|Home | About | Journals | Submit | Contact Us | Français|
This article will outline the current utility of magnetic resonance imaging (MRI) in relation to animal models of perinatal brain injury. This will be presented in two parts. In the first part the technical aspects of small animal imaging are reviewed. In the second, we discuss the application of specific MR methods to animal models of perinatal hypoxic-ischemic injury. The most common animal model reviewed is that of hypoxic-ischemic injury as described by Rice et al. (Rice et al., 1981) consisting in the occlusion of a common carotid artery followed by a variable period of hypoxia.
Obtaining images of sufficiently high spatial resolution from small animals is both challenging and critical, especially in rodent pups. The brain of a 7-day-old rat pup has a volume ~0.45 ml, whereas the volume of a mouse brain at the same age has a volume of ~0.2 ml. These volumes are 1200 and 2700 times smaller, respectively, than a human term infant brain. Typical high resolution MRI of a human neonate has a spatial resolution of 1 × 1 × 1 mm3. In contrast, the average resolution for rodent imaging ranges from 0.078 × 0.156 × 1 mm3 to 0.234 × 0.234 × 1 mm3. Even this remarkably high resolution is an order of magnitude lower, relative to brain volume, than that achievable in human infants. The undesirable effects of low spatial resolution are related mainly to partial volume averaging. This effect consists of inclusion of various tissue types, such as white matter, grey matter, and/or CSF, in a single image element or voxel. This leads to blurring of the image, and in some cases, small structures may not be detectable at all. The primary determinant of achievable spatial resolution is signal-to-noise ratio (SNR), which scales directly with voxel volume. Thus, increasing spatial resolution from 2 × 2 × 2 mm3 to 1 × 1 × 1 mm3 is associated with an 8-fold reduction in SNR. Interestingly, image contrast can sometimes be better at higher spatial resolution, despite the reduction in SNR, due to the associated reduction in partial volume averaging effects with less image blurring.
There is a variety of approaches to increasing SNR. One method is to signal average by repeating the image acquisition several times in succession and averaging the signals. SNR increases as the square root of the number of averages, but this approach can significantly lengthen imaging time. In the example above, restoring SNR after increasing spatial resolution from 2 × 2 × 2 mm3 to 1 × 1 × 1 mm3 would require a 64-fold increase in imaging time. Another means of increasing SNR is the use of radiofrequency (RF) imaging coils that are better matched to the size of the subject and provide better “filling factor.” The use of MRI systems with higher field strength also provides an increase in SNR. Practically speaking, SNR scales roughly proportionally with magnetic field strength.
Animal motion is a common artifact, which can be very detrimental to image quality, particularly at long imaging times, which are required for the high resolution imaging in the small immature brain. It typically arises from respiratory motion, spontaneous movement or from vibration of the MR scanner caused by torque on the magnetic field gradient set during image acquisition. Anesthetizing the animal is usually not sufficient to completely remove animal motion, making it is necessary to employ special restraint devices during imaging. For adult animals, such devices typically include tooth and ear bars. Newborn animals with soft skulls, no teeth, and very small ear canals present a special challenge. We designed an MR compatible holder with lateral bars that can be tightened around the head with a distance equivalent to the measured head diameter and a small front platform for the palate with an adjustable height. Standard T2 weighted images can be fairly forgiving in the case of motion, which is not the case for diffusion imaging where the slightest motion will greatly impair the computation of relative anisotropy.
As noted above, higher field strength provides higher SNR, but at a cost. As field strength increases, so does the frequency at which RF coils must operate to obtain the images. RF inhomogeneity becomes a significant issue at higher frequency. In our experience using an 11.74-T scanner operating at 500 MHz, dielectric effects cause significant image signal inhomogeneity for samples larger than the size of an adult rat brain. While there are means of partially compensating for these effects, such as the use of multi-element coils, they add an additional level of complexity to image acquisition and analysis. Magnetic susceptibility effects are also more severe at high field strength. These effects are related to inhomogeneity in the local magnetic field arising at boundaries between structures of differing magnetic susceptibility, such as the air-tissue interface of cranial sinuses. This effect causes significant distortions in images obtained with fast, gradient echo imaging methods such as echo-planar imaging. On the other hand, gradient effects also form the basis for signal intensity changes associated with brain activation in functional MRI. Thus, contrast in functional MRI studies is better at higher magnetic field strength.
