Pediatr Radiol (1999) 29: 199±205 Ó Springer-Verlag 1999
Jean-François Chateil Bruno Quesson Muriel Brun Eric Thiaudire Jean Sarlangue Christophe Delalande Claude Billeaud Paul Canioni François Diard
Received: 30 March 1998 Accepted: 24 September 1998 Presented at the 34th Annual Meeting of the European Society of Pediatric Radiology, Lugano, Switzerland, 28±30 May 1997
J.-F. Chateil ´ M. Brun ´ F. Diard Service de Radiologie A, Hôpital Pellegrin, Bordeaux, France
)
J.-F. Chateil ( ) UnitØ de RadiopØdiatrie, Hôpital Pellegrin, Place A Raba LØon, F-33 076 Bordeaux Cedex, France B. Quesson ´ E. Thiaudire ´ C. Delalande ´ P. Canioni RØsonance MagnØtique des Systmes Biologiques, UMR 5536 CNRS/UniversitØ V. SØgalen, Bordeaux, France J. Sarlangue ´ C. Billeaud Service de NØonatalogie, Hôpital Pellegrin, Bordeaux, France
Localised proton magnetic resonance spectroscopy of the brain after perinatal hypoxia: a preliminary report
Abstract Objectives. Perinatal hypoxic ischaemic injury is a significant cause of neurodevelopmental impairment. The aim of this study was to evaluate localised proton magnetic resonance spectroscopy (1H-MRS) after birth asphyxia. Materials and methods. Thirty newborn infants suspected of having perinatal asphyxia (Apgar score < 3) were studied. The mean gestational age was 37 weeks, mean age at the MR examination was 18 days and mean weight was 2.9 kg. A 1.5-T unit was used for imaging and spectroscopy. None of the babies had mechanically assisted ventilation. No sedation was used. Axial T1-weighted and T2-weighted images were obtained. 1H-MRS was recorded in a single voxel, localised in white matter, using a STEAM sequence. Results. Image quality was good in 25 of 30 babies. 1H-MRS was performed in 19 of 30 subjects, with ad-
Introduction Perinatal hypoxic ischaemic injury is a significant cause of cerebral insult. The initial clinical evaluation is difficult. Blood testing and imaging studies are not always sufficient, and there is a need for early identification of newborn infants at risk. In vivo magnetic resonance spectroscopy (MRS) can be used to assess cerebral metabolism. Phosphorus nuclear MRS was the first technique used [1, 2]. Localised proton MRS allows observations of cerebral metabolites in neonates and
equate quality in 16. Choline, creatine/phosphocreatine and Nacetylaspartate peaks and peak-area ratios were analysed. Lactate was detected in four infants. The Nacetylaspartate/choline ratio was lower in infants with an impaired neurological outcome, but the difference was not statistically significant. Conclusions. This study suggests that 1H-MRS may be useful for assessing cerebral metabolism in the neonate. A raised lactate level and decreased N-acetylaspartate/choline ratio may be predictive of a poor outcome. However, in our experience this method is limited by the difficulty in performing the examination during the first hours after birth in critically ill babies, the problems related to use of a monovoxel sequence, the dispersion of the ratios and the lack of determination of the absolute concentration of the metabolites.
might enable prediction of an adverse outcome [3±6]. This technique may play a role in guiding the application of new neuronal rescue strategies. The aim of the present study was to evaluate the feasibility and value of localised proton magnetic resonance spectroscopy (1H MRS) in such babies in clinical practice.
