Diffusion-Weighted Magnetic Resonance Imaging in Term Perinatal Brain Injury: A Comparison With Site of Lesion and Time From Birth
Objective. The aim of this study was to establish a more objective method for confirming tissue injury in term neonates who present with early seizures that are believed to be hypoxic-ischemic in origin.
Methods. We studied the relationship between contemporaneous diffusion-weighted magnetic resonance imaging and conventional magnetic resonance imaging in 63 symptomatic term-born neonates and 15 control term infants performed in the neonatal period. Apparent diffusion coefficients (ADC) were obtained for multiple regions of the brain.
Results. ADC values in the 15 control infants were 1 (1–1.15) (median [range]) × 10−3/mm2/second in the thalami and 1.1 (1–1.3) × 10−3/mm2/second in the lentiform nuclei, 1.5 (1.3–1.7) × 10−3/mm2/second in the centrum semiovale, 1.6 (1.46–1.7) × 10−3/mm2/second in the anterior white matter (WM), and 1.55 (1.35–1.85) × 10−3/mm2/second in the posterior WM with little variation over time. ADC values were significantly reduced in the first week after severe injury to either WM or basal ganglia and thalami (BGT), but values normalized at the end of the first week and then increased during week 2. ADC values were either normal or increased in moderate BGT and WM lesions when compared with controls. ADC values < 1.1 × 10−3/mm2/second were always associated with WM infarction and values <0.8 × 10−3/mm2/second with thalamic infarction.
Conclusion. A reduced ADC soon after delivery allows the presence of tissue infarction to be confirmed at a time when conventional imaging changes may be subtle. However, as both moderate WM and BGT lesions may have normal or increased ADC values, a normal ADC value during the first week does not signify normal tissue. ADC values should always be measured in combination with visual analysis of both conventional and diffusion-weighed images for maximum detection of pathologic tissue, and the timing of the scan needs to be taken into account when interpreting the results.
Magnetic resonance imaging (MRI) provides detailed information about the pattern of lesions after perinatal brain injury1–3 and is an excellent predictor of outcome in infants with hypoxic-ischemic encephalopathy.4–7 Bilateral basal ganglia and thalami (BGT) lesions are strongly associated with the development of motor impairment, and the extent of the lesions is closely related to the severity of the impairment. Severe white matter (WM) infarction leads to cognitive impairment but a less severe motor impairment.8 Perinatal stroke results in hemiplegia only when there is involvement of hemispheric WM, BGT, and posterior limb of the internal capsule (PLIC).9
MRI abnormalities may take several days to become obvious, a period during which maximum benefit from interventions designed to modify perinatal brain injury may be obtained and when important clinical decisions may have to be made. Very early confirmation of the site and severity of tissue injury would enable appropriate targeting of therapeutic interventions in a far more specific way than is currently available. In addition, even established abnormalities may not be obvious to the inexperienced radiologist. An additional and more objective method of assessing tissue integrity therefore would be very useful.
Diffusion-weighted MRI (DWI) techniques have been used to study the developing brain10–12 and have been shown to be of value in neonates with perinatal stroke.13,14 Abnormalities are most obvious 1 to 4 days after delivery, but the abnormal signal intensity gradually reduces by the end of the first week as the conventional imaging appearances become more abnormal. The evolution of the diffusion abnormality in perinatal stroke is consistent with and seems to be similar to that seen in adults.15,16
The majority of neonates with neurologic symptoms of probable hypoxic-ischemic origin do not have perinatal stroke but present with a neonatal encephalopathy (NE) resulting from a more global insult to the brain. The role of DWI in NE is not well defined. Only a few small studies have used DWI in this group of infants, and the findings have been inconsistent in terms of both visual analysis and the measurement of apparent diffusion coefficients (ADCs).14,17–22
The aim of the study was to obtain 1) an objective MRI method for confirming tissue injury, 2) information on the time course of change in ADC values in normal infants as compared with symptomatic infants, 3) a better understanding of ADC values in different patterns of lesion, and 4) insights into the mechanism of tissue injury and its timing.
