BOLD MRI in sheep fetuses: a non-invasive method for measuring changes in tissue oxygenation




The purpose of this descriptive study was to correlate changes in the blood oxygen level-dependent (BOLD) magnetic resonance imaging (MRI) signal with direct measurements of fetal tissue oxygenation.


Seven anesthetized ewes carrying singleton fetuses at 125 days' gestation (term 145 days) underwent BOLD MRI, covering the entire fetus in a multislice approach. The fetuses were subjected to normoxic, hypoxic and hyperoxic conditions by changing the O2/N2O ratio in the maternal ventilated gas supply. The partial pressure of oxygen (pO2) in the fetal liver was measured using an oxygen-sensitive optode. Maternal arterial blood samples were simultaneously withdrawn for blood gas analysis. These measurements were compared with BOLD MRI signals in the fetal liver, kidney, spleen and brain.


We demonstrated a consistent increase in the BOLD MRI signal with increasing tissue pO2. For the fetal liver, spleen and kidney we observed a clear association between changes in maternal arterial blood pO2 and changes in BOLD MRI signal. Interestingly, we found that the BOLD signal of the fetal brain remained unchanged during hypoxic, normoxic and hyperoxic conditions.


This experimental study demonstrated that BOLD MRI is a reliable non-invasive method for measuring changes in tissue oxygenation in fetal sheep. The unchanged signal in the fetal brain during altered maternal oxygen conditions is probably explained by the brain-sparing mechanism. Copyright © 2009 ISUOG. Published by John Wiley & Sons, Ltd.


Monitoring fetal well-being, identifying fetuses at high risk of poor perinatal outcome and timing of delivery are the main challenges in obstetrics. Doppler ultrasonography and cardiotocography are the preferred diagnostic tools for this purpose, providing immediate information about the fetal heart rate and blood flow1. These methods, however, do not offer direct information about the oxygen supply to the fetus nor the oxygen levels in the fetal organs. The only current method for obtaining this direct information is cordocentesis, which is a procedure associated with a risk of miscarriage. In this experimental study we introduce a non-invasive method for the direct measurement of changes in fetal tissue oxygenation.

The blood oxygen level-dependent (BOLD) effect in magnetic resonance imaging (MRI) is based on the magnetic properties of hemoglobin. Oxyhemoglobin has diamagnetic properties whereas deoxyhemoglobin has paramagnetic properties2. Changes in the blood saturation level will therefore affect the apparent transversal relaxation time (T2*) of the surrounding water molecules, providing a measurable change in the MRI signal using a T2*-weighted sequence. The content of deoxyhemoglobin is inversely related to T2*3, and a signal reduction in T2*-weighted images therefore corresponds to an increased deoxyhemoglobin content. Thus, assuming that the pO2 of capillary blood is in equilibrium with the pO2 in tissue, changes in BOLD signal relate to changes in tissue oxygenation4.

BOLD MRI provides an opportunity to monitor tissue oxygenation in vivo. This technique has allowed investigation of the fetal response following maternal hyperoxygenation5 as well as audio and visual fetal stimuli6–8. In an experimental study in pregnant sheep, Wedegärtner et al. showed a reduced BOLD signal in fetal organs suffering from hypoxia9–11. Although these observations justify the use of BOLD MRI as a novel diagnostic tool to identify critical changes in the fetal oxygenation level, the exact tissue oxygen content cannot be established. Hence, the aim of this study was to investigate the relationship between changes in BOLD MRI signal and direct tissue oxygen measurements in the fetal liver during hyperoxic, normoxic and hypoxic conditions. In addition, BOLD MRI signals in fetal organs were compared to maternal arterial pO2 in response to changes in the ventilated oxygen supply.


Seven healthy ewes (Gotland sheep) carrying singleton fetuses at 125 days' gestation (term 145 days) were included in the study. Prior to the experimental procedures, the ewes were cared for at our animal farm (Påskehøjgård, Aarhus University Hospital, Denmark), where they had free access to water and food. On the day of the experiment the ewe was sedated with atropine (Nycomed, Taastrup, Denmark) 2 mg and Rompun (Bayer Schering Pharma, Berlin, Germany) 2.5 mg and transported to the surgical unit. Anesthesia was induced with Ketamine (Pfizer, New York, NY, USA) 2 mg/kg and Dormicum (Roche, Basel, Switzerland) 0.25 mg/kg and maintained by Isoflurane (Baxter, Deerfield, IL, USA) 1.5% in an O2/N2O mixture. Orotracheal intubation (tube diameter 9 mm) was followed by artificial ventilation using a servoventilator (Abott, Solna, Sweden; tidal volume 600–800 mL, 18 breaths/min). Catheterization of the left common carotid artery was performed for withdrawal and blood gas analysis (Radiometer, Brønshøj, Denmark), a pulse oximeter sensor was attached to an ear, and a capnograph was used for end-tidal CO2 measurement.

