Skip to main content
NCBI home page
Journal of Cerebral Blood Flow & Metabolism logoLink to Journal of Cerebral Blood Flow & Metabolism
. 2014 Jul 9;34(9):1550–1557. doi: 10.1038/jcbfm.2014.116

Tissue oxygen saturation mapping with magnetic resonance imaging

Thomas Christen 1,2,9, Pierre Bouzat 1,2,3,9, Nicolas Pannetier 1,2, Nicolas Coquery 1,2, Anaïck Moisan 1,2, Benjamin Lemasson 1,2, Sébastien Thomas 1,2,3, Emmanuelle Grillon 4,5,6,7, Olivier Detante 1,2,8, Chantal Rémy 1,2, Jean-François Payen 1,2,3, Emmanuel Luc Barbier 1,2,*
PMCID: PMC4158672  PMID: 25005878

Abstract

A quantitative estimate of cerebral blood oxygen saturation is of critical importance in the investigation of cerebrovascular disease. While positron emission tomography can map in vivo the oxygen level in blood, it has limited availability and requires ionizing radiation. Magnetic resonance imaging (MRI) offers an alternative through the blood oxygen level-dependent contrast. Here, we describe an in vivo and non-invasive approach to map brain tissue oxygen saturation (StO2) with high spatial resolution. StO2 obtained with MRI correlated well with results from blood gas analyses for various oxygen and hematocrit challenges. In a stroke model, the hypoxic areas delineated in vivo by MRI spatially matched those observed ex vivo by pimonidazole staining. In a model of diffuse traumatic brain injury, MRI was able to detect even a reduction in StO2 that was too small to be detected by histology. In a F98 glioma model, MRI was able to map oxygenation heterogeneity. Thus, the MRI technique may improve our understanding of the pathophysiology of several brain diseases involving impaired oxygenation.

Keywords: brain, MRI, oxygenation, stroke, trauma, tumor

Introduction

The sensitivity of magnetic resonance imaging (MRI) to changes in oxygenation through the blood oxygen level-dependent (BOLD) effect1 is the basis of functional MRI. This technique allows a non-invasive exploration of the functions and dysfunctions of the human brain and has revolutionized cognitive neuroscience over the last 20 years. Less appreciated is the potential of BOLD MRI to study baseline brain oxygenation. In a clinical environment, MRI oximetry would offer definitive advantages over brain oxygenation mapping with positron emission tomography (PET),2 a technique not widely available and that requires ionizing radiations.

Several authors have laid out the foundations of a theoretical framework named quantitative BOLD (qBOLD) imaging that aims to extract quantitative vascular information from baseline scans.3, 4 By analyzing the magnetic resonance (MR) signal evolution through a biophysical model, cerebral venous blood volume and venous blood oxygen saturation maps have been obtained in both humans and animals.5, 6, 7 This qBOLD approach, however, requires a specific MRI sequence, which limits its dissemination. To ease its use and improve its spatial resolution, our group has recently proposed combining several estimates obtained with routinely used clinical sequences, the multiparametric qBOLD (mqBOLD).8 While promising results have been obtained in the rat brain, mqBOLD still requires a robust validation.

In this study, we firstly evaluated whether the mqBOLD approach could be used to map the tissue oxygen saturation (StO2, the blood oxygen saturation in the voxel, averaged over the arterio-venous network). For this purpose, the linearity of the relation between MR_StO2 (the estimate obtained with mqBOLD) and StO2 (derived from arterial and venous blood gas measurements) was assessed using rats submitted to several inspired oxygen fractions. Secondly, the impact on MR_StO2 of systemic changes in the hematocrit fraction (Hct) and subsequent brain tissue stress was monitored. Finally, MR_StO2 maps were obtained on models of three common brain injuries: (i) acute stroke (severe hypoxia), (ii) diffuse trauma (mild hypoxia) and (iii) highly heterogeneous brain tumor, and the results were compared to ex vivo hypoxia staining.

Materials and methods

Each study design was approved by the ‘Grenoble – Institute of Neuroscience' local committee for animal care and use (agreement number 04). Experiments were performed under permits (no. 380820 for CR and A3851610008 for experimental and animal care facilities) from the French Ministry of Agriculture.

Numerical Simulations

The protocol for numerical simulation has been extensively described and validated in a previous work.9 We summarize here only the principal components. All simulations were performed on a Dell Precision computer with double quad 2.33 GHz Xeon processors (Intel, Santa Clara, CA, USA) and 8 GB of RAM. Calculations were performed in the Matlab (Mathworks, Natick, MA, USA) environment using custom software. The numerical simulation tool uses a Fourier-based approach to compute the magnetic field distribution inside the voxel10 and a deterministic approach to account for relaxation and diffusion processes.11, 12 This allows high computation time efficiency and a free choice for the geometry of the vascular network and tissue. For the sake of simplicity, straight cylinders were used here as blood vessels and voxels were placed in an external magnetic field B0 aligned with the x-axis. A diffusion coefficient (D=10−9 m2/s) and a magnetic susceptibility—that of blood or of tissue—were associated to each point and the evolution of the MR signal was estimated over 90 milliseconds at 1-millisecond time intervals. The mqBOLD protocol was simulated for various networks (from large arterioles to large venules via capillaries, radius from 1 to 50 μm) and for tissue blood oxygenation saturation values ranging from 40% to 90% (oxygen saturation in the femoral vein (SvO2)=20–80%, oxygen saturation in the femoral artery (SaO2)=80–100%). The blood volume fraction was kept constant at 3.5%.

Animal Preparation

For all experiments, rectal temperature was monitored and maintained at 37.0°C±0.5°C. Male Wistar (Fischer for tumor) rats (Charles River, L'Arbresle, France) were anesthetized with isoflurane (5% for induction, 2% for maintenance) in air, mechanically ventilated using a rodent ventilator (SAR-830/P, CWE Inc., Ardmore, PA, USA) and equipped with a catheter in the femoral artery and in the femoral vein (unless mentioned otherwise).

