Introduction

       Perinatal asphyxia remains a major cause of pediatric mortality and morbidity, with possible neurologic sequels such as cerebral palsy, mental retardation, or epilepsy ^[1]. Hypoxic-ischemic encephalopathy (HIE) secondary to perinatal asphyxia occurs in 1 to 2 per 1000 live births ^[2]. Early and accurate diagnosis and prediction of outcome in neonates with encephalopathy is important for early clinical management and counseling of the parents. Moreover, early, accurate and quantitative assessment might serve as a decision point in determining if neuroprotection by using hypothermia is indicated. Therefore, it is critical to understand the evolution of early changes in the brain after the injury.

       MR has an increasing role thanks to its great sensitivity and specificity. Still, one issue with conventional T1 and T2-weighed MR imaging is subjectivity, meaning that the diagnostic accuracy relies on the ability of the reader to detect subtle abnormalities. Advanced MR techniques provide quantitative MR biomarkers such as apparent diffusion coefficient (ADC) on diffusion-weighted imaging (DWI), fractional anisotropy (FA) on diffusion tensor imaging (DTI) and neurochemical alterations on MR spectroscopy (MRS). Most of the centers perform neonatal clinical MR imaging at 0.5-1.5T. During the recent years, an increased utilization of 3.0T magnet in clinical settings has improved the image and diagnostic quality. Because the signal-to-noise ratio (SNR) is improved, the total data acquisition time is reduced. Furthermore, for MRS, higher magnetic field allows a larger separated spectroscopy of the metabolites. The radiofrequency power is also increased, whilst the specific absorption rates (SAR) at 3T are not prohibitive for neonatal scanning after adjusting for radiological safety guidelines^[3].

       The aim of this review is to present in a systematic order MR findings at 3T in neonatal HIE using conventional imaging and advanced MR techniques, such as DWI, DTI and MRS. The time course of lesions for each technique is emphasized, and correlated with the neurological outcome.

Practical issues for neonates at 3T MRI

       A successful acquisition relies on an immobile baby, so it requires a careful preparation. Neonates might be examined during natural sleep or under light sedation. Fast sequences such as DWI and DTI bring acoustic noise, and it is more intense at 3T, so that ear protection should be used. In our institution, MR acquisitions are performed early in the morning, when the neonates are spontaneously asleep, and they are positioned in a vacuum immobilization pillow to minimize body and head movements ^[4]. Ear-muffs are placed to minimize reactions to noise exposure. Oxygen saturation and electrocardiography are monitored throughout the acquisition. The sequences generating more noise are realized at the end of the exam and the total scanning time is reduced as short as possible in order to limit the motion artifact.

       Neonates with severe encephalopathy may be in coma or ventilator-dependent in the first few days, therefore MR-compatible ventilator equipment may be necessary during the MR acquisition. When not available, a hand-bag should be used during the short examination procedure.

3T neonatal MR imaging

       Imaging at 3T provides more details of the developing brain. The increased SNR allows high-resolution images within acceptable acquisition times. At 3T, an increased conspicuity of vessels is noticeable on T1 weighted images, so that an inversion recovery (IR) sequence could be a good alternative for T1 weighted images to avoid vessel contamination ^[3]; moreover it offers an extremely high contrast between gray and white matter and between myelinated and unmyelinated white matter.

       Diffusion imaging at 3T benefits from a higher SNR and less geometric distortion artifacts, when these acquisitions are combined to high accelerative parallel imaging techniques such as SENSE. Due to the improving SNR, the examination time is shortened but also the partial volume effects are reduced, which is beneficial for fibre tracking.

       For MRS, greater accuracy in quantitative measurements and minimized partial-volume effects are observed at 3T by reducing the voxel size. Increased chemical shift separation allows a clearer separation of peaks that are too close to each other at 1.5T, for example, glutamate, glutamine and gamma-aminobutyric acid (GABA) ^[5].

