 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
a Department of Paediatrics, MRC Clinical Sciences Centre, Imperial College Faculty of Medicine, Hammersmith Hospital, London, England
b Imaging Sciences Department, MRC Clinical Sciences Centre, Imperial College Faculty of Medicine, Hammersmith Hospital, London, England
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONREFERENCES |
|---|
METHODS. Twenty infants with HIE and 7 normal control infants were studied. All infants were born at >36 weeks' gestational age, and MRI scans were obtained within 3 weeks of delivery. Data were examined for normality, and comparisons were made using analysis of variance or Kruskal-Wallis as appropriate.
RESULTS. During the first week, FA values were decreased with both severe and moderate WM and BGT injury as assessed by conventional imaging, whereas ADC values were reduced only in severe WM injury and some severe BGT injury. Abnormal ADC values pseudonormalized during the second week, whereas FA values continued to decrease.
CONCLUSION. FA is reduced in moderate brain injury after HIE. A low FA may reflect a breakdown in WM organization. Moderate BGT injury may result in atrophy but not overt infarction; it is possible that delayed apoptosis is more marked than immediate necrosis, and this may account for normal early ADC values. The accompanying low FA within some severe and all moderate gray matter lesions, which is associated with significant later impairment, may help to confirm clinically significant abnormality in infants with normal ADC values.
Key Words: brain imaging • diffusion tensor imaging • hypoxic-ischemic encephalopathy • magnetic resonance imaging • neonates
Abbreviations: HIE—hypoxic-ischemic encephalopathy • GA—gestational age • DWI—diffusion-weighted imaging • ADC—apparent diffusion coefficient • WM—white matter • BGT—basal ganglia and thalami • ROI—region of interest • DTI—diffusion tensor imaging • FA—fractional anisotropy • RA—relative anisotropy • PLIC—posterior limb of the internal capsule • CSO—centrum semiovale • LN—lentiform nuclei • MT—medial thalamus • VLN—ventrolateral nuclei of the thalamus
Hypoxic-ischemic encephalopathy (HIE) is an important cause of mortality, morbidity, and adverse neurodevelopmental outcome in infants who are born at term, with an overall incidence between 1 and 2 per 1000 live births. Asphyxia accounts for up to 25% of total perinatal morbidity and mortality, as well as up to 15% of all cases of cerebral palsy.1
The pattern of injury after a hypoxic-ischemic event depends on the gestational age (GA) of the infant and also the duration and the severity of asphyxia to which they are subjected. Despite a range of suggestive clinical parameters, the diagnosis of HIE and the prediction of its eventual outcome remain notoriously challenging. Accurate initial diagnosis is also necessary to assess the effect of an intervention. Early imaging studies have shown that even with relatively strict entry criteria, the pattern of injury in infants with HIE may be very variable. It remains unclear whether this reflects differences in the timing of injury, the nature of the insult, or individual susceptibilities. These questions can be answered only by additional studies that improve both entry criteria and the description of the brain injury.
The sensitivity of MRI has been exploited in the study of normal brain development,2,3 characterizing changes in myelination and cortical folding. Given the varying and sometimes subtle patterns of injury in HIE, MRI is an attractive diagnostic and prognostic modality. Studies with conventional MRI have characterized HIE-associated lesions,4,5 and these have been shown to correlate with the type of hypoxic-ischemic insult, apparent clinical features,6 neurologic examination,7 and electroencephalographic measurements.8 Importantly, MRI has been shown to be a good predictor of neurodevelopmental outcome.9,10 However, abnormalities on conventional MRI may not be obvious within the first few days after delivery, particularly to those who are not experienced in assessing perinatal pathology. Therefore, there is a need for objective methods to improve the detection of ischemic tissue, to confirm suspected tissue injury as seen on conventional images, and to quantify these findings. For additional understanding of the evolution of imaging abnormalities after a variety of insults, a more accurate assessment for timing an injury would be extremely useful not only for clinical management but also for medicolegal issues.