By changing the O2/N2O ratio in ventilated gas, the fetus was subjected to normoxic, hypoxic and hyperoxic conditions. The oxygen content varied between 10% and 100%. Each ventilation level was maintained for 30 min to allow steady-state oxygenation conditions. In four ewes, the tissue oxygenation in the fetal liver was successfully measured using an oxygen sensitive optode. The experimental protocol was approved by the local animal ethics committee.

The invasive tissue oxygenation measurements were performed with an MRI-compatible optical oxygen-sensitive optode (Presens, Regensburg, Germany). The optodes used in this study consisted of a 15-m long, 0.9-mm thick fiber-optic cable, exposed and coated with oxygen-sensitive fluorescent sensor material at the tip, which was 0.1 mm in diameter. The principle of measurement is based on the effect of dynamic luminescence quenching of the fluorescent sensor material by molecular oxygen. The dynamic quenching is directly related to the local tissue pO2 and is measured in real time. The 90% response time of the optode was approximately 10 s12. Compared to conventional oxygen-sensitive (Clark-type) electrodes, optodes have the advantage of not consuming oxygen during measurements, such that long-term pO2 measurements in vivo can be performed in tissue, without depleting oxygen around the optode tip, even in regions where oxygen supply is mainly limited to molecular diffusion. One previous study has been published applying this fluorescent technique to fetal tissue oxygen measurement in vivo13. In this study a 1.2-mm diameter trocar was placed in the fetal liver using sonographic guidance and the optode was firmly inserted through it.

MRI was performed with a 1.5 T clinical system (Philips Medical Systems, Best, The Netherlands). The anesthetized ewe was examined in a left lateral position. The acquisitions were obtained with a phased-array abdominal radio-frequency coil. Fast axial and coronal scans of the fetus were initially performed to locate organs of interest. Next, a sequence was employed, sensitive to changes in magnetic susceptibility using strong T2*-weighting, allowing one to assess changes in the BOLD contrast. This was performed as a dynamic sequence, in which T2*-weighting was accomplished using a two-dimensional gradient-echo sequence with sequential echo-planar encoding. The sequence parameters included: eight slices with 3 mm of thickness; field of view 250 × 250 mm2; acquired spatial resolution 256 × 256 voxels; repetition time (TR) 1500 ms; flip angle 90°; and an echo time of 40 ms optimized for the average T2* value in vivo. Six transients were acquired for optimal signal-to-noise ratio, giving a single-frame scan time of 300 s. A total of 150 dynamic frames were recorded with an interval of 112 s.

The MRI data were transferred to an external workstation equipped with the Mistar software (Apollo Imaging Technology, Melbourne, Australia). All images displayed an adequate signal-to-noise ratio and did not show substantial artifacts. The signal intensity of the T2*-weighted images was examined in regions of interest (ROIs) as shown in Figure 1. Placement of ROIs in the fetal kidney, liver, spleen and brain was performed by a clinical radiologist, and movements were adjusted manually in each dynamic image. For each ROI, the mean ( ± SD) signal versus time curve was calculated.

Figure 1.

Blood oxygen level-dependent magnetic resonance image showing an axial view of the fetus with the fetal liver selected as the region of interest (ROI).

From a theoretical point of view the local magnetic susceptibility is the degree of magnetization of a material in response to a magnetic field, but the inherent differences in local magnetic field mean that it is not possible to directly compare the acquired BOLD signal between animals. Furthermore inter-animal variations are expected in hydration, fetal size, temperature, distance to the receiver coil etc. Owing to these expected inter-animal variations and the modest size of the dataset, we have taken a descriptive approach to data analysis.

Optode measurements of fetal pO2 were compared with levels of BOLD MRI of the fetal liver during different maternal ventilation oxygenation levels, and the BOLD MRI signals of all fetal organs were compared with maternal arterial pO2 levels and graphically presented.