MR-compatible monitoring devices are available from a few companies (Small Animal Instrument Inc., Biopac Systems Inc., FISO Technologies). Such systems allow monitoring of respiratory rate, body temperature, heart rate, EKG and blood pressure. In addition to providing information on the health of the animal under study, these data can be used to gate MRI acquisition to respiration and/or heartbeat, which may further reduce the effects of animal motion. Also, feedback circuits can be employed in conjunction with body temperature measurements and heating devices to maintain animal body temperature within a set range. Maintenance of body temperature is particularly crucial in small rodent pups that have a high body surface-to-volume ratio and are prone to rapid loss of body heat with associated hypothermia.
The age of the animal used for the experiment is an important aspect of the model, as the developing brain goes through differing stages of tissue vulnerability – both in the extent and the pattern of tissue injury. A difference of just one day has been shown to have a profound impact on the severity of the injury as assed by spectroscopy and histology (Malisza et al., 1999).
In relation to the distribution of the tissue injury, hypoxia on pregnant Sprague-Dawley rats will cause a predominant white matter injury with white matter cell death, increase number of macrophagic and microglial cells resulting in a gliotic scar with delayed myelination. Early imaging of these pups showed a T2 signal increase and a T1 signal decrease of the white matter with an associated increased ADC (Baud et al., 2004). In contrast, the same insult in the same rat model at 7-day-old will produce a predominant severe cortical insult (Nedelcu et al., 1999), with MRI and especially diffusion alterations being easily detected.
The duration of hypoxia will also have a great impact on the size and localization of injury. Qiao et al. (Qiao et al., 2004) found that 45 minutes of hypoxia following unilateral carotid artery ligation in 7-day-old Wistar rats produced predominant white matter injury as compared to 90 minutes of hypoxia which produced a cortical and white matter injury.
An important aspect in the consideration of the model and the application of MRI is the equivalency of the animal age to human gestation. Most of the studies referred to in this review article are performed on rodent day of life 7, which is commonly accepted to be equivalent to the human term newborn, though there are publications offering a wide variation in the calculation of the equivalent age between human and rodent (Hagberg et al., 1997, Clancy et al., 2001, Clancy et al., 2007). Craig et al. have compared oligodendroglial maturation in both human and rodents. They showed that a P2 mouse or rat had an age equivalent to preterm infants at high risk of periventricular leukomalacia (Back et al., 2001, Craig et al., 2003). Hagberg et al. compared neurochemical and metabolic data, EEG pattern, synapse formation, cell growth and proliferation, patency of the blood-brain barrier. Taking all those elements into consideration, they concluded that a 7 to 14-day-old rat brain is equivalent to a human brain at term (Hagberg et al., 1997).
In conventional MR imaging, the signal detected arises from 1H in 1H2O. For T1-weighted images, tissue contrast based on the T1 relaxation properties of the 1H atoms. The T1 relaxation time constant for 1H2O in CSF, for example, is greater than that for 1H2O in white matter due to differences in chemical environment. As a result, CSF appears dark relative to white matter on T1-weighted images. It is important to remember that both the T1 and T2 relaxation time constants of tissue water are significantly different between newborn and adult animals. Further, they change rapidly during early development. Consequently, grey/white contrast for newborn brain is actually reversed from that of adult brain. In humans, grey/white contrast takes on the adult form during the first year of life. A second effect of rapid developmental changes in T1 and T2 relaxation time constants is that the optimum MR image acquisition parameters, those at which tissue contrast-to-noise ratio is greatest, varies during development. In newborn rodents, the optimum parameters vary every few days during the first week of life. Thus, they must be carefully determined for the animal of interest for the age at which the imaging will be done. This is usually achieved through measurement of tissue T1 and T2 relaxation time constants and calculation of the best acquisition parameters (Haacke, 1999).
As noted above, it is desirable to obtain small animal images at high magnetic field strength for improvements in SNR. However, T1 varies with magnetic field strength, and image contrast in T1-weighted images suffers at higher field strength as the differences in T1 between water in white and grey matter are smaller. One solution to regain T1 contrast at high field was published in 1993 by Ugurbil et al. (Ugurbil et al., 1993). They designed an MR image acquisition sequence called MDEFT (Modified Driven Equilibrium Fourier Transform) that provides a broader range of acquisition parameters over which optimized image contrast can be obtained (Fig. 1).