Diffusion-weighted imaging (DWI) has been shown to identify ischemic tissue within hours of the onset of adult stroke11; it is able to detect ischemic lesions in perinatal stroke, but these studies have been performed after symptomatic onset, which is usually within days, not hours, of delivery. In perinatal stroke, DWI abnormalities are most notable in the first 4 days after birth but later pseudonormalize as conventional MRI scans first begin to exhibit pathology.12 However, perinatal stroke is not typical of the type of brain injury seen with HIE, in which the insult is global and followed by reperfusion. Only a few relatively small studies have investigated whether DWI may offer more immediate detection of tissue injury in infants with HIE, with somewhat conflicting conclusions.13–18 Of note, several of these studies did not measure apparent diffusion coefficient (ADC) values, reporting the sensitivity of DWI on the basis of visual analysis of the images only, which in some cases may be inconclusive. More recently, we demonstrated in a larger cohort of infants significantly reduced ADC values between control subjects and infants with severe white matter (WM) lesions and some severe basal ganglia and thalami (BGT) lesions in several regions of interest (ROI).12 However, values in infants with moderate but nonetheless clinically significant lesions were either normal or slightly raised. In addition, pseudonormalization, whereby abnormal ADC values approximate that of normal control subjects toward the end of the first week after birth, was also apparent. Although this phenomenon limits the usefulness of DWI later in the perinatal period, abnormalities on conventional imaging are usually apparent by this time.
Diffusion tensor imaging (DTI) allows the measurement of the directional diffusivity of water and may be more sensitive than DWI in detecting brain injury in neonates with HIE. At present, no studies have measured fractional anisotropy (FA) values in a cohort of infants with HIE. The objectives of this study were to ascertain whether DTI can improve the detection of brain injury in infants with HIE and, more specific, (1) to relate DTI findings to the pattern and severity of brain lesions, (2) to determine whether DTI measures of anisotropy improve the detection of severe and moderate BGT or moderate WM injury, and (3) to relate DTI findings to the timing of the scan from delivery.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
On the basis of our own and other previous DWI studies that show ADC values to evolve after birth in both encephalopathic and control infants, both cohorts were stratified on the basis of their postnatal age at scan: the early group included infants who were imaged during the first week after birth (
7 days), and the late group comprised those who were imaged in the second and third weeks (8–21 days).
Imaging
All infants were imaged at the Robert Steiner MRI Unit, Hammersmith Hospital. Patients were usually sedated using oral chloral hydrate (30–50 mg/kg), whereas control subjects were imaged during natural sleep. Infants wore ear protection that consisted of molded earplugs and specialist ear protection (Natus MiniMuffs; Natus Medical Inc, San Carlos, CA) and were monitored using pulse oximetry and electrocardiography throughout the scan. An experienced neonatologist, who was trained in MRI procedures, was in attendance throughout the imaging process.
MRI was obtained using a 1.5 Tesla Philips Eclipse scanner with a dedicated pediatric head coil. Conventional transverse T1-weighted spin echo (500/15 ms) and T2-weighted fast spin echo (4200/210 ms) with 192 x 256 matrix and 5-mm slice thickness, as well as 3D RF spoiled gradient echo images were obtained before DTI. Subsequently single-shot echo-planar imaging DTI was acquired in 12 (6 noncollinear) directions using a b value of 710 s/mm2, repetition time 6000 ms, echo time 100 ms, field of view 240 mm, matrix 100 x 100, and slice thickness 5 mm.
In-house software then was used to remove image distortion as a result of eddy currents and to construct ADC and FA maps on a per-pixel basis19:
ADC maps were calculated using
 | (1) |
 | (2) |
is gyromagnetic ratio for protons, G is amplitude of the pulsed gradient, 
is duration of the pulsed gradient, and 
is time between leading edges of the 2 pulsed gradients. The rotationally invariant measures of anisotropy, FA and relative anisotropy (RA), were used to characterize the diffusion tensor:
 | (3) |
 | (4) |
1, 2, and 3 are the principal eigenvectors of the diffusion tensor
 | (5) |
Analysis of Imaging
Conventional T1- and T2-weighted imaging was assessed subjectively for abnormal anatomy and/or signal intensity by 1 researcher (M.R.), who was experienced in interpreting neonatal brain MRI scans. For infants who were imaged very soon after birth and who survived, repeat imaging was obtained to confirm the pattern of lesions. Each infant was assigned a BGT grade of normal, mild, moderate, or severe. Mild lesions were small and focal with normal myelination in the posterior limb of the internal capsule (PLIC), moderate lesions were multifocal with equivocal or abnormal PLIC, and severe lesions showed complete BGT abnormality with abnormal PLIC. Similarly, WM was graded as normal, moderate (areas of increased T1 or T2), or severe (overt infarction). These grades were allocated on the basis of the specific appearance of each of the studied WM regions (Fig 1).