The Hospital Ethics Committee of the Hammersmith Hospital approved this study, and parental consent was obtained for all scans. Infants were eligible when they were delivered at >36 weeks' gestation and presented with seizures in the first 48 hours after delivery. The gestational age (GA) was assessed from maternal dates and early antenatal ultrasound scans. The infants were divided into two groups as previously described23:
Group 1: Infants with NE had abnormal tone patterns, feeding difficulties, and altered alertness and at least 3 of the following criteria: 1) late decelerations on fetal monitoring or meconium staining, 2) delayed onset of respiration, 3) arterial cord blood pH <7.1, 4) Apgar scores <7 at 5 minutes, and 5) multiorgan failure.
Group 2: Infants who had seizures within 72 hours of birth but did not have the criteria for NE as outlined above.
Infants were excluded when there was evidence of metabolic disease, congenital infection, major malformations, alcohol or drug embryopathies, hydrops, and chromosome abnormalities. They were also excluded when there was evidence of long-standing brain damage on their initial scans or evidence of a developmental abnormality.
These 2 groups were initially assessed separately as they have been shown to have different patterns of lesions on neonatal MRI.
Infants in group 1 tend to have a more global insult with a high incidence of basal ganglia lesions.1,4,23,24 Infants in group 2 do not have evidence for asphyxia and tend to have focal lesions in the form of arterial or parasagittal stroke,3,5,24 the imaging appearances of which are consistent with a perinatal injury. However, the groupings are not mutually exclusive but may represent 2 ends of the broad spectrum of perinatal brain injury.
Fifteen infants with normal brain imaging and normal neurologic examination25 were scanned. These infants were enrolled as part of an ongoing study on early brain development. None of these infants had needed resuscitation at birth, all had normal Apgar scores, and none had had seizures or other clinical neurologic symptoms.
All symptomatic infants underwent MRI, the majority within the first 2 weeks after delivery, as part of a detailed clinical neurologic evaluation conducted in our unit. Infants who had very early scans and survived were usually scanned later during the neonatal period to confirm the presence and demonstrate the evolution of lesions. Infants were examined during natural sleep or after a feed or were sedated with oral chloral hydrate (30–50 mg/kg). Infants wore ear protection in the form of individually molded earplugs made from dental putty and neonatal ear protectors (Natus Mini Muffs, Carlos, CA). Infants were monitored with pulse oximetry and electrocardiogram, and a pediatrician was present throughout the scan. The same procedure was followed for the control infants except that none was sedated for the scan.
MRI was performed using a 1.5 Tesla Philips Eclipse Scanner. Conventional images were obtained with T1 (SE 500/15) and fast T2 (FSE4200/210) weighted spin echo sequences (both with a 192 × 256 matrix, 5-mm slice thickness, and 2 averages). DWI was acquired using single-shot echoplanar imaging at multiple levels. Fifteen slices of 5-mm thickness were obtained (repetition time 6000 ms, echo time 110 ms, field of view 24 cm, b values of 0 and 1000 mm2/second) in 3 orthogonal directions. The total acquisition time was 37 seconds.
All conventional images were assessed for the presence of normal anatomy and abnormal signal intensities (SIs) separately and before diffusion analysis.
The PLIC, which usually shows a small linear region of short T1 and short T2 corresponding to myelin within the posterior half, was described as normal, abnormal, or equivocal.4 Equivocal SI was usually asymmetrical or had too much or too little SI for age. An abnormal PLIC showed loss of the normal SI from myelin, most easily seen on T1-weighted images (Figs 1a and 2a) . An abnormal SI with the posterior limb in a term infant is an excellent predictor of neuromotor impairment.4
Infants with abnormal SI within the BGT were also graded with a BGT score as mild, moderate, or severe, according to our previously described criteria.6 Mild lesions are focal SI abnormalities, may be transient, and are associated with a normal appearance to the myelination in the PLIC. Moderate lesions are multifocal abnormalities and usually associated with an equivocal or abnormal appearance to the PLIC. Severe lesions involve the entire BGT and an abnormal appearance to the PLIC (Figs 1a and b and 2a and b).