All the animals underwent BOLD MRI during normoxic, hyperoxic and hypoxic conditions. In four out of seven animals an oxygen-sensitive optode was successfully inserted into the fetal liver for direct measurement of tissue oxygen. For the fetal liver, changes in the BOLD signal were associated with changes in the direct optode measurements of tissue pO2 (Figures 2 and 3). Within a few min after changing the maternal oxygen supply (performed by changing the ventilated O2/N2O ratio), we observed a corresponding response in the optode measurements as well as in the acquired BOLD signal of the fetal liver. Because of progressive lung atelectasis, the maternal arterial pO2 did not return to the starting level when the maternal ventilation was returned to normoxic conditions (O2/N2O ratio = 1.0) at the end of the experiment, and consequently the BOLD signal of the fetal liver and the optode measurements also remained at a lower level in the last period of the experiment.

Figure 2.

Changes in fetal liver blood oxygen level-dependent (BOLD) magnetic resonance signal (a), fetal liver optode oxygen measurement (b) and maternal arterial pO2 (c) over time for one sheep (Animal 8) during varying oxygen levels (14–100%) in maternal ventilated gas.

Figure 3.

Association between the blood oxygen level-dependent (BOLD) magnetic resonance imaging (MRI) signal of the fetal liver and the direct optode measurements from the fetal liver in four animals. ⧫, Animal 2; ▪, Animal 3; ▴, Animal 5; ×, Animal 8.

Figure 4 demonstrates a linear relationship between maternal arterial pO2 and optode measurements of tissue oxygen levels, justifying the use of maternal arterial pO2 as a reference for organs without optode measurements. Figure 5 illustrates the BOLD signal in one of the seven fetuses during periods of different maternal arterial pO2 level. The BOLD signal of the fetal kidney and the spleen changed in patterns similar to the one seen in the liver. In contrast to this, the signal of the brain remained constant during normoxic, hypoxic and hyperoxic conditions.

Figure 4.

Association between optode measurements of pO2 in the fetal liver and pO2 measured in maternal arterial blood samples in four animals. ⧫, Animal 2; ▪, Animal 3; ▴, Animal 5; ×, Animal 8.

Figure 5.

Blood oxygen level-dependent (BOLD) magnetic resonance signals from the fetal brain (

Thumbnail image of

), spleen (

Thumbnail image of

), liver (

Thumbnail image of

) and kidneys (equation image) in one animal (Animal 5) during varying oxygen levels in maternal ventilated gas (10–100%).

In all the animals we observed that changes in the BOLD signal of the fetal spleen, kidney and liver were closely associated with changes in maternal arterial oxygenation (Figure 6). As shown in Figures 3 and 6, the BOLD MRI data from each organ in relation to optode/maternal arterial measurements seem to describe a curved graph, which is steeper at lower oxygen levels and flattens at higher oxygen levels. The only exception to this pattern is the BOLD response from the brain, the BOLD signal of which remained constant.

Figure 6.

Association between blood oxygen level-dependent (BOLD) magnetic resonance signals from the fetal brain (a), spleen (b), liver (c) and kidney (d) and maternal arterial pO2. ▪, Animal 1; □, Animal 2; ▴, Animal 3; ▵, Animal 5; ●, Animal 6; ○, Animal 7; ×, Animal 8.

The BOLD MRI signal response was expected to vary between the animals. However, as shown in Figures 3 and 6 there was only a small amount of variation between the different animals, thus the data seem to be reproducible.


This experimental study in pregnant sheep demonstrated that changes in the BOLD MRI signal of the fetal liver were closely related to changes in tissue oxygenation determined by an oxygen-sensitive fluorescence optode. We also observed that changes in BOLD MRI signal were reproducible among animals. Thus, BOLD MRI seems to constitute a reliable non-invasive technique for the determination of short-term changes in fetal tissue oxygenation.