Manganese enhanced MRI (MEMRI) also can be employed to enhance T1 contrast. MgCl2 is typically administered intra-peritoneally, where it is absorbed into the vascular system and passes through the blood brain barrier to enter the brain. Contrast in the hippocampus, pituitary gland, olfactory bulb and cerebellum is enhanced. There are no hypoxic-ischemic studies in rat pups to date using MEMRI or MDEFT, but there is a very elegant description of enhancement of cerebellar cortex in mice pups following manganese administration. (Wadghiri et al., 2004) (Fig 2).
T2 relaxation time constants are less field dependent than T1, though there is a tendency for T2 values to be lower at higher field (de Graaf et al., 2006). Similarly to human newborns, immature rodent have an inverse grey-white matter contrast (Fig. 3). T2 relaxation time constants change quickly during maturation, with a normal contrast already present by three weeks of age (Fig. 3). It is therefore necessary to optimize image acquisition parameters to the model being used. Even with this optimization, it is often difficult to optimize acquisition parameters for both grey/white contrast and normal/injured tissue contrast (Fig. 4). As a result, the acquisition parameters used are often a compromise between optimization of these two contrasts.
An increase in the local tissue water T2 relaxation time constant, which causes areas of injury to appear bright on T2-weighted images, is often detectable in a matter of hours following ischemia in the immature rodent brain (Fig 5 and Fig 6) (Table 2) (Albensi et al., 1998, Aden et al., 2002). This contrasts greatly with changes observable on T2-weighted images from human neonates, which are usually evident only after 7 days following a perinatal ischemic insult (Rutherford et al., 2006).
It is important to appreciate that the lesion size as well as the T2 values change over the first 7 days following an ischemic insult in the immature brain (Aden et al., 2002). This is presented in Figure 6 where both the lesion size as well as the T2 values rise rapidly from 6–24 hours following an ischemic insult. Thus, delaying MR imaging until 24 hours after the injury may increase the visibility of lesions, particularly if the injury is more subtle or diffuse in nature.
The correlation of lesion size from MR T2 weighted imaging with histology has been studied. In newborn rodents, the area of T2 hyperintensity measured 3 and 6 days after injury has been shown to correlate with infarct size evaluated by histology (Albensi et al., 1998, Stieg et al., 1999). Aden et al. showed that T2 changes as early as 3 to 6 hours post injury correlate well with the histopathological score at 4 weeks (Aden et al., 2002). Antier et al. (Antier et al., 1999) used both TTC staining and T2-weighted imaging to evaluate injury size in 7-day-old rat pups subjected to unilateral carotid ligation and hypoxia. Their data showed good qualitative agreement between TTC staining at 72 hours and T2-weighted imaging at 5 weeks. Thus, lesion size on MR imaging early appears to be well defined in relation to histology.
In diffusion-based MR imaging, image contrast is based on the microscopic displacements of water molecules rather than the T1 or T2 relaxation properties of water in different chemical environments. The displacements measured are remarkably small, on the order of 10 µm, and are affected by cell membranes and constituents in a way that can be applied to learn about local tissue microstructure.
The basic measure of water displacements is the apparent diffusion coefficient, or ADC. ADC has units of mm2/s and varies during development. In general, ADC values decrease as the brain matures and water content decreases. ADC values also decrease within minutes of brain injury, providing a robust and widely used means for the early detection of brain injury. It is important to note that water ADC values decrease with both reversible and irreversible injury. For example, ADC decreases in response to administration NMDA in a mouse pup, but returns to normal when the excitotoxic action of NMDA is blocked pharmacologically (Dijkhuizen et al., 1996). Water ADC values are also anisotropic, meaning that they are not equivalent in all directions in space. In white matter, for example, water displacements parallel to myelinated fibers are greater than perpendicular to them. This is likely because water motion perpendicular to fibers is hindered by layers of myelin. To move in this direction, water molecules must pass through or around myelin layers. Similarly, water ADC values are anisotropic in immature cerebral cortex. In this case, displacements are greatest in a radial direction, reflecting the radial organization of the apical dendrites of pyramidal cells and radial glia. As the brain matures, anisotropy increases in white matter and decreases in cortex.
In practice, diffusion images are usually acquired using “diffusion tensor imaging” or DTI. With this approach, a series of diffusion measurements are made for each imaging slice. For each measurement, diffusion is evaluated along a different spatial orientation. These measurements are then combined to create a 3-dimensional representation of water displacements for each image element or voxel. This representation is described mathematically as a 3 × 3 matrix or tensor – hence the name. For areas of high anisotropy, the shapes tend to be cigar-like. In white matter, the long axis of the cigar is oriented parallel to the direction of the axons. In immature cerebral cortex, the long axis is oriented radially or perpendicular to the cortical surface. For areas in which diffusion is isotropic, such as in mature cortex or CSF, the shapes are spherical. For area of high anisotropy, it is common to show the long axis of the ellipsoid as a short line or “whisker” (Fig. 9).