200
Table 1 Clinical, MRI and MRS summaries of the patients. (GA gestational age, BW birth weight, FD fetal distress, AS Apgar scores at 1 and 5 min, PAE postnatal age at time of examination, P present, NA not available, N normal, CI cerebral infarct, BG abnorm-
alities in basal ganglia, WM abnormalities in white matter, MNI mild neurological impairment, SNI severe neurological impairment)
Infant
GA (weeks)
BW (kg)
FD
AS 1¢, 5¢
PAE (day)
MRI diagnosis
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
37 42 41 38 42 36 36 41 40 40 41 37 40 36 37 36 38 40 36 37 40 27 38 40 37 40 27 34 39 34
2.23 2.78 3.80 1.90 3.40 2.70 2.40 3.70 3.30 3.90 4.06 1.90 3.14 2.50 3.10 2.20 2.90 3.20 1.97 2.65 3.00 0.86 3.50 3.00 2.20 4.00 0.90 1.90 3.80 1.50
P P P
1 1, 7 2, 8 1, 4 1, 1 2, 7 2, 8 10, 10 0, 10 4, 6 4, 9 10, 10 1, 5 0, 7 1, 5 10, 10 1, 5 1, 8 1, 8 2, 8 1 10, 10 1 1 4 1 0, 0 8, 10 1, 6 3, 10
40 7 12 7 12 20 20 14 12 14 2 15 8 28 7 26 11 12 26 13 10 58 12 38 4 5 90 45 6 16
NA NA NA N N N N CI NA N N N NA CI N NA BG NA BG N N N N N NA N N N WM N
a b
P P P P P P P P P P P P P P P P P P P
NAA/Cho
NAA/Cr
Cho/Cr
1.59
1.44
0.91
1.28 0.96 1.22 0.74
4.62 3.55 1.89 0.94
3.61 3.72 1.55 1.26
0.92
1.11
1.21
0.97
2.05
2.11
0.93/0.51a 0.91
1.32/1.36a 0.86
1.42/2.67a 0.95
1.16/1.03b
1.72/1.26b
1.48/1.22b
0.75 1.02 1.37 0.78
1.70 3.65 1.69 1.05
2.27 3.57 1.23 1.35
0.93 0.64
1.70 0.92
1.82 1.42
Lactate peak
P
P P
P
Outcome NA N N N N MNI MNI N N N N MNI N MNI N N N NA N N N N N N MNI N SNI MNI N N
Second voxel studied in the temporal lobe Second voxel studied in the basal ganglia
Material and methods Thirty infants, whose average gestational age was 37 weeks and whose average birth weight was 2.8 kg, were studied. Fetal distress had been suspected in 22 of the babies. All were suspected of having suffered birth or post-natal asphyxia, with a corresponding Apgar score below 3 in most and a need for resuscitation after birth, with tracheal intubation in some cases (Apgar score at 5 min in these cases was not always recorded). Some had a normal Apgar score at birth, but suffered postnatal distress, temporarily requiring assisted ventilation in an intensive care unit. Clinical features of these infants are given in Table 1. The magnetic resonance study was performed at an average age of 18 days of life. No subjects required mechanical ventilation during the procedure. The MR study was conducted with a 1.5-T unit (Magnetom SP 63, Siemens) and a standard quadrature transmit/receive head coil. The examination was performed during natural sleep without sedation. The infants were monitored with pulse oximetry and ECG (Maglife, Brucker) under clinical supervision. After a 3D localisation sequence, T1-weighted (T1-W) spin echo (TR/TE, 450/15 ms) and T2-weighted (T2-W) spin echo
(TR/TE, 2500/90 ms) images were obtained in the axial plane. The specific absorption rate was systematically evaluated in view of the low weight of the subjects. It was always below the maximum value of 2.4 W/kg (average, 0.41 W/kg). MR spectroscopy was performed with a STEAM sequence, with a TR of 2 s, a TE of 135 ms, and a bandwidth of 1 kHz. A single voxel of 8 cm3 was studied. Two measurements were obtained in two infants and one measurement in the others. The common region of interest was the paraventricular white matter. In most cases, the right side of the brain was studied, but in some children, because of the orientation of the head in the coil, the left side was chosen. Regardless of the side involved, the voxels were located in the centrum semiovale at the same distance from the brain midline. A second region of interest in one baby was the basal ganglia and the temporal lobe in another, where abnormal areas had been detected by imaging. The localised shim was manually optimised, resulting in a mean value of 8 Hz line-width at half maximum for the unsuppressed water signal. Water resonance was suppressed by a specific pulse, manually adjusted for each sequence. One-hundred-twenty-eight or 256 acquisitions were averaged. This sequence lasted from 4 to 8 min. The total imaging session lasted from 20 to 90 min, with an average duration of approximately 1 h.