The sensitivity of BOLD MRI is enhanced by the relative hypoxic conditions of the normal fetus. Despite the left shift of the dissociation curve of fetal hemoglobin, as compared to adult hemoglobin, the hemoglobin saturation in the umbilical vein is only 80%14. The fetal arterial, mid-capillary and venous oxygen concentrations correspond with the steeper part of the hemoglobin dissociation curve. Therefore, with any given change in oxygen concentration, a more pronounced change in the BOLD signal is expected in the fetus as compared to the adult. Keeping the sigmoidal shape of the dissociation curve in mind, one should expect a more pronounced change in BOLD signal during lower oxygen levels and the opposite effect for higher oxygen levels. Our work supports this theory, as the BOLD data from each organ in relation to optode/maternal arterial measurements seem to describe a curved graph that is steep at lower oxygen levels and flattens at higher levels. This is in conflict with a study by Wedegärtner et al.9 that suggested a linear association between maternal arterial saturation and the BOLD signal of selected fetal organs. This difference is probably explained by the different experimental designs of the two studies, as we, unlike Wedegärtner et al., also performed experimental measurements during hyperoxic conditions.

Another variable of importance is the ‘fractional blood volume’. As the BOLD MRI signal reflects the deoxyhemoglobin concentration, it depends not only on hemoglobin saturation but also on the fractional blood volume, i.e. if the vessels dilate the amounts of both oxyhemoglobin and deoxyhemoglobin increase within the voxel even in case of unchanged hemoglobin saturation15. The oxygen-sensitive optodes used in this study, however, operate independently of changes in fractional blood volume. This indicates that the changes in BOLD MRI signal truly reflect oxygenation changes.

As we only succeeded in inserting oxygen-sensitive optodes into the fetal liver of four animals, we used the maternal arterial pO2 as a reference for the remaining fetal organs. This was justified by the observed linear relationship between the optode measurements and the maternal arterial pO2. In the fetal liver, spleen and kidney we found a clear association between changes in the BOLD signal and changes in maternal arterial pO2. In contrast to this, the BOLD signal of the fetal brain remained unchanged during the normoxic, hypoxic and hyperoxic conditions—a consistent finding in all our experiments. We explain this interesting phenomenon as a result of the brain-sparing mechanism, characterized by increased blood flow to the brain during hypoxic conditions, thus compensating for the reduced oxygen content of the circulating blood. This autoregulation is very efficient, and protects the brain from extreme conditions. Our results indicate that the brain was also spared from hyperoxic conditions.

Apart from the fetal brain, fetal oxygenation increased during hyperoxic maternal conditions. This consistent finding is supported by other studies. Semple et al.5 performed BOLD MRI during hyperoxia in normal pregnancies, showing an increase in BOLD signal in five out of nine pregnancies. The lack of response in four pregnancies was probably due to fetal movements interfering with the MRI signal5. In growth-restricted fetuses maternal hyperoxygenation seems to improve umbilical pO2 as determined by cordocentesis16, normalize brain sparing demonstrated by Doppler flow17 and increase fetal movements18. The ‘oxygen test’ was introduced by Arduini et al.19; if fetal brain sparing is normalized by maternal hyperoxygenation the test is positive, and if brain sparing remains unaffected the oxygen test is negative. A negative test has been reported to be a strong predictor of adverse perinatal outcome19, 20. A positive oxygen test cannot be explained by increased oxygenation of maternal hemoglobin as the arterial saturation is close to 100% even during normoxic conditions. Therefore, an increased oxygen partial pressure gradient across the placental membrane is likely to be of importance, as the hemoglobin saturation of venous umbilical blood is only 80%14.

Differences between the human placenta and the sheep placenta complicate extrapolation of our experimental findings to the human situation. The histological structure of the human placenta can be represented by the multivillous model, in contrast to the sheep placenta, which is represented by the venous equilibration exchange model21. The sheep placenta is less effective at transporting oxygen to the fetus than is the human placenta. As a consequence of the reduced O2 extraction coefficient the sheep fetus is slightly more hypoxic (umbilical vein, 4.7 kPa; umbilical artery, 2.7 kPa) compared to the human fetus (umbilical vein, 5.5 kPa; umbilical artery, 3.9 kPa). The neuroendocrine and cardiovascular responses in hypoxia are very similar in sheep and human fetuses, including the brain-sparing mechanism22. The long duration of anesthesia might have affected our results, but not the conclusions drawn from them.

In conclusion, BOLD MRI is a reliable non-invasive method for measuring regional changes in tissue oxygenation in fetal organs in vivo. This method provides new opportunities for fetal monitoring and fetal testing.


This work was supported by Forskningsinitiativet, Aarhus University Hospital. Thanks to Peter Årslev Nielsen for support and advice in anesthesiological matters.