setting.
On the other hand, susceptibility-weighted imaging protocols (SWIp) are increasingly used in the clinic to calculate SWI images, for which multi-echo gradient-echo data are intrinsically acquired. Hence, this study aimed to assess whether the multi-echo signal magnitude from a clinical SWIp sequence could be used to calculate 3D maps of SvO₂ based on R2*, in large veins, in a clinical cohort of Multiple Sclerosis patients. In addition to using the SWIp sequence, we exploited the administration of gadolinium concentration to permit us to segment the large cerebral veins from T1-weighted images.

Methods

Thirty-one subjects with relapsing-remitting Multiple Sclerosis (age: 36.4 ± 12.4 years, 14/16 M/F, disease duration: 12.0 ± 10.1 years, Expanded Disability Status Scale (EDSS) score = 3.3 ± 1.9) underwent MRI on a Philips Achieva 3 T system using a 32-channel receive-only head coil. As part of the imaging protocol, the following sequences were used: 3D T1-weighted before and after gadolinium-based contrast injection and 3D T2*-weighted multi-echo susceptibility-weighted imaging (SWIp) before contrast injection. The acquisition parameters are shown in Table 1. For the SWIp acquisition, the raw complex images were obtained via delayed reconstruction.
Images from each subject were processed as follows. First, a parametric map of the transverse relaxation rate (R2*) was calculated by voxel-wise linearly fitting the log-linearised signal magnitude over echo times (TEs) according to:
$$ln(S) = ln(S_0) - TE\cdot R2^* \quad (1)$$
Second, a preliminary segmentation of the cerebral vessels was calculated by rigidly aligning (FSL FLIRT^2,3) the T1-weighted image acquired before contrast injection to the one acquired after contrast injection, and then subtracting the images. The resulting difference image was thresholded empirically and binarised. The resulting vessel mask was then rigidly aligned to the first-echo SWIp magnitude image using FSL FLIRT. Based on this preliminary segmentation in SWIp space, the superior sagittal sinus, straight sinus, transverse sinuses, and internal cerebral veins were delineated using the manual segmentation tool in ITK-SNAP⁴.
For each subject and in each venous region of interest (ROI), voxel-wise venous oxygen saturation (SvO₂) values were calculated based on the R2* map and the following calibration relation¹:
$$R2^* = A^*+C^*(1-SvO_2)^2 \quad (2)$$
Here, A* = 12.7 Hz and C* = 101.1 Hz assuming a haematocrit value of 0.39¹. Median and standard deviation values of SvO₂ were also calculated in each venous ROI both globally and across subjects.

Results and Discussion

A representative segmentation of the major veins is shown in Figure 1, overlaid on the first-echo magnitude image from the SWIp protocol. These results indicate that combining the pre-contrast and post-contrast T1-weighted images enables segmentation of the main cerebral veins. A representative SvO₂ map is shown in Figure 2, overlaid on the corresponding R₂* map. Here, most venous ROIs exhibited a fractional SvO₂ in the range 0.5-0.7. Histograms of all voxels of all subjects and histograms of the median values in all subjects are respectively shown in Figures 3 and 4, providing a quantitative summary of SvO₂ in this cohort. In the venous ROIs, the global median SvO₂ ranged between 0.60 and 0.68 (Figure 3). Similarly, the median SvO₂ calculated in individual subjects ranged between 0.58 and 0.68 (Figure 4). These results were in agreement with SvO₂ values previously reported in the literature, including those measured by Mao et al.¹ This simple approach remains to be validated against dedicated SvO₂ measurements such as those provided by T2-Relaxation-Under-Spin-Tagging (TRUST)⁵.

Conclusions

Clinical protocols acquiring SWIp and pre-contrast and post-contrast T1-weighted images may enable calculating parametric maps of venous oxygenation in the major cerebral veins without adding a dedicated sequence. This approach may open the way to a more widespread evaluation of cerebral oxygen extraction and cerebral oxygen metabolism for clinical research.