201
b
c MR imaging
a Fig. 1 a±c Patient 14. a axial T1-W MR imaging shows an ischaemic right temporal lesion. b Voxel in the parietal white matter shows normal spectra (Myo myo-inositol, Cho choline, Cr creatine, NAA N-acetylaspartate) c Voxel in the right temporal area; localised spectra shows a low NAA peak and inverted lactate peak (lac) at 1.3 ppm
Data were processed using exponential filtering (4±15 Hz line broadening), manual phasing (order 0 and 1) and, in 14 cases, spline baseline correction. The chemical shifts were referenced to that of N-acetylaspartate (NAA) at 2 ppm. Spectral peaks were attributed to the following metabolites, according to their chemical shift: · · · · ·
myo-inositol ± 3.5 ppm choline-containing compounds ± 3.2 ppm creatine plus phosphocreatine ± 3 ppm N-acetylaspartate ± 2 ppm lactate ±1.3 ppm
Areas of the majors peaks were measured, and the ratios of NAA/ choline, NAA/creatine and choline/creatine were calculated. Neurological outcome of these infants was clinically evaluated at 1 year of age and classified as normal or mild or severe neurodevelopmental delay. No controls were enrolled into this study. The data were analysed by means of Student's t-test to compare the findings in subsequently normal and deficient infants. The study was approved by the local ethics committee and parental consent was obtained for each infant.
Results Twenty-five studies were obtained. Five were invalid because of poor quality of the images and spectra due to head motion during the examination.
MRI was abnormal in five cases. In case 8, left parietal focal cortical and subcortical hyperintensity was detected on T1-W images in relation to a haemorrhagic infarct. In case 14, right temporal hypointensity was obtained on T1-W images (Fig. 1 a), corresponding to temporal infarction. In two cases (17 and 19), MRI showed abnormal high signal in basal ganglia related to ischaemic lesions (Fig. 2 a,b). In case 29, patchy subcortical parieto-occipital white matter hyperintensities were present on T1-W images, in relation to subacute petechial haemorrhage (Fig. 3 a). MR spectroscopy MR spectroscopy was carried out in 19 cases. In 3 cases, spectra were of poor quality. The three major peaks were visible, but the quality of the data was not sufficient to execute reproducible calculation of the peak areas. In the other 16 babies, spectral resolution allowed peak area calculations (Fig. 1 b). In these 16 patients, the voxel studied was located in 12 cases on the right side and in 4 cases on the left side. Peak-area ratios and average are given in Tables 1 and 2. No significant difference was demonstrated between the ratios calculated on the right and those calculated on the left. Among these 16 cases, a lactate peak, which was inverted, was detected in 3 at approximately 1.3 ppm. In the 2 cases in which a second voxel was studied, a lactate peak was detected in one (Fig. 1 c). Neurodevelopmental outcome In 2 of the 30 children, a 1-year follow-up evaluation was not obtained. Twenty-one of the babies did not develop neurodevelopmental impairment. Seven developed impairment, which was minor in 6 and severe in 1. Among the 7 infants who showed impairment, MRS was available in 5.
202
Fig. 2 a±d Patient 17. a Axial T1-W MR imaging shows bilateral high signal in the basal ganglia. b Axial T2-W MR imaging confirms bilateral hyperintense lesions in the putamena. c Voxel in the right parietal white matter shows normal localised spectra. d Voxel in the right basal ganglia shows normal localised spectra (Cho choline, Cr creatine, NAA N-acetylaspartate)
a
b
c
d
Comparison between MRI and MRS We compared the spectroscopy results with the imaging findings. MRI was abnormal in five infants, among whom 1H MRS was obtained in three (Figs. 1±3). Abnormal findings were shown by both examinations in only one case, in which MRS detected a lactate peak in the abnormal area demonstrated by MRI (Fig. 1). MRI did not show any abnormalities in ten cases, among whom 1H MRS was available. Average ratios of metabolites are given in Table 2; there was no statistical difference between these babies and the entire series. Comparison between MRS and clinical outcome Among the five children who had neurodevelopmental delay and a valid MRS result, three had a lower NAA
peak (Fig. 1 c) than the infants with normal neurological outcome, There was no statistically significant difference between the average ratios of NAA/choline, NAA/creatine and choline/creatine of the five children with impairment and the respective average ratios of the children with normal outcome. Two of the infants with neurodevelopmental delay had a detectable level of lactate revealed by 1H MRS.