Transient Hyperoxia/Transient Hypoxia

Experiment 1. Superior longitudinal sinus versus femoral vein

Rats (n=12; cf. the section ‘Animal preparation') were equipped with an additional catheter in the superior longitudinal sinus. By varying the inspired oxygen fraction (FiO2) in N2, rats were submitted to hyperoxia (PaO2=350 mm Hg), normoxia (PaO2=150 mm Hg), moderate hypoxia (PaO2=65 mm Hg), and severe hypoxia (PaO2=40 mm Hg). One blood sample was collected from each vessel 10 minutes after setting the FiO2 level for blood gas analysis (see the section ‘Blood gas analysis' below).

Experiment 2. MR_StO 2 versus femoral vein

Rats (n=17; cf. ‘animal preparation') were submitted to the same FiO2 levels as in Experiment 1. For each FiO2 level, MR_StO2 was mapped and blood samples from the femoral vein and from the femoral artery were analyzed. Each rat was submitted to two PaO2 levels. The striatum, cortex, and corpus callosum regions of interest (ROIs) were manually defined in the MR images.

Experiment 3. Brain tissue oxygen tension (PbtO 2 ) versus femoral vein

Rats (n=5; cf. ‘animal preparation') were equipped with a PbtO2 probe (see the section ‘Determination of PbtO2' below). After 1 hour stabilization, rats were submitted to the same FiO2 levels as in Experiment 1. For each level, blood gas was measured in each vessel and the PbtO2 was noted after a 10-minute stabilization delay.

Figure 1.

Figure 1

Open in a new tab

Effect of the inspired oxygen fraction on MR_StO2 (blood tissue oxygen saturation) estimates (n=17). (A) Representative T2-weighted image, T2* image, cerebral blood volume (CBV, %) map, and MR_StO2 (%) map obtained in a healthy rat under normoxic conditions (experiment 2). CBV and MR_StO2 color maps were overlaid onto the T2-weighted images. (B) Partial arterial pressure of oxygen (PaO2; mm Hg) values obtained for the four inspired oxygen fractions, i.e., severe hypoxia, moderate hypoxia, normoxia, and hyperoxia. (C) PaO2 (mm Hg), partial arterial pressure of carbon dioxide (PaCO2, mm Hg), and mean arterial pressure (MAP, mm Hg) under the four oxygenation conditions of experiment 2. (D) Correlation between MR_StO2 (%, magnetic resonance imaging (MRI) measurement, experiment 2) in the cortex and StO2 calculated from arterial and venous blood gases (StO2=1/3 × SaO2+2/3 × SsO2). Representative MR_StO2 color maps (overlaid onto the T2-weighted images) for the different oxygenation conditions.

Intermittent Hypoxia and Hemodilution

Three groups of rats (n=19, 333±34 g; cf. ‘animal preparation' but without intubation) were studied:

  • Control group (n=8): rats were used as control.

  • Intermittent hypoxia group (n=6): for 14 consecutive days before the MR session, the rats were housed in a custom-made hypoxia chamber and submitted to change in inspired FiO 2 over a period of 1 minute (40 seconds at FiO2=5% followed by 20 seconds at FiO 2=21%) for 8 hours during daytime13 (see Figure 2B). This process has been shown to induce a global increase in Hct.13

  • Hemodilution group (n=5): isovolemic hemodilution was produced by the withdrawal (1 hour before MRI) of 5 mL of arterial blood (rate of 1 mL/min), which were concurrently replaced by the same volume of serum albumin (5%) injected at the same rate (see Figure 2A).

Figure 2.

Figure 2

Open in a new tab

Effect of hematocrit level on MR_StO2 (blood tissue oxygen saturation) estimates (n=19). Image depicting the protocols for the hemodilution (A) and intermittent hypoxia (B) experiments. (C) Hematocrit fraction (%) obtained in the three groups of rats (intermittent hypoxia, hemodilution, and control). (D) MR_StO2 (%) color maps overlaid onto the T2-weighted images of one representative rat from each group. (E) Blood gas analyses in the femoral vein (SvO2) and in the femoral artery (SaO2), calculated StO2 using SsO2 (oxygen saturation in the superior longitudinal sinus) and SaO2, and MR_StO2 in the gray matter. Data are presented as mean±s.d.

Blood samples were collected and analyzed prior and after the imaging session. MRI data were processed as described below using the Hct individually determined. Finally, the mean MR_StO2 was measured in a large ROI manually drawn over the cortex.

Stroke

Rats (n=9, 396±9 g; cf. ‘animal preparation') underwent transient focal brain ischemia induced by intraluminal occlusion of the right middle cerebral artery (MCAo).14 Briefly, the right carotid arterial tree was isolated. A monofilament (silicon rubber-coated monofilaments: 0.37 mm diameter, Doccol, Sharon, MA, USA) was advanced from the lumen of the external carotid artery into the internal carotid artery up to 5 mm after the external skull base. Five minutes before occlusion, rats were injected intravenously with pimonidazole (60 mg/kg; Chemicon International; Temecula, CA, USA) diluted in saline (60 mg/mL). After 30 minutes of occlusion, the MRI session began (cf. below) (see Figure 3A). Blood samples were collected and analyzed at the end of the imaging session. Rats were euthanized after 2 hours of MCAo. After killing the mice, their brain was quickly removed, frozen in −40°C isopentane and stored at −80°C until processing to determine the Pimonidazole necrotic–hypoxic fraction (see the section ‘Immunohistology' below). Finally, the lesion hemisphere was manually contoured in the MR anatomical T2-weighted (T2w) images. Within this area, the surface with MR_StO2 below 40% was measured.15 The ratio between the MR_StO2 surface below 40% and the hemisphere surface was denoted by the MRI necrotic–hypoxic fraction.

Figure 3.

Figure 3

Open in a new tab

Stroke experiment (n=9). (A) Image depicting the protocol for the stroke experiment. Transient focal brain ischemia was induced by intraluminal occlusion of the right middle cerebral artery (MCA). (B) Correlation between necrotic-hypoxic fractions estimated with magnetic resonance imaging (MRI) (fraction of the hemisphere with an MR_StO2 (blood tissue oxygen saturation)<40%) and histology (ratio between the positive pimonidazole-labeled surface plus the necrotic area and the hemisphere surface). (C) T2-weighted image, apparent diffusion coefficient (ADC, μm2/s) map, MR_StO2 map (%), and corresponding pimonidazole-stained slice in a rat bearing a stroke lesion. ADC and MR_StO2 color maps were overlaid onto the T2-weighted images.

