|Home | About | Journals | Submit | Contact Us | Français|
Enormous progress has been made in assessing the neonatal brain, using magnetic resonance imaging (MRI). In this review, we will describe the use of MRI and proton magnetic resonance spectroscopy in detecting different patterns of brain injury in (full-term) human neonates following hypoxic–ischaemic brain injury and indicate the relevance of these findings in predicting neurodevelopmental outcome.
Fifty years ago, experiments have demonstrated that perinatal asphyxia could induce brain injury in primates . Since then, different patterns of brain injury have been established , which were dependent on the severity and duration of the hypoxic–ischaemic insult. These findings in animal studies were comparable to postmortem findings in asphyxiated human neonates.
During the 1960s to 1980s, imaging of brain injury was performed in human neonates using positron emission tomography scans and ultrasound. Since the introduction of magnetic resonance imaging (MRI), the knowledge of localisation and severity of brain injury following perinatal asphyxia (hypoxic–ischaemic brain lesions) in surviving neonates has expanded tremendously [3–5]. Diffusion-weighted MRI has enabled us to diagnose lesions much earlier than conventional MRI. MRI can reliably predict neurodevelopmental outcome and may serve as an early biomarker. In addition, phosphorus and proton MR spectroscopy (MRS) have enabled us to detect metabolic changes in the neonatal brain following hypoxia–ischaemia [6, 7]. In this review, we will describe the use of MRI and proton MRS in detecting different patterns of brain injury in (full-term) human neonates following hypoxic–ischaemic brain injury and indicate the relevance of these findings in predicting neurodevelopmental outcome.
Although cranial ultrasound can still play a role in the full-term infant with hypoxic–ischaemic brain injury , MRI is the method of choice to obtain more detailed and accurate information . The American Academy of Neurology recommended in 2002 that a computed tomography (CT) should be performed to detect haemorrhagic lesions in encephalopathic term infants and MRI only if findings are inconclusive . We do, however, recommend the use of MRI in all full-term infants who present with neonatal encephalopathy and/or seizures. The use of CT is restricted to infants who present with a large intracranial haemorrhage with a shift of the midline seen on cranial ultrasound. These infants may need neurosurgical intervention and access to a CT may be easier, and associated skull fractures may also be better visualised than with MRI [10, 11].
The increased use of neuroimaging techniques and MRI, in particular, has been a tremendous help in timing of brain injury and recognising patterns of injury [4, 12, 13]. Performing MRI within the first 2 weeks after birth, Cowan et al.  were able to show that more than 90% of affected newborns had evidence of perinatally acquired lesions on their MRI, with a very low rate of established antenatal brain injury. The presence of ventricular dilatation, widening of the subarachnoid space and interhemispheric fissure, presence of germinolytic cysts or cystic lesions in the white matter, seen at birth or during the first week, are all suggestive of an antenatal insult or an underlying problem, for instance a metabolic disorder . Performing cranial ultrasound on admission is also of value, as most of these lesions suggestive of an antenatal insult or an underlying problem will be recognised with ultrasound. The presence of increased echogenicity in the white matter on a day 1 ultrasound examination is also strongly suggestive of an antenatal insult as this echogenicity takes time to develop. These findings may be important for genetic counselling as well as for medicolegal issues (Fig. 1).
The use of diffusion-weighted imaging (DWI) has also greatly improved our ability to time the onset of brain lesions. A reduced apparent diffusion coefficient can be calculated, showing reduced values (restricted diffusion) during the first few days after the insult, with pseudonormalisation by the end of the first week [15–17]. Sequential imaging has shown that lesions in the basal ganglia may increase in size and site during the first week after birth [18, 19]. Barkovich et al.  performed sequential imaging in ten newborn infants and noted that patterns of injury varied considerably during the first 2 weeks after injury. The appearance of new areas of reduced diffusion simultaneous with the pseudonormalisation of areas that had reduced diffusion at earlier times could result in an entirely different pattern of injury on ADC maps acquired at different time points. One therefore needs to be aware of these evolving patterns.