Discussion Acute cerebral hypoxia during birth is a major cause of brain injury. Data have previously been reported suggesting that 1H MRS might be helpful in predicting neurological outcome in such babies [3, 4, 7]. The aim of the present study was to assess the practical value of 1H MRS in a clinical setting. In this context, several re-
203
Fig. 3 a, b Patient 29. a Axial T1-W MR imaging demonstrates patchy parieto-occipital white matter hyperintensities (arrow). b Localised spectra, voxel in the parietal white matter shows no significant decrease in the NAA peak (Myo myo-inositol, Cho choline, Cr creatine, NAA N-acetylaspartate)
a
b
Table 2 Average ratios with standard deviation of NAA/choline, NAA/creatine and choline/creatine in the 16 patients evaluated with MRS, in the 10 patients who had normal MRI and statistical
NAA/choline NAA/creatine Choline/creatine
comparison between infants with normal (11) as opposed to impaired (5) neurological development (Student's t -test, NS not significant)
Complete series
Patients with normal MRI
Normal outcome
Neurodevelopmental delay
0.98 ± 0.27 1.92 ± 1.07 1.94 ± 0.94
0.99 ± 0.24 2.33 ± 1.31 2.17 ± 1.08
1.05 ± 0.26 2.20 ± 1.26 2.02 ± 1.15
0.86 ± 0.24 NS 1.43 ± 0.45 NS 1.73 ± 0.55 NS
marks are prompted by the experience gained during the present study. The first difficulty involved the reliability of the criteria used for recruitment of the infants. All were suspected to have hypoxic ischaemic brain injury, with Apgar scores below 3. The interpretation of an Apgar score, however, may be problematical because its determination cannot be dissociated from the active resuscitation measures which may, fortunately, alter the score. The presence of clinical signs of acute hypoxic encephalopathy (such as exaggerated response to stimulation, convulsions and abnormally reduced level of consciousness) may also provide valuable prognostic information to indicate the need for a complementary investigation. The technical limitations of the MR examination must be considered. Monitoring and intensive care MRI and MRS are not easy to perform during the first hours of life in babies with assisted ventilation, or when the infant weighs less than 2 kg. To study critically ill infants, it is necessary to use a specialised non-magnetic incubator equipped with mechanical ventilation and a complete system for physiological monitoring. Our MR
unit is equipped for such monitoring of vital signs throughout the procedure, but specific incubators are not marketed. Systems designed to handle these types of indications need to be developed. In the present series, it was not possible to perform the study during the first hours of life, or during the critical early period in the most severely affected infants. Most of our babies had a normal neurological outcome and none died, indicating that moderate neonatal distress predominated in our patients. It will be difficult to reconcile early urgent care with optimum use of 1H MRS, which is believed to be most relevant when performed within the first hours after birth [3, 4]. Motion artefacts are difficult to avoid and sedation may be dangerous in non-intubated infants. Because of such difficulties, only 16 infants (53 %) were studied by MRS in the present study. Choice of sequence and region of interest The two main techniques for single voxel 1H MRS are PRESS (point-resolved spectroscopy) and STEAM (stimulated-echo acquisition mode) sequences. Signal/ noise ratio is better with the former, but localisation of the voxel is more precise with the latter. Most of the pre-
204
viously published studies have been performed with a TE of 130±270 ms. Shorter TE prevents signal losses due to T2 relaxation and are probably more suitable for the detection of glutamate and glutamine. In the current study, data were collected using a STEAM sequence with a TE of 135 ms. With TE shorter than this, residual eddy currents were difficult to eliminate with the system used by the current investigators. Monovoxel sequences require selection of the voxel position. Previous studies have been performed with a voxel centred on the thalamus and/or the parieto-occipital white matter. Several factors entered the selection of the voxel location in the current study. Basal ganglia are known to be particularly at risk from acute asphyxia [5], but parasagittal and subcortical regions are also susceptible to injury in term infants. In the present study, when the voxel was centred on the thalamus, partial volume effects from adjacent structures, including the circle of Willis, midbrain tectum and cisterns, were observed. Therefore, we chose to place the voxel in the centrum semiovale. The right side of the brain was investigated in most cases, but in some patients the voxel was localised on the left side. No controls were assessed to determine whether there might have been a significant difference between the two hemispheres. Regarding the possibility of lipid contamination, the voxel position was chosen so that the centre of the voxel was located 2 cm or more from the scalp. The gradient strength used for spatial selection was 2.7 mT/m, corresponding to 1150 Hz/cm. Under such conditions, lipid contamination, if any, would appear in spectral regions that do not overlap the 0±4 ppm (250 Hz) region analysed in the present