Brain Trauma

Rats (350–450 g, n=5 per group; cf ‘animal preparation') were ventilated with additional oxygen to compensate for the atelectasis occurring during mechanical ventilation and not with spontaneous breathing. Ventilation was adjusted to maintain PaCO2 at 30 to 35 mm Hg. Traumatic brain injury (TBI) was induced according to the impact-acceleration model16 by dropping a 500-g mass through a vertical plexiglas guide from a height of 1.5 m onto the metallic disc glued onto the skull (Figure 4A). After the impact, the metallic disc was removed and the scalp sutured. MRI was performed 1 hour after the trauma in the TBI group and 1 hour after a reference time in a sham-operated group. The mean MR_StO2 and the mean apparent diffusion coefficient (ADC) were measured across four ROIs manually drawn in the left and right neocortex, and left and right caudoputamen cortex. One rat from each group was injected intravenously with pimonidazole before imaging and MRI, and the Pimonidazole necrotic-hypoxic fraction was eventually derived (see the section ‘Stroke'). Blood samples were collected and analyzed at the end of the imaging session.

Figure 4.

Figure 4

Open in a new tab

Traumatic brain injury (TBI) experiment (n=10). (A) TBI was obtained with the impact-acceleration model. A 500-g mass is dropped from a height of 1.5 m onto a metallic disc glued onto the skull. (B) MR_StO2 ((blood tissue oxygen saturation; mean±s.d.) across the cortex and the caudoputamen of the sham-operated and TBI groups. (C) T2-weighted image, apparent diffusion coefficient (ADC, μm2/s) map, MR_StO2 map (%), and corresponding pimonidazole-stained slice 1 hour after diffuse TBI in one rat, where no hypoxia was detected. ADC and MR_StO2 color maps were overlaid onto the T2-weighted images.

Tumor

F98 cells17 were implanted in the brain of male Fisher 344 rats (244±12 g, Charles River, n=10) as previously described.18 Tumor cell inoculation was performed under anesthesia with the following parameters: 5% isoflurane for induction and 2.5% for maintenance in 100% air. Five μl of cell suspension in serum-free RPMI1640 medium containing 5 × 103 cells was inoculated into the right caudate nucleus through a 1-mm burr hole 3.5 mm lateral to the bregma. After injection, the burr hole was filled, the skin incision sewed, and rats revived in an incubator before returning to the animal facility.18 Twenty-two days after tumor implantation, rats (cf. ‘animal preparation') underwent MRI (see Figure 5A). Oxygen (20%) was provided during tumor imaging to avoid the loss of potentially fragile rats due to the presence of a large tumor mass. Blood samples were collected after the imaging session and pimonidazole was injected in seven animals (see the section ‘Stroke'). Tumor and contralateral cortex ROIs were manually defined on the anatomic T2w images.

Figure 5.

Figure 5

Open in a new tab

Tumor experiment (n=7). (A) Image depicting the protocol for the tumor glioma experiment. (B) T2-weighted image (green line delineates the tumor region of interest (ROI)), apparent diffusion coefficient (ADC, μm2/s) map, cerebral blood volume (CBV, %) map, and MR_StO2 (blood tissue oxygen saturation; %) map in two rats bearing a F98 glioma implanted 22 days before the MRI experiment. ADC and MR_StO2 color maps were overlaid on the T2-weighted images. White arrows indicate tumor regions of low ADC, low CBV, and low MR_StO2 values in rat 1 and indicate regions of high ADC, high CBV, but normal MR_StO2 values in rat 2. (C) Correlation between necrotic–hypoxic fractions estimated with MRI (fraction of the hemisphere with an MR_StO2<40%) and histology (ratio between the positive pimonidazole-labeled surface plus the necrotic area and the hemisphere surface).

MRI Data Acquisition

MRI was conducted with a horizontal-bore 4.7-T Biospec animal imager (Bruker Biospin, Ettlingen, Germany) with an actively decoupled cross-coil setup (a body coil for radiofrequency transmission and a quadrature surface coil for signal reception). The MRI protocol was as follows:

  • Anatomical T2w images were acquired using a spin-echo MRI sequence (repetition time (TR)/echo time (TE)=4,000/33 milliseconds, 13 slices with a field of view (FOV)=30 × 30 mm2, matrix=256 × 256 and voxel size=117 × 117 × 1,000 μm3).

  • For tumor, stroke, and trauma rats, the ADC was mapped using a diffusion-weighted, spin-echo, single-shot echo-planar imaging (TR/TE=2,500/30 milliseconds, 8 averages, 5 slices with FOV=30 × 30 mm2, matrix=64 × 64 and voxel size 468 × 468 × 1,000 μm3). This sequence was applied four times, once without diffusion weighting and three times with diffusion weighting (b=800 seconds/mm2) in three orthogonal directions.

  • Multi-spin-echo 2D (MSME) (TR=1,500 milliseconds, 20 spin echoes, ΔTE=12 milliseconds, 5 slices with FOV=30 × 30 mm2, matrix=128 × 128 and voxel size 234 × 234 × 1,000 μm3).

  • Multi-gradient-echo 3D (MGE3D) (TR=400 milliseconds, 30 gradient echoes, ΔTE=2.5 milliseconds, 30 slices with FOV=30 × 30 mm2, matrix=256 × 256 and voxel size 117 × 117 × 200 μm3).

  • Injection of 200 μmol/kg of iron oxide particles (USPIO: P904, Guerbet S.A., Aulnay-sous-bois, France).

  • Repetition of MGE3D.

The entire MRI protocol lasted less than 1 hour in each rat.

MRI Data Processing

MRI data were processed in Matlab 8 (Mathworks) environment, using homemade software. ADC maps were computed as the mean of the ADCs observed in the three principal directions of the gradient system.