Comparable to studies in a primate model , two main patterns of injury can be distinguished with MRI in the full-term neonate:
Recently developed techniques, such as diffusion tensor imaging (DTI), will allow quantification and visualisation of white matter pathways in vivo . DTI characterises the 3D spatial distribution of water diffusion in each MRI voxel. Water diffuses preferentially along the direction of the axons and is restricted perpendicular to axons by myelin. This directional dependency is referred to as anisotropy. Directionality encoded colour maps (red–green–blue) or fibre trackings are commonly used. A FA map can show asymmetry of the PLIC as early as the neonatal period. In a study of 15 patients with congenital hemiparesis due to different causes, studied at a median age of 2 years and compared with 17 age-matched controls, clinical severity of hemiparesis was noted to correlate with asymmetry in fractional anisotropy (p<0.0001), transverse diffusivity (p<0.0001) and mean diffusivity (p<0.03) . Another promising technique is volumetric determination of stroke volumes, which was noted to predict motor outcome in animal studies . Functional MRI tends to be used in childhood or adolescence to study reorganisation of the sensorimotor cortex [47, 48], but it was recently shown that passive unilateral sensorimotor stimulation is feasible even in the preterm infant resulting in bilateral activation of the sensorimotor cortex [49, 50].
The term PHS was recently coined by Armstrong-Wells et al.  and appeared to include term infants with parenchymal haemorrhage due to different underlying problems. They found a population prevalence of 6.2 in 100,000 live births. All infants presented with encephalopathy and more than half (65%) with seizures. Perinatal haemorrhagic stroke was typically unifocal (74%) and unilateral (83%). Etiologies included thrombocytopenia (n=4) and cavernous malformation (n=1); 15 (75%) were idiopathic. Foetal distress and postmaturity were found to be independent predictors. While some of the larger lesions will be recognised with ultrasound, MRI will provide more detailed information, and early DWI will be able to show associated areas with restricted diffusion.
Cerebral sinovenous thrombosis (CSVT) occurs in 0.41 per 100,000 liveborn infants . This diagnosis should especially be considered in infants who present without a history of perinatal asphyxia but with seizures and/or lethargy, sometimes in the context of infection or dehydration. Wu et al.  first pointed out that CSVT should always be considered in the presence of an intraventricular haemorrhage (IVH), especially when associated with a unilateral thalamic haemorrhage. Thirty-one percent of 29 infants born >36 weeks gestation, who were diagnosed with CSVT, presented with an IVH. Thalamic haemorrhage was diagnosed in 16% of infants with CSVT [53, 54].
Cranial ultrasound may detect CSVT, particularly in the presence of a midline thrombus in the superior sagittal sinus, or a unilateral thalamic haemorrhage. Power Doppler may be superior to colour Doppler when available [55, 56]. Additional imaging is required to exclude CSVT in more peripheral locations and to confirm the extent of the thrombus. Unenhanced CT may detect a thrombus and contrast-enhanced CT may show the ‘empty delta’ sign which is a filling defect in the posterior portion of the superior sagittal sinus due to thrombus. There are false positives and missed diagnoses in up to 40% of children with CSVT . MRI and MR venography or CT venography are needed to confirm the diagnosis . Susceptibility weighted imaging has recently been reported as another useful sequence in confirming the presence of CSVT and following for progression or resolution [59, 60].
The technique of MRS enables us to detect different molecules in tissue. Nuclei that have been used clinically for MRS are 31P and 1H. Thereby, in vivo brain metabolism can be assessed, and changes can be documented. MRS has been used to study changes in cerebral metabolism of neonates following (perinatal) hypoxia–ischaemia.
31P-MRS was one of the first MRS techniques to be used in neonates more than two decades ago [6, 7]. With this technique, metabolites such as high-energy phosphates (phosphocreatine (PCr), nucleotide triphosphates (NTP)) and inorganic phosphate (Pi), phosphomonoesters and phosphodiesters can be detected. It is time-consuming to measure absolute concentrations, and therefore, metabolite ratios have been calculated used instead to demonstrate changes in the brain. In addition, intracellular pH can be calculated using the formula published by Petroff et al. .
It has been shown that these metabolite ratios change during development, especially during the first years of life . In animal experiments hypoxia decreased PCr/Pi and NTP/total phosphate ratios . Full-term neonates with perinatal asphyxia have also been studied [64, 65]. After a successful resuscitation, brain energy metabolism returned to normal to become abnormal after 6–12 h to decrease even further after 24–48 h [65–67]. This coincided with clinical deterioration such as the development of seizures. The concept of ‘secondary energy failure’ has been elaborated in animal models, in particular in newborn piglets [68–70]. Using 31P-MRS in these animal models, neuroprotective strategies like hypothermia or 2-iminobiotin could be tested [71, 72].