Infants with abnormal SI within the WM were graded as moderate or severe.6 Moderate abnormalities included the presence of areas of increased T1 and T2 and/or small hemorrhagic lesions. Severe changes consisted of areas of overt infarction whether unilateral or bilateral. This is seen as a loss of gray matter/WM differentiation on early scans or an exaggeration of gray matter/WM differentiation on second week scans.
The images were assessed for the presence of a normal cortical configuration, loss of cortical markings as seen with focal infarction, and the presence of abnormal high SI on T1- weighted images, so called “cortical highlighting” (CH; Fig 1a)
Analysis of DWI
The DWI obtained in each of the 3 planes and the combined trace ADC map were obtained using the following equation
where S = signal in the DW image, S0 = signal in the reference image, and b is given by the following equation
where γ = gyromagnetic ratio for protons, G = amplitude of the pulsed gradient, δ = duration of the pulsed gradient, and Δ = time interval between the leading edges of the two pulsed gradients.26
Analysis of the DWI was undertaken by 1 experienced researcher (M.R.) using Philips proprietary software. Diffusion-weighted and ADC trace maps were visually assessed for abnormal high SI or abnormal low SI, respectively, consistent with restricted diffusion and therefore probable tissue infarction (Figs 1c and 2c). The trace maps were used to produce ADC values within all regions of interest (ROIs). ROIs were identified on reference images and transferred to trace maps. Circular ROIs were used for all areas except the PLIC.
ROIs included WM in the centrum semiovale (CMO), anterior and posterior WM at the level of the BGT, central sulcus, lentiform nucleus, medial and lateral thalamus, PLIC, brainstem, and cerebellum. At least 2 measurements were taken for each ROI, and the average was used in the results. Measurements were made for both right and left sides of the brain. When these values differed by <0.1, the average of the 2 was used.
Comparisons were made across lesion groups and with the age at scan by dividing infants into those who were scanned during the first week and those who were scanned after the first week. This division was undertaken because of the known process of pseudonormalization of ADC values that occurs at ∼7 days15,16
Intraobserver error was measured using κ statistics. A value of .62 was obtained consistent with good intraobserver agreement. All data were checked for normality using the Shapiro Wilkes test. Comparisons between groups were then made using an unpaired t test for normal distributed data or the Kruskal Wallis test for nonnormally distributed data, with corrections for multiple comparisons.
The 15 control term-born infants had a median GA at delivery of 38.5 weeks (range: 36–43) and were scanned at a median of 6 days (range: 1–38) after delivery. Nine infants were imaged in week 1, and 6 were imaged after week 1.
A total of 63 patients were scanned at a median of 6 days (range: 1–26) after delivery. In group 1, 49 infants presented with NE. These infants had a median GA at delivery of 40 weeks (range: 36–42.5) and a median age at time of scan of 5 days (range: 1–26). Twenty-nine were imaged in week 1, and 34 were imaged after the first week.
In group 2, 14 infants presented with isolated seizures. These infants had a median GA at delivery of 40 weeks (range: 37–41) and were scanned at a median of 10 days (range: 4–24). Five were imaged in week 1, and 9 were imaged after the first week. There was no significant difference in GA between the 2 clinical groups and the control subjects.
The 15 control infants had conventional imaging that was within normal limits for age.
Of the 49 infants who presented with NE, 6 had normal BGT and 43 had abnormal BGT. Of the 6 with normal BGT, 2 had parasagittal WM infarction; 1 had punctate WM hemorrhages; and 3 had some long T1, long T2, in the WM only. The 6 infants with normal BGT were imaged between 3 and 18 days when abnormalities in BGT should have been obvious if present.