Cerebral blood volume (CBV) maps were estimated using the steady-state approach described by Tropres et al19 and computed using

graphic file with name jcbfm2014116e1.jpg

where γ=2.67502 × 108 rad/s per Tesla, B0=4.7 T and ΔχUSPIO, the susceptibility difference between blood in the presence and in the absence of USPIO, was set to 0.28 × 10−6 (CGS), a value previously obtained in similar experimental conditions.20, 21

MR_StO2 maps were estimated using the multiparametric qBOLD approach.8 A brief description of the protocol can be found in (Supplementary Figure 1a). A T2 map was derived from the multi-spin-echo 2D sequence using a non-linear fitting algorithm. To correct for the macroscopic inhomogeneities of B0, the MGE3D was spatially averaged as described in Christen et al.8 Thus, a MGE3D data set with a spatial resolution matching that of the T2 and of the CBV maps was obtained. The following equation was eventually fitted pixelwise to the MR signal decay s(t) collected beyond 10 milliseconds TE of the corrected MGE3D:8

graphic file with name jcbfm2014116e2.jpg

where Δχ0, the difference between the magnetic susceptibilities of fully oxygenated and fully deoxygenated hemoglobin, was set to 0.264 × 10−6 (CGS),22 and using T2 and CBV from the corresponding maps. Cte is the proportionality constant and Hct was set to 0.42 unless mentioned otherwise. A ratio of 0.85 was then employed to convert the large-vessel Hct to small-vessel Hct.23

Blood Gas Analysis

Blood samples (<0.1 mL) were analyzed using a blood gas analyzer (ABL 510, Radiometer, Copenhagen, Denmark). As this equipment is dedicated to human use, the measured pO2 was first converted to human scales using a rodent-to-human conversion factor of 0.694 before being translated into SaO2 (or SvO2, depending on the sample type) based on the oxygen dissociation curve after correcting for pH and temperature.24 From each blood sample, the partial pressure in O2 and in CO2, the hemoglobin content, and the pH were derived.

Determination of PbtO 2

To estimate PbtO2, we used the OxyLite system (Oxford Optronics, Abingdon, UK), an oxygen monitor based on fluorescence quenching. The optode (a 0.25-mm-diameter optical fiber at the tip of which a chromophore is fixed) was stereotaxically inserted at a depth of 6 mm into the striatum. After 1-hour stabilization, a transient increase in inspired oxygen fraction was performed to test the functioning of the PbtO2 probe.

Immunohistology

Coronal cryosections (12 to 20-μm-thick sections) were cut along the entire lesion for immunohistological analysis. Pimonidazole immunodetection was performed on the slice with the largest lesion area, using the Hypoxyprobe-1 kit (Chemicon International) according to the manufacturer's instructions. The primary antibody was mouse anti-pimonidazole diluted 1:10 and the secondary biotin-conjugated antibody was rabbit anti-mouse diluted 1:200 (B-8520, Sigma-Aldrich, St Louis, MO, USA). Extradivin-conjugated peroxidase was added 1:500 (E-2886, Sigma-Aldrich) and detected with the DAB Peroxidase Substrate Kit (SK-4100, Vector Laboratories, Burlingame, CA, USA). For brain tissue tumors, the adjacent slice was stained by hematoxylin/erythrosine. On the pimonidazole-stained slice with the largest lesion, the lesion hemisphere and the positive pimonidazole-labeled surface were manually contoured. The ratio between the positive pimonidazole-labeled surface and the hemisphere surface was denoted the Pimo necrotic–hypoxic fraction.

Statistical Analysis

An MRI estimate obtained from one rat corresponds to an average value obtained across a ROI. Data were compared across groups using unpaired Student t-tests (healthy rats) or Mann–Whitney tests (trauma rats). Comparison between data arising from the lesioned and contralateral sides was performed using the paired Student t-test. Parametric Pearson tests were used for the correlation analysis. Statistical significance was set to P=0.05. Statistical analysis was conducted with statistical software package (SPSS, Chicago, IL, USA). Data are expressed as mean±s.d.

Results

Numerical Simulations

The mathematical model plays a key role in the analysis and its assumptions can significantly impact the mqBOLD results. Following our previous work,9 we first tested the accuracy of the model in silico by simulating the MR protocol on simple virtual vascular networks. As the type of microvessel varies across the voxel (from large arterioles to large venules via capillaries) and thus so too does the oxygen saturation level, we found that the voxel T2 remained stable across microvessel types whereas the voxel T2* decreased (Supplementary Figure 1b), in line with the theory. We also found a strong linear correlation between the theoretical StO2 and MR_StO2 when simulations were performed at B0=3 T (corresponding to clinical scanners) (R2>0.99, P<0.001) and 4.7 T (the preclinical scanner used here) (R2>0.98, P<0.001) (Supplementary Figure 1c). Although basic geometry was used for blood vessel simulations, preliminary reports using data from high-resolution microscopy suggest that more realistic blood vessels will not significantly impact the results.9 We further explored the error on MR_StO2 estimates using numerical simulations and the following parameter ranges: T2 (10–200 milliseconds), T2* (10–200 milliseconds), CBV (1–15%), and MR_StO2 (20–100%). Signal to noise ratio was chosen to be 40 for the multi-spin-echo sequence and 55 for the multi-gradient-echo sequence, according to in vivo measurements. We obtained the following mean errors: T2=1.5%, T2*=1%, CBV=5.8%, and MR_StO2=23.8%. For a brain tissue characterized by CBV=4% and MR_StO2=70% (healthy) we obtained an error on MR_StO2 of ∼10% (in line with the observed variability on healthy rats (8%), see the results below). As MR_StO2 or CBV decreases, the error on MR_StO2 increases (e.g., for CBV=1.5% and MR_StO2=20%, the error on MR_StO2 becomes 40%).

Baseline MR_StO 2 Measurements

The mqBOLD approach was first tested in vivo in a cohort of 17 animals breathing air. One MR_StO2 map (obtained with mqBOLD), obtained in a representative rat, is shown in Figure 1A. The corresponding anatomic image (T2w), the T2* map, and the CBV map are also presented. MR_StO2 values were found to be homogeneous in the gray matter with an average of 76%±6%. The spatial resolution (234 × 234 μm2) allows the distinction of the corpus callosum, which exhibits lower MR_StO2 values. Fine detail and great contrast between gray and white matter can also be seen in the CBV map.