|
|
Statistics
Intraobserver error was expressed by calculating coefficients of variability after repeating measurements for 6 infants who were chosen at random. All data were subjected to the Shapiro-Wilk test for nonnormality and then as appropriate unpaired t and analysis of variance methods (for parametric data) or Mann-Whitney U and Kruskal-Wallis tests (for nonparametric data).
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Visual Analysis of Conventional Imaging
The imaging of all control infants was within normal limits for their age. Of the 20 encephalopathic infants, 17 had moderate or severe BGT abnormality. Only 2 infants had moderate BGT abnormalities; therefore, all BGT lesions were grouped together. WM abnormalities were more variable: whereas 4 infants had normal WM, 14 had severe abnormalities in the WM of the CSO, 7 had severe anterior WM abnormalities, and 8 had severe posterior WM abnormalities. Two infants had BGT abnormality in the absence of any WM changes, and 1 encephalopathic infant had normal BGT and WM imaging.
Four infants, 1 control infant and 3 patients, were imaged within 48 hours. In 2 of these, repeat imaging confirmed the initial imaging appearances at 2 and 6 weeks, respectively. The remaining 2 were followed up but did not have repeat imaging. The control infant had normal development, and the patient with BGT lesions had developed a motor impairment. The visual analysis of DWI and ADC trace images is not included in this study.
ADC and Anisotropy Measurements
A single researcher (P.W.) produced measurements to minimize observer error; the average coefficient of variability over all ROIs was 1.5% for ADC, 3.89%, for FA and 4.26% for RA values. For assessment of any influence of GA at delivery on measurements, changes in ADC and anisotropy in all of the WM regions were examined in the control infants. These areas are known to be actively myelinating during the neonatal period and infancy (and thus rapidly changing in terms of diffusion characteristics).
ADC Values
In the control infants, although both WM and BGT ADC values tended to decrease with increasing age at scan, this observation did not reach significance over the relatively small range of GA in this study. Similarly, there was no difference between the early (imaged <7 days' postnatal age) and late (imaged 8–21 days) control groups. When multiple regression was used to examine simultaneously the effect of both GA at delivery and postnatal age at scan, again no significant differences were found. The ADC values for all patients and control subjects are shown in Table 1. The larger range of values in the patients represents varying pathologic severity between infants, as well as their age at the time of scan, consistent with our previous work.
|
|
|
|
White Matter
In the first postnatal week, FA values were significantly decreased not only in infants with severe WM abnormality but also in those with moderate abnormality. In the anterior WM, median anisotropy was 0.110 (0.089–0.164; P = .0271) and 0.114 (0.098–0.140; P = .0315) for patients with moderate and severe WM pathology, compared with 0.157 (0.137–0.193) for control infants. Aberration in posterior WM anisotropy was even more pronounced (Fig 6), with values of 0.144 (0.096–0.175; P = .0007), 0.138 (0.124–0.198; P = .0016), and 0.225 (0.184–0.254) in the moderate pathology, severe pathology, and control groups, respectively. In the CSO WM, infants with severe abnormality had significantly decreased FA in the first week (0.167 [0.091–0.279] vs control infants 0.289 [0.210–0.314]; P = .0032). Of particular interest, given the phenomenon of pseudonormalization described with ADC values, is that compared with similarly aged control infants, CSO WM anisotropy was also significantly decreased in patients who had severe abnormality and were imaged during the second and third weeks (0.113 [0.08–0.127] vs 0.221 [0.177–0.259]; P = .006). Indeed, there was significant difference between the values of patients who were imaged during the first week and those who were imaged during the second and third weeks (P = .015), but, more important, anisotropy became more deranged rather than pseudonormalized. In patients with lesions but some areas of normal-appearing WM, values in usually normal WM were as follows: median FA was 0.132 (0.126–0.153) in the anterior WM, 0.137 (0.136–0.160) in posterior WM, and 0.151 (0.119–0.175) in CSO WM. In the last 2 ROI, FA was significantly reduced compared with that in control infants (P < .03).