An example of the changes in 31P-MRS during cerebral hypoxia–ischaemia in the newborn piglet is presented in (Fig. 7). With the increasing experience in human neonates, 31P-MRS was found to correlate with long-term neurodevelopmental outcome, especially when performed during the first few weeks after the insult [66, 67]. In neonates with a very poor outcome, elevated intracellular pH, the so-called pH paradox, existed even for weeks after the insult, possibly due to changes in the Na+/H+ transporter.
Unfortunately, with the magnetic field strength of the current clinical systems, it is impossible to perform localised 31P-MRS, so only relatively large brain areas can be examined. Due to these limitations, 31P-MRS has not become a routine clinical tool to assess asphyxiated full-term neonates.
Proton MRS has also been used in neonates since the early 1990s [62, 73]. Metabolites that can be demonstrated with 1H-MRS and echo times of 136 or more milliseconds are choline (Cho), creatine and phosphocreatine as a single peak (Cr), N-acetylaspartate (NAA) and—when present—lactate (Lac). With shorter echo times, myo-inositol and the combined glutamate–glutamine–GABA peak can be measured. Given the large amounts of water in the neonatal brain, high-quality water suppression is essential for 1H-MRS. The detection level of the aforementioned metabolites is in the millimolar range. As with 31P-MRS, providing absolute concentrations of metabolites is difficult, needing internal or external standards. Although some have used water as an internal standard, this may not be appropriate in asphyxiated full-term neonates with changing amounts of intracellular and extracellular brain water .
Therefore, metabolite ratios such as NAA/Cho, NAA/Cr or Lac/NAA are used instead . NAA is found mainly in neurons and oligodendroglial precursors. It has been used as a marker of neuronal integrity. With selective loss of neurons, the NAA/Cho ratio will decrease (Fig. 8).
Previously, the changes in 1H-MRS during normal development have been demonstrated . In a previous study, we have shown an increase in the NAA/Cho ratio between preterm at a gestational age of 32 weeks and term equivalent age . Others have shown that lactate is a normal component in the preterm brain . We have demonstrated lactate in the brain of the asphyxiated full-term neonate days after the insult . Others have shown that this may persist even for weeks . Experimental work in animals showed accumulation of lactate in the brain during secondary energy failure which could be decreased with appropriate neuroprotective strategies .
An advantage of 1H-MRS over 31P-MRS is the possibility to perform localised examinations. Volumes of brain tissue as small as 1 cc can be assessed in 1.5-T MR systems. Thereby, changes in brain areas that are particularly vulnerable, such as the basal ganglia and thalamus, can be demonstrated.
In addition, chemical shift imaging enables us to measure metabolites in a matrix of voxels overlying the brain. Thereby, localised elevations of lactate can be demonstrated . An example of this is given in Fig. 9, where 1H-MRS chemical shift imaging is presented of a neonate with a large infarct in the territory of the right middle cerebral artery.
One of the limitations of 1H-MRS is the loss of quality by susceptibility. Thereby, it is impossible to examine the cortex reliably.
We and others have shown decreases in NAA/Cho or NAA/Cr and elevations of Lac/NAA to predict a poor neurodevelopmental outcome [75, 80]. These studies have recently been summarised . The timing of 1H-MRS is less critical than of 31P-MRS. Measurements performed after the second week of life may show normalised PCr/Pi ratios. MR imaging performed before the fourth day after the insult may not yet demonstrate changes on T1- and T2-weighted images. Diffusion-weighted images may show pseudo-normalisation, which complicates assessment of the severity of the insult. Abnormalities using 1H-MRS may only be missed, when this technique is performed very early, i.e. during the first day of life, before the development of secondary energy failure .
Based on these aspects, it has been suggested that 1H-MRS is the best MR biomarker to predict neurodevelopmental outcome in asphyxiated full-term neonates . However, since brain metabolite ratios may vary between MR systems and coils, development of normal values in one’s own setting is required, and support of physicists is mandatory. The limiting factor in the development of these normal values may be the lack of ‘normal’ neonates who will undergo MR examinations.
Overall, there has been tremendous progress in the MRI technique over the last few decades. Both MRI and proton MRS can help in detecting different patterns of brain injury in (full-term) human neonates following hypoxic–ischaemic brain injury and are extremely useful in predicting neurodevelopmental outcome. We recommend to perform MRI and MRS in every full-term infant with encephalopathy and/or seizures admitted to a neonatal intensive care unit.
The authors would like to thank the technicians and radiologists of the department of radiology for their help.
Conflict of interest statement We declare that we have no conflict of interest.
Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.