Effects of Inspired Oxygen Fraction Challenge on MR_StO 2

Three separate experiments (see Materials and methods) were conducted to investigate the relationship between measurements of MR_StO2, SvO2, oxygen saturation in the superior longitudinal sinus (SsO2), and/or oxygen pressure in brain tissue (PbtO2) in rats subjected to changes in the inspired fraction of O2 (FiO2) (Figure 1B).

In the first experiment, we found a good correlation between SsO2 and SvO2 (R2=0.957, P<0.001, n=12) (Supplementary Figure 2b). When SsO2 measurements could not be directly performed (e.g., rat in the MRI scanner or implanted with an optical probe), we computed an estimate of SsO2 from this calibration curve and the measurement of SvO2.

In the second experiment, we found a linear correlation between MR_StO2 in the striatum and blood gas estimates of StO2 (R2=0.83, P<0.001). The latter parameter was derived by the combination of SsO2 and SaO2 values using StO2=(1/3 × SaO2)+(2/3 × SsO2).25 The slope of the relation (1.05), close to unity, and the intercept (6.26), close to 0, suggest that MR_StO2 estimates reflect the tissue oxygen saturation level. As a comparison, MR estimates of StO2 in the cortex were also well correlated with the SsO2 values (R2=0.84, P<0.001) although the slope of the linear fit (1.26) indicated that MR_StO2 was greater than SsO2. These correlations were found to be weaker in the white matter (corpus callosum; R2=0.63, P<0.001, Supplementary Figure 2c). In this experiment, blood gases indicated that varying the FiO2 had a significant impact on the level of PaO2 level in each group (P<0.01, Figure 1C) but not on that of hemoglobin, pH, PaCO2, or mean arterial pressure (Figure 1C and Supplementary Figure 2a). The parametric maps obtained in four representative rats (Figure 1D) show global variations of MR_StO2 across the brain. The CBV was stable across PaO2 levels (Supplementary Figure 2d). Altogether, these results indicate that the MR_StO2 is greater than venous oxygen saturation and is strongly correlated to the tissue oxygen saturation across a large range of blood oxygenation saturation levels.

In the third experiment, a linear correlation was found between PbtO2 and SsO2 (R2=0.86, P<0.001, gray square in Supplementary Figure 2b), suggesting that MR_StO2 might reflect brain tissue pO2 in the healthy brain within the 5–60 mm Hg PbtO2 range.

Effect of Hematocrit Level on MR_StO 2

We found significant differences in hematocrit fraction between the three groups of animals, with Hct measured in the femoral vein ranging from 39%±3% in the control group to 31%±3% (P<0.001) and 51%±3% (P<0.001) in the hemodilution and intermittent hypoxia groups, respectively (Figure 2C). On the contrary, PaO2 and PaCO2 showed no significant change with Hct variations (Supplementary Figure 3a). By comparison to the control group, the local blood oxygen saturation estimated by MR in the cortex (MR_StO2) decreased in the hemodilution group and increased in the intermittent hypoxia group: 52%±5% and 74%±3%, respectively, versus 62%±4%, P<0.01) (Figures 2D and 2E). These significant differences were also found in blood gas estimates of StO2 (Figure 2E), while the SaO2 remained stable between groups. The lower StO2 value obtained in this control group (62%±4%) compared to that previously reported in the ‘Baseline MR_StO2 measurements' (76%±6%) may be ascribed to the lower depth of anesthesia to maintain spontaneous breathing in non-intubated rats.

MR_StO 2 Mapping During Stroke

During the MRI session (i.e., in the acute phase of occlusion), PaO2 was 129±49 mm Hg and PaCO2 was 37±13 mm Hg, consistent with the values found in healthy rats breathing air. We found regions of reduced ADC in the right (591±74 μm2/s) compared to the contralateral (879±54 μm2/s) hemisphere. Significant differences between MR_StO2 values were also found between the right and left hemispheres (48%±10% versus 76%±10%, respectively). The spatial extents of low MR_StO2 regions did not however necessarily match the low-ADC regions (Figure 3C), suggesting that MR oximetry could contribute to distinguish between the ischemic core (non salvageable) and the penumbra (where a proper oxygenation is maintained). Reduced CBV was observed in the lesion (manually delineated on the ADC maps) compared to the contralateral hemisphere (2.3%±0.9% and 4.3%±1.3%, respectively). In the injured hemisphere, a good correlation (R2=0.83; P<0.001) was found between the necrotic–hypoxic fraction defined with histology (pimonidazole staining) and that with MRI (MR_StO2<40%) (Figure 3B).

MR_StO 2 Mapping Following Traumatic Brain Injury

We found similar physiological parameters (PaO2, PaCO2) between the TBI group and the sham-operated group (Supplementary Figure 3b). Mean ADC (taken as a marker of the cellular edema) in the cortex and caudoputamen was, however, lower in the TBI group (790±70 μm2/second) compared with the sham-operated group (910±50 μm2/second). MR_StO2 was lower 1 hour after the insult in the TBI group (50%±9%) compared with the sham-operated group (73%±7%, P=0.03, Figure 4B), while CBV was similar in both groups (4.1%±1.0% and 3.9%±0.5%). Pimonidazole staining was negative in both the TBI and the sham-operated groups (Figure 4C), in agreement with the measured MR_StO2 (Figure 4C). The pimonidazole staining is described to highlight deep-tissue hypoxia (pO2<10 mm Hg),26 while in our experiment the MR_StO2 values in the trauma area stayed above the limit of deep hypoxia (>40%).15 These data are thus consistent with the presence of a mild brain hypoxia in the impact-acceleration model that is not deep enough to retain pimonidazole.27