|
|
|
Predictive Value of FA
Using data from all of the infants, regardless of postnatal age at scan or the severity of any WM or BGT injury as assessed visually, the predictive power of FA for tissue injury as identified on conventional imaging analysis exceeded that of ADC in all ROIs. Of all WM regions, FA was most predictive in the posterior WM, where a value of <0.16 was 81% sensitive and 85% specific for some degree of abnormality. In the BGT, a value <0.260 in the VLN was 82% sensitive and 86% specific for moderate or severe BGT injury. Importantly, the sensitivity and the specificity of these values improved when the data were limited to infants who were imaged during the first week (sensitivity: >85%; specificity: 100%), indicating that the high level of predictive power can be attributed to the sensitivity of FA, rather than simply the lack of pseudonormalization after insult. FA of <0.380 in the PLIC was >80% sensitive and specific for BGT injury but reduced to 75% when based on first week data alone.
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
A large range of ADC values was evident in the patient group, suggesting that infants had lesions that varied not only in their location and severity but also in their degree of evolution at the time of imaging. ADC values were decreased in the first postnatal week in the WM regions of patients with severe WM abnormality and in the LN and PLIC of those with severe BGT pathology, before normalizing as exhibited by infants who were imaged during the second and third weeks. This pseudonormalization obviously limits the absolute sensitivity of the parameter, but by the time it has occurred, lesions are usually more obvious on conventional imaging. In addition, infants with moderate pathology demonstrated normal or marginally increased ADC values compared with control infants. These findings are in agreement with our previous, larger DWI study as well as other investigations with focal infarction in neonates and adult patients.
Given the concurrence of the present ADC data with previous work, the observed patterns in anisotropy are also relevant. In the control infants, FA was highest in the WM of the PLIC (median: 0.439), which is in keeping with the fact that this area consists of tightly packed and parallel fibers that are actively myelinating at term; this organization and the multiple lipid layers of myelin tend to restrict the direction across axonal tracts. Experimental demonstration of anisotropy in nonmyelinated WM indicates that axonal diameter and membrane properties and the activity of oligodendrocytes modulate diffusion characteristics.21,22 These principles underlie observations that anisotropy increases with increasing brain development21,23 and that the change is greater in WM than in central gray matter.24
Anisotropy was significantly decreased in patients with not only severe but also moderate WM pathology and not only in infants who were imaged in the first postnatal week but also in infants who were imaged in the second and third weeks. Abnormally low FA values have also been documented in chronic infarction in adult stroke studies.25 The persistently abnormal FA differentiates the characteristics of the parameter from ADC in 2 respects: first, that it seems to be more sensitive, given the detection of moderate injury, and, second, that the parameter remains abnormal when ADC values have pseudonormalized.