Conventional MR imaging in neonatal HIE

       For more than a decade, T1- and T2-weighted imaging has been considered an essential modality for imaging the neonatal brain ^[6,7,8]. The basic information for the diagnosis is provided by T1- and T2-weighted images, depicting typical MRI features of hypoxic-ischemic damage. The imaging pattern of HIE has been classified into three types ^[7,8,9]: parasagittal lesions involving vascular boundary zones, profound lesions located in the basal ganglia or thalamus and multicystic encephalomalacia. Parasagittal lesions are usually associated with milder HIE, and typically involve the parasagittal cortico-subcortical regions. In the acute phase, the affected cortex shows areas with high T2 signal intensity and low T1 signal intensity, due to the presence of edema. As the lesion evolves, cortical thinning becomes evident. Profound lesions in the basal ganglia and thalamus are often associated with moderate or severe HIE. Increased signal intensity on T1 can be detected 3-7 days after the insult, and decreased T2 signal intensity can be seen a little later after 6-10 days ^[8]. A complete loss or change in the normal signal intensity of the posterior limb of internal capsules (PLIC) may be seen in case of perinatal asphyxia. The posterior part of PLIC have already myelinated in full-term neonates, therefore the signal from myelin may be diminished due to an anoxic injury ^[10]. Multicystic encephalomalacia occurs, when the anoxic insult is particularly severe and a diffuse damage ensues in the whole brain. Diffuse T1 hypointensity and T2 hyperintensity can be detected in the first 2-3 days due to the brain swelling, followed by marked T1 hyperintensity and T2 hypointensity in the basal ganglia and thalami. The cavitations are formed rapidly and calcification can be detected on CT in 3-4 weeks.

       The changes seen on routine T1- and T2-weighted images are characteristic, but are often subtle and may not be obvious until several days after the injury ^[11,12,13]. At the acute stage, brain swelling on T1- or T2-weighted images may be associated with some loss of gray/white matter differentiation, and with the evolution of the diseases, abnormal signals could be apparent in the following weeks ^[13]. Because of the unmyelinated white matter, potential signal abnormalities on T2-weighted images can be masked by the high T2 signal of white matter. During the first weeks of life, T1 shortening (hypersignal on T1-weighted images) lesions can be seen before becoming obvious on T2-weighted images. These signal changes might be due to lipid breakdown products of myelin or mineralization ^[8].

       Recently a meta-analysis ^[14] was reported on the prognostic utilities of various quantitative cerebral MR biomarkers in neonatal encephalopathy. Thirty-two studies were included between January 1990 and July 2008. For predicting adverse outcome, conventional MR during the first 30 days had a pooled sensitivity of 91% (95% confidence interval [CI]: 87%-94%) and specificity of 51% (95% CI: 45%-58%). The pattern of injury identified with conventional MRI may provide diagnostic and prognostic information for the neonatal encephalopathy ^[9,15]. The infants who show abnormal signal intensity in the basal ganglia/ thalami or the PLIC usually undergo a poor neurodevelopmental outcome ^{[10, 15]}.

DWI in neonatal HIE

       Diffusion imaging is based on the random thermal movement of molecules (i.e. Brownian motion). The technique uses fast (echo-planar) imaging methods which make it less sensitive to patient motion, with an acquisition time limited to less than one minute^[16]. Quantification of this motion can be measured by ADC, which presents the averaged magnitude of water diffusion. Hypoxia or ischemia may damage the brain by causing focal lesions or diffuse injury. One study indicated that a hypersignal in DWI with a decrease of ADC was detected in focal lesions; this might be due to the restriction of extracellular water motion by cell swelling resulting in cytotoxic edema ^[17].

       DWI has proven to be very sensitive in early detection of acute cerebral ischemia, and may detect the lesions in HIE when for conventional MRI it is too early to show any tissue abnormalities. A number of studies have been performed to evaluate the ADC changes in different time course. Decreased ADC and visual abnormalities on DWI are most detected during 1-4 days after delivery ^[13] at the moment when conventional imaging may not be that obviously abnormal (Fig 1). Over time, ADC values might return to normal "pseudonormalization" at approximately 7 days after injury. Following pseudonormalization, ADC values become higher than normal during week 2 ^[13].

       As reported by the meta-analysis study ^[14], ADC had a sensitivity of 66% (95% CI: 52%-79%) and a specificity of 64% (95%CI: 35%-87%) for predicting later neurological outcome. Meanwhile, the location of abnormal diffusion parameters is also an important factor; in term infants with HIE, low values of ADC in the PLIC predict relatively poor outcome ^[18].

Click for larger

Fig 1  Term neonate with a history of birth asphyxia. MRI was performed at day 4 of life (a-d) and day 30 of life (e-h). Axial T2-weighted image (a) and IR image (b) showed loss of gray/white matter differentiation in the left frontal lobe. Diffusion-weighted imaging (c) showed very obvious high signal intensity in the left frontal lobe and bilateral parieto-occipital lobe [hypersignal, black arrow] consistent with restricted diffusion in ADC map (d) [hyposignal, white arrow]. Later T2-weighted image (e) at day 30 showed more obvious high signal intensity in the left frontal lobe, but no abnormal signals were detected in DWI (g) or in ADC map (h).