MR_StO 2 Mapping in Glioma

During MRI, the mean tumor size was 104.9±28.0 mm3 (Figure 5A), PaO2 was 126±22 mm Hg, and PaCO2 was 31±6 mm Hg. In this glioma model, the mean CBV values measured in the tumor and in the contralateral cortex were similar (4.0%±0.5% versus 4.5%±0.6%, P=0.082, respectively), while the ADC was higher in the tumor than in the contralateral cortex (1,184±79 versus 780±28 μm2/s, respectively, P<0.001). MR_StO2 was lower in the tumor than in the contralateral cortex (62%±8% versus 79%±5%, respectively, P<0.01). However, average values in the ROIs did not reflect the dispersion of values inside the lesions. As an example, one can clearly distinguish in Figure 5B a core region with very low MR_StO2 values in rat 1, surrounded by regions of better-oxygenated tissues. At the edge of the tumor (as defined by anatomic T2w images), the MR_StO2 values are similar to the contralateral values. Other tumors exhibited homogeneous and normal MR_StO2 values, while CBV or ADC maps showed regions with higher values than in the contralateral cortex (see rat 2 in Figure 5B). The larger CBV in tumor could be ascribed to angiogenesis or to a leakage of the iron-oxide particles. Such leakage has however not been observed in previous experiments using other tumor models.20 Across all rats, the mean standard deviation within a ROI, an estimate of the heterogeneity of that region was ∼2.7 times larger in the tumor than in the contralateral area for MR_StO2 (this factor was 5.2 for ADC). In the lesioned hemisphere, the surface with MR_StO2 <40% was correlated with the surface labeled with pimonidazole, but the strength of the relation (R2=0.69; P<0.001) was weaker than the one found in the stroke model (Figure 5C).

Discussion

Several studies have recently been conducted to quantify the BOLD signal (see the review in Yablonskiy et al28). Based on the qBOLD approach,3, 5 the mqBOLD technique presented here includes two major modifications. First, mqBOLD combines several MR estimates obtained from sequences widely available on preclinical and clinical scanners. In this study, CBV was mapped using a steady-state approach and iron oxide particles, which yielded accurate maps with high spatial resolution. This approach has recently become available29, 30 for human experiments and may also be replaced by a dynamic susceptibility contrast approach using a bolus of routinely used gadolinium.31 Second, mqBOLD uses the total CBV instead of the deoxygenated blood volume.8 As such, mqBOLD aims to yield values equal to the weighted sum of arterial and venous oxygenation levels and above the venous oxygenation level.5 We evaluated the validity of this assumption using healthy rats submitted to various brain oxygenation challenges and found a strong correlation between MR_StO2 (from mqBOLD) and StO2 estimates from blood gas in the range of 20–90%. This technique could be further refined by taking into account the variations of magnetic susceptibilities across tissue types, as recently proposed in Yablonskiy et al.28 However, this would firstly require a procedure to assign a tissue type to each voxel and to handle partial volume effects. This refinement could be particularly useful in white matter, where the low MR_StO2 values observed with our approach might be due to the presence of myelin fibers that also shorten T2*.32 This refinement would prove particularly useful for human studies. More generally, this technique is sensitive to change in the tissue magnetic susceptibility and to change in hematocrit. MR_StO2 measurements might also be combined with (i) measurements of cerebral blood flow, available on MRI scanners, (ii) an estimate of the arterial oxygen fraction, available from a pulse oxymeter, and (iii) an estimate of the arterial/venous blood volume ratio, available in the literature,25 to estimate cerebral metabolic rate of oxygen (CMRO2), a physiologic estimate obtained with PET imaging.

Beyond the MR_StO2 map, would it then be possible to obtain a brain tissue pO2 map? The oxygen transport to the tissue is driven by the difference between blood pO2 and tissue pO2. However, the relation between blood pO2 and blood SO2, given by the hemoglobin dissociation curve,33 depends on several parameters such as pH, temperature and type of hemoglobin.34 While the relation between MR_StO2 and pO2 is not direct, it may nevertheless yield some information about the hypoxic status. In an acute stroke model, we obtained a strong correlation between the hypoxic areas determined from the MR_StO2 maps and from sections stained with pimonidazole, a gold standard marker of hypoxia that accumulates in tissues with a pO2 below ∼10 mm Hg.26 In the most hypoxic rats of our gas challenge study, PbtO2 was 6.2±1.6 mm Hg and SvO2 was 41.4%±4.4%, in good agreement with the oxygen saturation threshold for tissue hypoxia found elsewhere.15 Altogether, our results suggest that MR_StO2 can be a surrogate marker of mild and severe tissue hypoxia. In patients with stroke, this could be useful to better delineate the ischemic penumbra,35 as clinicians cannot currently clearly define the tissue at risk.36 In patients with severe TBI, it could be used in place of the current approaches such as near infrared spectroscopy, which cannot probe deep-brain structures, or PbtO2 probing, of which positioning remains controversial and it does not characterize the heterogeneity in brain oxygenation.37 Finally, our results in a glioma model suggest that the high spatial resolution accessible with mqBOLD can help assess the spatial heterogeneity of tumor tissue, which exhibits complex metabolic behavior.38 As mqBOLD is a non-invasive approach, it may be used to orient the therapeutic strategy and monitor the impact of antitumoral drugs on tumor oxygenation.39

In conclusion, MR_StO2 mapping appears as a robust and accurate imaging approach to map tissue oxygen saturation in the brain, whether in normoxic, mild, or severe hypoxic conditions. This in vivo imaging technique, which can be readily performed on existing MRI scanners, appears as a promising tool to improve our understanding of the pathophysiology of several brain diseases involving brain oxygenation impairments such as stroke, trauma, or cancer.

Acknowledgments

The authors acknowledge the excellent technical support of the MRI Facility of Grenoble (UMS IRMaGe), Pr. Christophe Ribuot from Inserm/UJF U1042 for providing animals exposed to intermittent hypoxia, and Guerbet SA for providing contrast agents.

The authors declare no conflict of interest.

Footnotes

Supplementary Information accompanies the paper on the Journal of Cerebral Blood Flow & Metabolism website (http://www.nature.com/jcbfm)

Financial support for this project was provided by ANR Imoxy program 2011-BSV5-004, Lyon Biopôle, ‘Fondation Gueules Cassées', and ‘fondation ARC pour la recherche sur le cancer'.