study. Multi-voxel localisation techniques have been introduced. Multi-dimensional chemical-shift imaging techniques allow reconstruction of metabolite maps that represent the spatial intensity of individual metabolite resonances [7±9]. In the study of Hanrahan et al. [4], the size of the matrix was 32 ´ 16, with a spatial resolution of 15 ´ 15 mm. However, spectral resolution chemical-shift imaging is less precise than monovoxel studies, water suppression is more difficult to obtain and acquisition times are longer. Lipid signals caused by subcutaneous fat are also difficult to suppress. In practice, when the concentrations of water and lipids are relatively high, the performance of the pulses used for water suppression and volume selection is critical [10]. Improved resolution will be necessary for chemical-shift imaging to contribute to clinical practice. Processing of data For each patient, an optimum non-automated localised shim in the voxel had to be obtained, and the value of the pulse for the water resonance suppression had to be
adjusted. In the current series, data were processed by exponential filtering of line broadening before Fourier transformation, manual phasing and, in most cases, manual baseline correction. The peaks were not fitted and peak areas were determined using the software of the system. Thus, the present spectral analysis was `operator-dependent'. The main metabolites observed were: · myo-inositol, which is thought to be specific to the glia · choline, which is found in pathways of synthesis and breakdown of myelin · creatine, which is found in all cells · NAA, which is predominantly neuronal, but is also found in immature oligodendrocytes in the neonatal brain · lactate, the peak of which, with a TE of 135 ms, is inverted in phase Metabolite peak-area ratios were used to analyse the spectra. The present study confirms previous reports concerning the relatively low peak of NAA in the normal neonatal brain, compared to older infants. Maturation of the brain is associated with changes in biochemical composition and metabolism, with an increase of the NAA/ choline ratio [11, 12]. In the series presented here, the metabolite ratios were quite dispersed, and there was no clear distinction between the children with favourable outcome and the others. We found lower NAA/choline and NAA/creatine ratios in babies with impaired outcome, but the difference was not statistically significant. Infants in the current study probably did not have extensive brain lesions. Other studies have shown that a low NAA/choline ratio during the first 2 weeks of life indicates a poor outcome [7, 13±15]. However, this was not confirmed by a study performed at an even earlier age in which no correlation was found [3]. Increased lactate may have been caused by anaerobic glycolysis, or have been produced by phagocytes invading necrotic areas. Lactate is generated early after hypoxia and ischaemia, and a second increase takes place within 24±48 h after the insult, associated with the socalled `secondary energy failure' [3, 16]. Observation of lactate in term neonates indicates a poor long-term prognosis [3, 5, 7]. In preterm infants, the significance of lactate is unclear and may be an aspect of normal brain metabolism. It has been reported that lactate can be used in the brain as a metabolic fuel [17]. In the present series, lactate was detected in four infants, two of whom showed neurodevelopmental delay. However, the babies were older at the time of examination than in comparable series, and most of the current patients were not severely affected. Analysis of metabolite ratios, as performed in this study, allows only relative changes to be detected; peakarea ratio can vary due to altered relaxation times without any changes in metabolite concentrations. Quantita-
205
tive determination of regional absolute concentrations of metabolites, compared with an internal or external reference, would resolve this problem [18]. Determination of the T2 of each metabolite is also necessary to estimate its absolute concentration; variations of T2 may relate to increase in the cytosolic water, possibly reflecting neuronal oedema [9]. With a system that provides absolute values of the concentrations of metabolites, comparisons with normal values would be possible. Finally, a comparison with diffusion-weighted magnetic resonance imaging and MRS has to be evaluated to confirm the accuracy of these two methods for the early diagnosis of human neonatal cerebral hypoxic ischaemic injury [19, 20]. In conclusion, in addition to MR imaging, proton MR spectroscopy may prove to be useful in neonates,
but in our study the latter examination showed no significant difference between normal and mildly impaired babies. A longer follow-up will be necessary to determine whether a correlation exists between the MRS results and clinical outcome. Technical improvements, including specific incubators, head coils designed specifically for MRS in babies, chemical-shift imaging sequences, reproducible evaluation procedures and, if possible, determination of the absolute concentration of each metabolite will be necessary for wider clinical use of this tool. Acknowledgements This work was supported by the Centre National de la Recherche Scientifique and the Centre Hospitalier Universitaire de Bordeaux.