The change in ADC values following brain injury is a dynamic process. For both newborn infants and rodent pups (Fig 7 and Fig 8), ADC values tend to decrease quickly following injury and then return to normal after a period of time. This return to normal is known as “pseudonormalization.” Following pseudonormalization, ADC values increase to higher than normal and stay elevated as cell debris is cleared away by macrophages. Thus, as with conventional MR imaging, the sensitivity of DTI to injury varies with time after injury. As shown on figure 8, pseudonormalization takes place more rapidly in animal models of perinatal ischemia than in human newborns (Fig. 7).
Application of diffusion methods to animal models of perinatal injury has been fairly limited. Tuor et al. (Tuor et al., 1998) reported a reduction of ADC values in neonatal rats following hypoxic/ischemic injury. In contrast to humans, ADC and T2 changes occurred in parallel (Fig. 6). Sizonenko et al. (Sizonenko et al., 2007) evaluated the effects of unilateral carotid ligation and exposure to hypoxia on the cerebral cortex of newborn rats. They showed a loss of cortical anisotropy in association with injury (Fig. 7). Drobyshevsky et al. (Drobyshevsky et al., 2007) in a model of rabbit uterine ischemia showed a correlation between low fractional anisotropy in corpus callosum, internal capsule and corona radiata in hypertonic kits which coincided in immunostaining with a loss of phosphorylated neurofilaments.
Spectroscopy offers information about metabolite levels, which in turn reflect the metabolic state of the brain. The greatest difference between spectroscopy and imaging lies in the relative concentrations of the molecules under study. The concentration of 1H in brain water is on the order of 100 M, whereas the concentrations of metabolites such as lactate are in the tens of mM range. As a result, a typical “voxel” for spectroscopy is much larger than that for imaging, often 2 cm on a side. Spectroscopy benefits markedly from high field strength with better SNR and improved spectral dispersion. Better spectral dispersion provides better separation of resonance peaks in MR spectra and allows for more accurate quantitation of resonance amplitudes, from which metabolite concentrations can be calculated. One potential challenge for spectroscopy at high field is that it can be difficult to eliminate magnetic field distortions related to magnetic susceptibility effects. There are automatic shimming techniques like FASTMAP(Gruetter, 1993), which help greatly in obtaining a homogeneous field. The line-width of the water peak is used to assess the quality of the shim and should be reported. It is also worth noting that a variety of nuclei are MR detectable in addition to 1H, including 31P, 23Na, 19F and 133Cs.
There are several metabolites that can be detected by proton spectroscopy. Changes in N-acetylaspartate (NAA) and lactate following a hypoxic-ischemic insult have been analyzed in detail. NAA is considered as a marker of neuronal integrity, whereas lactate is a marker of anaerobic metabolism (Malisza et al., 1999). Following hypoxic-ischemic injury, spectroscopy typically shows a decrease of NAA and an increase of lactate. Malisza et al. (Malisza et al., 1999) showed in the rat pup that increased lactate concentration can be detected during hypoxia-ischemia and remains high during 48 hours. NAA has a different profile. It initially decreases during hypoxia-ischemia, recovers in the following hours and decreases again at 24h and 48h. Vial et al. (Vial et al., 2004) in a piglet model found that lactate remains elevated over a longer period with lactate/NAA lactate/choline and lactate creatine ratio increased up to 7 days after injury.
Proton spectroscopy can also be used to measure brain temperature as the water resonance frequency is temperature dependent. Since the NAA resonance frequency is not temperature dependent, the chemical shift between the water and NAA resonances is usually used to measure temperature (Corbett et al., 1995).
31P spectroscopy measures phosphorylated metabolites. 31P spectroscopy in the rat pup and in the piglet has been uses to quantify acute changes of brain energy metabolism following hypoxic/ischemic injury. In a piglet model of hypoxia, high energy metabolite levels are monitored during hypoxia via 31P spectroscopy in an effort to maximize the reproducibility of the model. The duration of hypoxia is guided by nucleoside triphosphate and phsophocreatine levels (Thornton et al., 1997, Nedelcu et al., 1999, Vial et al., 2004). 31P spectroscopy during hypoxic/ischemic injury shows a rapid reduction of the ratio of phosphocreatine to inorganic phosphate that normalizes in a few hours and is followed by a more prolonged reduction of the ratio (Nedelcu et al., 1999) that correspond to secondary energy failure (Fig. 10).