Supplementary Material

Supplementary Figure S1
Click here for additional data file. (4.4MB, tif)
Supplementary Figure S2
Click here for additional data file. (3.8MB, tif)
Supplementary Figure S3
Click here for additional data file. (1.2MB, tif)
Supplementary Figure Legends
Click here for additional data file. (26.5KB, doc)

References

  1. Ogawa S, Lee TM, Kay AR, Tank DW. Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc Natl Acad Sci USA. 1990;87:9868–9872. doi: 10.1073/pnas.87.24.9868. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Frackowiak RS, Lenzi GL, Jones T, Heather JD. Quantitative measurement of regional cerebral blood flow and oxygen metabolism in man using 15O and positron emission tomography: theory, procedure, and normal values. J Comput Assist Tomogr. 1980;4:727–736. doi: 10.1097/00004728-198012000-00001. [DOI] [PubMed] [Google Scholar]
  3. He X, Yablonskiy DA. Quantitative BOLD: mapping of human cerebral deoxygenated blood volume and oxygen extraction fraction: default state. Magn Reson Med Off J Soc Magn Reson Med Soc Magn Reson Med. 2007;57:115–126. doi: 10.1002/mrm.21108. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. An H, Lin W. Quantitative measurements of cerebral blood oxygen saturation using magnetic resonance imaging. J Cereb Blood Flow Metab Off J Int Soc Cereb Blood Flow Metab. 2000;20:1225–1236. doi: 10.1097/00004647-200008000-00008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. He X, Zhu M, Yablonskiy DA. Validation of oxygen extraction fraction measurement by qBOLD technique. Magn Reson Med Off J Soc Magn Reson Med Soc Magn Reson Med. 2008;60:882–888. doi: 10.1002/mrm.21719. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. An H, Liu Q, Chen Y, Lin W. Evaluation of MR-derived cerebral oxygen metabolic index in experimental hyperoxic hypercapnia, hypoxia, and ischemia. Stroke J Cereb Circ. 2009;40:2165–2172. doi: 10.1161/STROKEAHA.108.540864. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. An H, Lin W. Cerebral oxygen extraction fraction and cerebral venous blood volume measurements using MRI: effects of magnetic field variation. Magn Reson Med Off J Soc Magn Reson Med Soc Magn Reson Med. 2002;47:958–966. doi: 10.1002/mrm.10148. [DOI] [PubMed] [Google Scholar]
  8. Christen T, Lemasson B, Pannetier N, Farion R, Segebarth C, Rémy C, et al. Evaluation of a quantitative blood oxygenation level-dependent (qBOLD) approach to map local blood oxygen saturation. NMR Biomed. 2011;24:393–403. doi: 10.1002/nbm.1603. [DOI] [PubMed] [Google Scholar]
  9. Christen T, Zaharchuk G, Pannetier N, Serduc R, Joudiou N, Vial J-C, et al. Quantitative MR estimates of blood oxygenation based on T(2) *: A numerical study of the impact of model assumptions. Magn Reson Med Off J Soc Magn Reson Med Soc Magn Reson Med. 2012;67:1458–1468. doi: 10.1002/mrm.23094. [DOI] [PubMed] [Google Scholar]
  10. Marques JP, Bowtell R. Application of a Fourier-based method for rapid calculation of field inhomogeneity due to spatial variation of magnetic susceptibility. Concepts Magn Reson Part B Magn Reson Eng. 2005;25B:65–78. [Google Scholar]
  11. Klassen LM, Menon RS. NMR simulation analysis of statistical effects on quantifying cerebrovascular parameters. Biophys J. 2007;92:1014–1021. doi: 10.1529/biophysj.106.087965. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Bandettini PA, Wong EC. Effects of biophysical and physiologic parameters on brain activation-induced R2* changes: simulations using a deterministic diffusion model. Int J Imaging Syst Technol. 1995;6:133–152. [Google Scholar]
  13. Joyeux-Faure M, Stanke-Labesque F, Lefebvre B, Béguin P, Godin-Ribuot D, Ribuot C, et al. Chronic intermittent hypoxia increases infarction in the isolated rat heart. J Appl Physiol. 2005;98:1691–1696. doi: 10.1152/japplphysiol.01146.2004. [DOI] [PubMed] [Google Scholar]
  14. Longa EZ, Weinstein PR, Carlson S, Cummins R. Reversible middle cerebral artery occlusion without craniectomy in rats. Stroke J Cereb Circ. 1989;20:84–91. doi: 10.1161/01.str.20.1.84. [DOI] [PubMed] [Google Scholar]
  15. Kurth CD, Levy WJ, McCann J. Near-infrared spectroscopy cerebral oxygen saturation thresholds for hypoxia-ischemia in piglets. J Cereb Blood Flow Metab Off J Int Soc Cereb Blood Flow Metab. 2002;22:335–341. doi: 10.1097/00004647-200203000-00011. [DOI] [PubMed] [Google Scholar]
  16. Marmarou A, Foda MA, van den Brink W, Campbell J, Kita H, Demetriadou K. A new model of diffuse brain injury in rats. Part I: Pathophysiology and biomechanics. J Neurosurg. 1994;80:291–300. doi: 10.3171/jns.1994.80.2.0291. [DOI] [PubMed] [Google Scholar]
  17. Benda P, Someda K, Messer J, Sweet WH. Morphological and immunochemical studies of rat glial tumors and clonal strains propagated in culture. J Neurosurg. 1971;34:310–323. doi: 10.3171/jns.1971.34.3.0310. [DOI] [PubMed] [Google Scholar]
  18. Bräuer-Krisch E, Requardt H, Brochard T, Berruyer G, Renier M, Laissue JA, et al. New technology enables high precision multislit collimators for microbeam radiation therapy. Rev Sci Instrum. 2009;80:074301. doi: 10.1063/1.3170035. [DOI] [PubMed] [Google Scholar]
  19. Troprès I, Grimault S, Vaeth A, Grillon E, Julien C, Payen JF, et al. Vessel size imaging. Magn Reson Med Off J Soc Magn Reson Med Soc Magn Reson Med. 2001;45:397–408. doi: 10.1002/1522-2594(200103)45:3<397::aid-mrm1052>3.0.co;2-3. [DOI] [PubMed] [Google Scholar]
  20. Valable S, Lemasson B, Farion R, Beaumont M, Segebarth C, Remy C, et al. Assessment of blood volume, vessel size, and the expression of angiogenic factors in two rat glioma models: a longitudinal in vivo and ex vivo study. NMR Biomed. 2008;21:1043–1056. doi: 10.1002/nbm.1278. [DOI] [PubMed] [Google Scholar]