FA was also lower throughout the BGT and also the PLIC. One would expect this in the PLIC given its known role as a predictor of neurodevelopmental outcome in HIE9 As with WM regions, FA was most abnormal and remained abnormal after the first week in the VLN. Anisotropy in the LN and MT, although low in the first week, was normal in the infants who were scanned later. The PLIC and VLN have the highest FA values in the control infants; therefore, it is possible that disturbance of structure is easier to detect. It is of interest that the FA values within the VLN are relatively high compared with the LN and even with the MT. Differences between thalamus and lentiform nucleus anisotropy have been shown in other pediatric and adult studies,26,27 but these have not specifically measured different regions of the thalamus. Relatively high FA values in the VLN may be explained by the presence of fibers within these nuclei, which are already myelinated in the term neonate.3 The degree of anisotropy within the nuclei will reflect the relative orientation of these fibers, and this has been shown on diffusion tractography in the adult brain.28 The sensitivity of diffusion values in the VLN is consistent with its role within the pyramidal system and the sensitivity of this actively myelinating system to injury. The evolution of our FA changes may be explained by irreversible ischemia and subsequent infarction within the VLN, whereas changes in the MT and the LN may comprise in part extracellular edema that settles after week 1. Conventional images in infants with BGT lesions often show swollen medial thalami with long T1 and long T2, which could be attributable at least in part to extracellular edema. It is also feasible that we were unable to sample the most vulnerable region of the LN, the posterior putamen, because we were attempting to avoid partial volume effects on adjacent WM in the PLIC. In addition, differences between the composition and thus diffusion properties of WM and gray matter may explain these different diffusion properties; reductions in ADC and anisotropy are known to be less in central gray matter structures (in some studies nonexistent) and also possibly pseudonormalize earlier.25
Of interest, FA values in the cerebellar hemispheres of patients with moderate/severe BGT injury were reduced. Given the relatively short period of study, however, this is unlikely to be attributable to delayed injury, although it remains unclear whether this is attributable to an acute cerebellar injury or a secondary effect from diaschisis. ADC values in the cerebellar hemispheres were not reduced acutely, although this does not exclude a moderate acute injury. Correlation between perinatal BGT injury and subsequent impaired cerebellar development has been documented.29 The use of DTI in a cohort of infants who are imaged during a longer temporal period is indicated, given that reduced FA has been shown in areas of Wallerian degeneration after adult stoke.30
In this study, we used the severity of injury as visually assessed on conventional imaging to index the sensitivity of the diffusion parameters; therefore, it is clear that FA and particularly ADC become more aberrant with increasing visual abnormality and injury. However, it is interesting to observe the variation of ADC and FA in the 5 infants with areas of apparently normal WM. Where ADC was similar to that in control infants, FA was reduced, significantly in the posterior and CSO WM. Although we appreciate the small number of patients in this subcohort, this trend suggests that a quantitative measure of anisotropy may detect abnormality when it is not obvious on conventional imaging. Therefore measuring FA would provide a useful adjunct to the visual analysis of conventional MRI, particularly to those who are not experienced in assessing perinatal pathology.
Other literature using DTI in a similar cohort of infants with HIE is sparse. Given known parallels between the evolution of ADC values after infarcts in adults and neonates, it is reasonable to relate our findings similarly. In acute adult stroke, anisotropy changes in distinct temporal phases after the onset of ischemia.
During the first 24 hours, compression of the extracellular space after cellular swelling during necrotic cell death may lead to slightly increased anisotropy, particularly in WM tracts.25 In a study of adult infarction, however, there were no hyperacute changes in anisotropy within 6 hours of stroke onset.31 Buijs et al32 investigated focal neonatal brain ischemia and documented increased FA immediately after the insult. However, it is unusual for perinatal stroke to present within 12 hours of delivery, and it is possible that focal perinatal injury that was already obvious within 24 hours may have occurred before labor. In our study, despite that 3 patients were imaged on the first day after birth, there were no elevated FA values; this may represent normal variation in anisotropy, but more likely, it demonstrates the different patterns of injury resulting in focal infarction and the more global insult associated with HIE. Measuring the separate eigenvectors of the diffusion tensor and thereby assessing their individual effect on the overall FA may help to explain the relationship among injury type, age, and FA.