Peeters-Scholte et al (Peeters-Scholte et al., 2003) applied both T2-weighted imaging and 31P spectroscopy to a newborn piglet model of hypoxia ischemia. They found significant improvement of MR parameters following allopurinol or deferoxamine, i.e. no decrease the phophocreatine/inorganic phosphate ratio and less T2 increase after injury (Table 3). There was no improvement in their histopathology evaluation with no difference in caspase-3 activity, TUNEL labeling or light microscopy. This suggests that MR methods may be more sensitive than histological approaches under some circumstances.
Cell tracking can now be undertaken with MRI by labeling the cells with ultrasmall superparamagnetic particles of iron oxide (USPIO) particles. USPIO causes distortions of the local magnetic field that cause signal dropout in MR images obtained with T2* weighting. This allows one to detect the cells. While there are no studies so far in neonatal models of hypoxic/ischemic injury, Wiart et al. have shown the distribution and evolution of inflammation of adult mice with stroke (Wiart et al., 2007), (Fig. 11). USPIO can also be incorporated in selected cell populations such as stem cells, thus broadening the utility of the method.
Functional MRI (fMRI) is possible in small animals although quite challenging. It relies on regional changes of cerebral blood flow occurring in concurrent changes of neuronal activity. Signal intensity varies in relation to the local concentration of paramagnetic deoxyhemoglobin. fMRI requires higher resolution for immature brains. Shen et al. (Shen et al., 2005) have shown loss of activation in somatosensory cortex after forepaw stimulation in a model of ischemic brain injury in adult rat at 4.7 T. Considerable work will be require to adapt those sequences to the developing rodent brain, as a spatial resolution 5 to 10 times higher than that used by Shen et al. would be required to identify cortical structures with precision. However, it is likely this technical hurdle will eventually be overcome.
It is also worth mentioning a form of functional MRI – functional connectivity or fcMRI – that has strong potential for application in animals and young children. For fcMRI studies, the subject does not perform a task during scanning, and data are collected with the subject at rest. The “noise” in fMRI data is of physiologic origin, and this aspect of the “resting” signal is analyzed. Correlations are sought between low frequency (< 0.1 Hz) signal fluctuations in the fMRI signal from a seed region (e.g., motor cortex) and other brain regions. For regions that are functionally connected, the signal fluctuations are temporally correlated (Biswal et al., 1995). When the signal in the seed area shows a positive fluctuation, the signal in the connected area does as well. Further, it has been shown that the fcMRI signal fluctuations in regions that undergo task-related deactivation are temporally anticorrelated with those of regions that undergo task-related activation (Fox et al., 2005). When the signal in the seed area shows a positive fluctuation, the signal in the connected area shows a negative fluctuation. In this fashion, fcMRI shows both correlations within networks and anticorrelations between networks. In addition, functional connectivity MRI effects persist under a variety of conditions. During the performance of low-level tasks, correlations corresponding to networks can be detected during both rest and stimulus epochs (Arfanakis et al., 2000; Greicius M.D., Menon V., 2004). The effects are also present, becoming more prominent, under midazolam sedation (Kiviniemi et al., 2005). To date, this method has been applied primarily in human studies, but will likely be adapted to animal studies in the near future.
Imaging of neonatal animal models following hypoxic-ischemic injury remains a challenge, but there are already an extensive number of studies in this domain. Recent access to very high magnetic field has improved image quality and the potential for defining the changes occurring after a hypoxic-ischemic insult. Moreover MRI is a fast developing field with new approaches being implemented that may also help in better defining injury and regeneration. Overall, diffusion studies (particularly ADC measurements) are effective for showing injury in the acute stage (within hours of injury). At later time points, conventional T1- and T2-weighted imaging shows the extent of injury and subsequent atrophy. Spectroscopy can be employed for evaluating changes in metabolic state, particularly energy status with 31P spectroscopy. Microstructural changes can be evaluated through measurement of diffusion anisotropy. In the future, changes in neural circuitry may also be approachable through fcMRI, and the presence of inflammatory or stem cells may be detected with T2*-weighted imaging. MR imaging also provides an important bridge between human clinical studies, for which developmental outcomes are often available, and animal studies, for which histology data are often available. It is likely that MRI will play a pivotal role in the evaluation of neuroprotective agents in both animal models and human trials.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.