References 1. Cady EB, Costello AM, Dawson MJ, et al (1983) Non-invasive investigation of cerebral metabolism in newborn infants by phosphorus nuclear magnetic resonance spectroscopy. Lancet 8333: 1059±1062 2. Roth SC, Edwards AD, Cady EB, et al (1992) Relation between cerebral oxidative metabolism following birth asphyxia, and neurodevelopmental outcome and brain growth at one year. Dev Med Child Neurol 34: 285±295 3. Penrice J, Cady EB, Lorek A, et al (1996) Proton magnetic resonance spectroscopy of the brain in normal preterm and term infants, and early changes after perinatal hypoxiaischemia. Pediatr Res 40: 6±14 4. Hanrahan JD, Sargentoni J, Azzopardi DA, et al (1996) Cerebral metabolism within 18 hours of birth asphyxia: a proton magnetic resonance spectroscopy study. Pediatr Res 39: 584±590 5. Leth H, Toft PB, Peitersen B, et al (1996) Use of brain lactate levels to predict outcome after perinatal asphyxia. Acta Paediatr 85: 859±864 6. Shu SK, Ashwal S, Holshouser BA, et al (1997) Prognostic value of 1HMRS in perinatal CNS insults. Pediatr Neurol 17: 309±318 7. Groenendaal F, Veenhoven RH, Van Der Grond J, et al (1994) Cerebral lactate and N-acetyl-aspartate/choline ratios in asphyxiated full-term neonates demonstrated in vivo using proton magnetic resonance spectroscopy. Pediatr Res 35: 148±151
8. Groenendaal F, Van Der Grond J, Witkamp TD, De Vries LS (1995) Proton magnetic resonance spectroscopic imaging in neonatal stroke. Neuropediatrics 26: 243±248 9. Cady EB (1996) Metabolite concentrations and relaxation in perinatal cerebral hypoxic-ischemic injury. Neurochem Res 21: 1043±1052 10. Nelson SJ, Vigneron DB, Star-Lack J, Kurhanewicz J (1997) High spatial resolution and speed in MRSI. NMR Biomed 10: 411±422 11. Van der Knapp MS, Van der Grond J, Van Rijen PC, et al (1990) Age-dependent changes in localized proton and phosphorus MR spectroscopy of the brain. Radiology 176: 509±515 12. Hüppi PS, Fusch C, Boesch C, et al (1995) Regional metabolic assessment of human brain during development by proton magnetic resonance spectroscopy in vivo and by high-performance liquid chromatography/gas chromatography in autopsy tissue. Pediatr Res 37: 145±150 13. Peden CJ, Cowan FM, Bryant DJ, et al (1990) Proton MR spectroscopy of the brain in infants. J Comput Assist Tomogr 14: 886±894 14. Peden CJ, Rutherford MA, Sargentoni J, et al (1993) Proton spectroscopy of the neonatal brain following hypoxicischaemic injury. Dev Med Child Neurol 35: 502±510
15. Kimura H, Fujii Y, Itoh S, et al (1995) Metabolic alterations in the neonate and infant brain during development: evaluation with proton MR spectroscopy. Radiology 194: 483±489 16. Lorek AK, Takey Y, Cady EB, et al (1994) Delayed (ªsecondaryº) cerebral energy failure after acute hypoxia-ischemia in the newborn piglet: continuous 48-hours studies by phosphorus magnetic resonance spectroscopy. Pediatr Res 36: 699±706 17. Leth H, Toft PB, Pryds O, et al (1995) Brain lactate in preterm and growth-retarded neonates. Acta Paediatr 84: 495±499 18. Kreis R, Ernst T, Roos BD (1993) Development of the human brain: in vivo quantification of metabolite and water content with proton magnetic resonance spectroscopy. Magn Reson Med 30: 424±437 19. Cowan FM, Pennock JM, Hanharan JD, et al (1994) Early detection of cerebral infarction and hypoxic ischemic encephalopathy in neonates using diffusion-weighted magnetic resonance imaging. Neuropediatrics 25: 172±175 20. Thornton JS, Ordidge RJ, Penrice J, et al (1998) Temporal and anatomical variations of brain water apparent diffusion coefficient in perinatal cerebral hypoxic-ischemic injury: relationships to cerebral energy metabolism. Magn Reson Med 39: 920±927