After the acute phase of adult stroke, FA is decreased compared with control tissue,33 probably reflecting cell death and loss of structural integrity, resulting, again particularly in WM, in an environment that is more permissive to multidirectional (isotropic) diffusion rather than the vastly unidirectional, anisotropic diffusion seen in intact WM tracts. Contemporaneously, ADC values tend to pseudonormalize and subsequently increase compared with control values.25
Our current study clearly demonstrates that these concepts occur in a cohort of HIE patients, and that trends in ADC and anisotropy follow different patterns after insult gives 2 important advantages. First, one may differentiate infants whose conventional imaging will eventually show moderate or severe WM lesions, those who are more likely to sustain a severe injury will have both decreased ADC and anisotropy in the first week, whereas those with moderate injury will have a grossly normal ADC but decreased anisotropy. Second, the timing of the causative insult may be approximated by comparing ADC, which may be abnormal or have pseudonormalized, and FA, which remains or even becomes increasingly abnormal in the first weeks after birth. We do appreciate, however, the need to characterize the evolution of both parameters more closely, and that natural variation may limit the ability to make accurate assumptions that would be required, for example, in the medicolegal arena.
The sample size of the study prohibited several comparisons. Although our findings with WM injury are extremely encouraging, we realize most importantly a lack of infants who have particularly moderate BGT lesions and may still experience adverse neurodevelopmental outcome. This undoubtedly reduced the sensitivity of thalamic ADC values; with a larger cohort, as was available in our previous study, a more thorough BGT lesion classification could have been used. Of critical importance is the availability of control data to establish normal ranges, but the recruitment of true control infants within the postnatal first month is notoriously difficult. It was interesting to note the greater variation in control FA values compared with ADC values, but this may be a manifestation of a more sensitive parameter combined with normal variation.
The ability of FA to predict both WM and BGT abnormality either in the first postnatal week or in fact at any time within the first 3 weeks of postnatal life with >80% sensitivity and specificity is impressive. Despite the small control cohort, this is testament to the possible applications of DTI in the investigation of HIE. These results need to be tested prospectively under the same imaging conditions and in a similar cohort of infants.
It is also important to emphasize that the rotationally invariant parameters FA and RA are only 2 of the functions that may be used to describe the diffusion tensor. Although aberrant FA values seem to be a sensitive detector of abnormal tissue, especially in WM, one could use the relative magnitudes of the 3 principal eigenvectors to describe diffusion in and around a lesion. Given the ordered axonal tract structure of WM, departures from the expected linear diffusion pattern, to either planar or isotropic, may give detailed visualization of the evolution of a lesion, delayed injury, and the response of the brain to the insult. Furthermore, altering the imaging sequence to increase sensitivity to more slowly diffusing water by increasing the b value will alter the diffusion tensor. The effect of different b values on the quantification of anisotropy and their subsequent relation to pathology should be investigated.
The mainstream application of DTI in neonatology, as with other MRI techniques, has been delayed given the relative wealth of studies using the modality in the study of acute adult ischemia. Anisotropy is a sensitive index of adult ischemia and importantly has been shown to correlate with eventual motor outcome. Evidence from this study shows not only that FA varies in a similar manner to experimental and adult human ischemia studies but also that in combination with quantification of ADC values and supporting evidence from conventional imaging, anisotropy may be an extremely useful tool in ascertaining the severity and the temporal evolution of a perinatal hypoxia-ischemia-induced lesion.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Address correspondence to Mary Rutherford, MD, FRCR, Imaging Sciences Department, Robert Steiner MR Unit, Hammersmith Hospital, Du Cane Rd, London W12 0HS. E-mail: m.rutherford{at}imperial,ac.uk
The authors have indicated they have no financial relationships relevant to this article to disclose.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  S. Wang, E. X. Wu, C. N. Tam, H.-F. Lau, P.-T. Cheung, and P.-L. Khong Characterization of White Matter Injury in a Hypoxic-Ischemic Neonatal Rat Model by Diffusion Tensor MRI Stroke, August 1, 2008; 39(8): 2348 - 2353. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A.I. Bartha, K.R.L. Yap, S.P. Miller, R.J. Jeremy, M. Nishimoto, D.B. Vigneron, A.J. Barkovich, and D.M. Ferriero The Normal Neonatal Brain: MR Imaging, Diffusion Tensor Imaging, and 3D MR Spectroscopy in Healthy Term Neonates AJNR Am. J. Neuroradiol., June 1, 2007; 28(6): 1015 - 1021. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2006 American Academy of Pediatrics. All rights reserved. |
||
|
||
, infants with severe WM abnormality; 
, infants with moderate WM abnormality.