  21. Lemasson B, Christen T, Tizon X, Farion R, Fondraz N, Provent P, et al. Assessment of multiparametric MRI in a human glioma model to monitor cytotoxic and anti-angiogenic drug effects. NMR Biomed. 2011;24:473–482. doi: 10.1002/nbm.1611. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Spees WM, Yablonskiy DA, Oswood MC, Ackerman JJ. Water proton MR properties of human blood at 1.5 Tesla: magnetic susceptibility, T(1), T(2), T*(2), and non-Lorentzian signal behavior. Magn Reson Med Off J Soc Magn Reson Med Soc Magn Reson Med. 2001;45:533–542. doi: 10.1002/mrm.1072. [DOI] [PubMed] [Google Scholar]
  23. Grubb RL, Jr, Phelps ME, Ter-Pogossian MM. Regional cerebral blood volume in humans. X-ray fluorescence studies. Arch Neurol. 1973;28:38–44. doi: 10.1001/archneur.1973.00490190056006. [DOI] [PubMed] [Google Scholar]
  24. Hall FG. Minimal utilizable oxygen and the oxygen dissociation curve of blood of rodents. J Appl Physiol. 1966;21:375–378. doi: 10.1152/jappl.1966.21.2.375. [DOI] [PubMed] [Google Scholar]
  25. Watzman HM, Kurth CD, Montenegro LM, Rome J, Steven JM, Nicolson SC. Arterial and venous contributions to near-infrared cerebral oximetry. Anesthesiology. 2000;93:947–953. doi: 10.1097/00000542-200010000-00012. [DOI] [PubMed] [Google Scholar]
  26. Gross MW, Karbach U, Groebe K, Franko AJ, Mueller-Klieser W. Calibration of misonidazole labeling by simultaneous measurement of oxygen tension and labeling density in multicellular spheroids. Int J Cancer J Int Cancer. 1995;61:567–573. doi: 10.1002/ijc.2910610422. [DOI] [PubMed] [Google Scholar]
  27. Bouzat P, Millet A, Boue Y, Pernet-Gallay K, Trouve-Buisson T, Gaide-Chevronnay L, et al. Changes in brain tissue oxygenation after treatment of diffuse traumatic brain injury by erythropoietin. Crit Care Med. 2013;41:1316–1324. doi: 10.1097/CCM.0b013e31827ca64e. [DOI] [PubMed] [Google Scholar]
  28. Yablonskiy DA, Sukstanskii AL, He X. Blood oxygenation level-dependent (BOLD)-based techniques for the quantification of brain hemodynamic and metabolic properties - theoretical models and experimental approaches. NMR Biomed. 2013;26:963–986. doi: 10.1002/nbm.2839. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Christen T, Ni W, Qiu D, Schmiedeskamp H, Bammer R, Moseley M, et al. High-resolution cerebral blood volume imaging in humans using the blood pool contrast agent ferumoxytol. Magn Reson Med. 2013;70:705–710. doi: 10.1002/mrm.24500. [DOI] [PubMed] [Google Scholar]
  30. Varallyay CG, Nesbit E, Fu R, Gahramanov S, Moloney B, Earl E, et al. High-resolution steady-state cerebral blood volume maps in patients with central nervous system neoplasms using ferumoxytol, a superparamagnetic iron oxide nanoparticle. J Cereb Blood Flow Metab Off J Int Soc Cereb Blood Flow Metab. 2013;33:780–786. doi: 10.1038/jcbfm.2013.36. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Wintermark M, Sesay M, Barbier E, Borbély K, Dillon WP, Eastwood JD, et al. Comparative overview of brain perfusion imaging techniques. Stroke J Cereb Circ. 2005;36:e83–e99. doi: 10.1161/01.STR.0000177884.72657.8b. [DOI] [PubMed] [Google Scholar]
  32. Lee J, Shmueli K, Kang B-T, Yao B, Fukunaga M, van Gelderen P, et al. The contribution of myelin to magnetic susceptibility-weighted contrasts in high-field MRI of the brain. NeuroImage. 2012;59:3967–3975. doi: 10.1016/j.neuroimage.2011.10.076. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Barcroft J, Hill AV. The nature of oxyhaemoglobin, with a note on its molecular weight. J Physiol. 1910;39:411–428. doi: 10.1113/jphysiol.1910.sp001350. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Pittman RN.Regulation of Tissue Oxygenation Morgan & Claypool Life Sciences: San Rafael, CA, USA; 2011 . http://www.ncbi.nlm.nih.gov/books/NBK54104/ (accessed 5 September 2013). [PubMed] [Google Scholar]
  35. Jensen-Kondering U, Baron J-C. Oxygen imaging by MRI: can blood oxygen level-dependent imaging depict the ischemic penumbra. Stroke J Cereb Circ. 2012;43:2264–2269. doi: 10.1161/STROKEAHA.111.632455. [DOI] [PubMed] [Google Scholar]
  36. Heiss W-D. The concept of the penumbra: can it be translated to stroke management. Int J Stroke Off J Int Stroke Soc. 2010;5:290–295. doi: 10.1111/j.1747-4949.2010.00444.x. [DOI] [PubMed] [Google Scholar]
  37. Beynon C, Kiening KL, Orakcioglu B, Unterberg AW, Sakowitz OW. Brain tissue oxygen monitoring and hyperoxic treatment in patients with traumatic brain injury. J Neurotrauma. 2012;29:2109–2123. doi: 10.1089/neu.2012.2365. [DOI] [PubMed] [Google Scholar]
  38. Vander Heiden MG, Cantley LC, Thompson CB. Understanding the Warburg effect: the metabolic requirements of cell proliferation. Science. 2009;324:1029–1033. doi: 10.1126/science.1160809. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Emblem KE, Mouridsen K, Bjornerud A, Farrar CT, Jennings D, Borra RJH, et al. Vessel architectural imaging identifies cancer patient responders to anti-angiogenic therapy. Nat Med. 2013;19:1178–1183. doi: 10.1038/nm.3289. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Figure S1
Click here for additional data file. (4.4MB, tif)
Supplementary Figure S2
Click here for additional data file. (3.8MB, tif)
Supplementary Figure S3
Click here for additional data file. (1.2MB, tif)
Supplementary Figure Legends
Click here for additional data file. (26.5KB, doc)

Articles from Journal of Cerebral Blood Flow & Metabolism are provided here courtesy of SAGE Publications

RESOURCES