DTI and tractography in neonatal HIE

       DTI is currently the best available non-invasive technique to explore white matter structure in newborns. DTI is a diffusion-weighted technique with at least 6 directions, however generally, 32 directions or more are necessary. The DTI indices, such as FA which expresses the fraction of diffusion anisotropic and mean diffusivity (MD) corresponding to the directionally averaged magnitude of water diffusion, allow us to assess and quantify water diffusion at a microstructural level. Diffusion tensor tractography consists in the reconstruction of the principal white matter tracts in preterm infants' brain, even before myelination becomes histologically evident ^[19] (Fig 2).

       Like ADC, MD is less sensitive to perinatal brain injury after 7 days as compared to the conventional MR because of transient pseudonormalization. Consequently, ADC or MD has been combined with other MR parameters to evaluate the entire temporal evolution of lesions. Diffusion anisotropy values (e.g. FA) decline and remain reduced during the sub-acute and chronic phases of cerebral ischemia, due to disruption of the cytoarchitecture ^{[20, 21]}.

       Fewer studies of diffusion anisotropy were reported concerning its predictive value of long term follow-up compared to the studies of ADC. Arzoumanian and colleagues ^[22] evaluated DTI studies at term-equivalent age and follow-up at a corrected gestational age of 18-24 months on 137 preterm neonates, and found that FA values in the right PLIC were significantly lower for preterm infants with cerebral palsy compared with the control group. Neonatal diffusion tensor imaging may allow earlier detection of specific microstructural abnormalities in infants at risk of neonatal HIE. However, future studies are necessary to evaluate the predictive value of the diffusion anisotropy on long-term neurological outcomes.

Click for larger

Fig 2  Term neonate with HIE and periventricle haemorrhage lesions in the right side. Cortico-spinal tracts (CST) were shown on axial FA map (a) and coronal FA map (b) by probabilistic tractography. Decreased FA value and tract volume were found in the right CST, and later the infant resulted in left hemiplegia.

MRS in neonatal HIE

       MRS is a noninvasive tool that allows analyzing the neurochemicals and metabolites in the human brain. Due to the rapid growth and turnover of the membranes in the neonatal brain, the N-acetylaspartate (NAA; 2.01 ppm) level is much smaller than the Choline (Cho; 3.02 ppm) ^{[23,24,25,26]}. During the early brain development, NAA increases, reflecting the active myelination. Choline (Cho) is a marker for membrane synthesis and Creatine (Cr; 3.0 ppm) was chosen as the metabolite of reference because of its stability. With the brain development, there is a rapid increase of NAA/Cho and NAA/Cr ratios in the first 3 months after birth ^[27] (Fig 3).

       Lactate (Lac; 1.33 ppm) presents very small amounts in the term neonates and is hardly detectable by MRS ^[28]. However, in injured neonates with HIE, a large quantity of lactate is typically present ^[28,29,30], due to anaerobic respiration. MRS shows elevated Lactates, decreased NAA and Cr in an early stage, before anomalies are detected on conventional MR imaging. Abnormal metabolite ratios could be detected on MRS as soon as within the first 24 hours after brain injury ^[29]. Later on, Lac/NAA ratio in the basal ganglia and thalami increased during the first 4 days, and tended to worsen until about day 5 and then normalize; still, sometimes, abnormal metabolite ratios persisted ^[30] (Fig 3).

       Several studies found a reduced NAA and increased Lac in the basal ganglia, which were correlated with poor neurological outcomes ^{[28,29,30,31,32]}. The meta-analysis in neonatal encephalopathy ^[14] concluded that in deep gray matter, Lac/NAA peak-area ratio had an 82% sensitivity (95% CI: 74%-89%) and a 95% specificity (95% CI: 88%-99%). Lac/NAA carried a better diagnostic accuracy compared to conventional MRI as well as brain ADC evaluation at any time during neonatal period ^[14]. Thus, whenever performing a MRS, the region of interest of signal voxel or multiple voxel should include the basal ganglia/thalamus. Moreover, the use of a long TE corresponding to the inversed peak of lactate is preferable for a better evaluation of the metabolic changes.

Click for larger

Fig 3  A term neonate with birth asphyxia. MRI was performed on day 4 and day 30 of life. Multiple voxel proton MR spectroscopy at echo time 144ms. An average spectrum over the right basal ganglia shows an inverted lactate peak on day 4 (a), and the lactate decreased on day 30 (b). Moreover, the NAA peak has obviously increased and the NAA/Cho ratio decreased with maturation process.

Conclusion

       The neonatal brain can be imaged safely at 3T; the increased SNR allows faster imaging with improved image quality. Conventional MR imaging and advanced techniques play an important role in the evaluation of neonatal HIE. Diffusion imaging and MRS allow early detection of the disease and improve the accuracy of prognosis. Conventional MR with DWI and MRS is now recommended to be performed between days 2 and 8 after birth in order to establish the pattern of injury and to predict the neurologic outcome ^[15].