Of the 43 infants with abnormal BGT, all had bilateral abnormality. In addition, 7 had minimal WM or minimal CH only, 8 had more marked long T1 and long T2 in the WM, 8 had CH with some abnormal SI in the subcortical WM, 2 had focal hemorrhagic lesions, and 18 had loss of gray matter/WM differentiation consistent with overt infarction. Therefore, of the encephalopathic group, 88% had BGT lesions. Nearly one half of the infants with NE and BGT lesions had significant WM damage in addition.
Of the 14 infants who presented with isolated seizures, 2 had normal images, 6 had focal infarction in an arterial territory distribution, 2 had bilateral parasagittal WM infarction, and 4 had moderate bilateral WM abnormalities. Four of the 6 infants with focal infarction had unilateral abnormal SI within the BGT, and 1 of the infants with bilateral WM infarction had a unilateral BGT abnormality. Focal infarction was the main diagnosis in this group. All conventional imaging abnormalities were consistent with a perinatal onset for the infarction. None of the infants had bilateral BGT abnormality, and none had evidence of antenatal damage to the brain.
Analysis of the DWI and ADC Values
The analysis of the DWI and measurement of the ADC values was performed according to the pattern of lesions on conventional imaging. The 2 clinical groups therefore have been combined.
Visual Analysis of DWI
A visual analysis was done on the DWI and the ADC trace images. The appearances in the control infants were normal. All infants with focal WM infarction on conventional imaging had visually abnormal DWI and ADC trace images. The SI abnormalities were variable in their appearance even within 1 area of infarction. The ADC trace images identified more abnormalities than the DWI. As expected, some infants without obvious signs of early WM infarction on conventional imaging had abnormal DWI.
In infants with widespread bilateral WM and BGT changes, it was difficult to interpret visually the DWI or ADC trace maps, as there was no normal tissue for comparison (Fig 2c). When compared with a normal set of images, it became obvious that the entire image was abnormal. BGT lesions that were visualized on conventional images were not always visible even on the trace images, particularly when small.
The ADC values in all measured brain regions in the control infants are shown in Table 1 and in Fig 3. Of the 15 control infants, 9 were imaged during the first week after delivery and 6 were imaged after the first week. The values for ADCs in the BGT were constant during this period. There was a trend for ADC values in the WM to decrease after the first week, but this did not reach statistical significance. There was no correlation between GA at birth and ADC values in the 3 different WM regions in these control infants. The ADC values for all patients are shown in Table 1.
ADC Values in Infants With BGT Lesions
To compare ADC values with lesions identified on conventional imaging, we combined the 2 patient groups. Scatter plots (Figs 4–6) showed a wide range of ADC values with increasing postnatal age in the patient groups, confirming the need to divide patients into those with early scans (<7 postnatal days) and those with later scans (Table 2). Twenty-nine infants had normal BGT (15 control subjects and 14 patients), 22 had moderate BGT lesions, and 27 had severe BGT lesions.
Infants with severe BGT lesions had significantly reduced ADC values in the thalami (median: 0.77; range: 0.5–1.27; P = .0004) and in the lentiform nuclei (median ADC: 0.78; range: 0.5–1.35; P < .0001) during the first week when compared with control subjects. These values “pseudonormalized” during week 2 and then increased to a median of 1.23 (range: 0.9–1.4) in the thalami and a median of 1.23 (range: 0.75–1.5) in the lentiform nuclei. This increase reached significance when compared with week 1 (P = .016) and with control values (P < .001) in the thalami but not in the lentiform nuclei (P = .63; Fig 4). In some infants with BGT lesions and ADC values that had pseudonormalized, there was still obvious abnormal SI on trace images (Fig 1c).
In contrast, some infants with moderate BGT lesions had higher thalamic ADC values compared with control subjects during week 1, but this did not reach significance across the group median ADC (1.2; range: 0.84–1.4). There was an additional increase in week 2 with a median ADC 1 (0.95–1.4) resulting in a significant difference from control subjects (P = .034; Fig 2). ADC values in the lentiform nuclei in infants with moderate BGT were increased from control subjects, but this did not reach significance in week 1 (P = .25) or in week 2 (P = .31)
ADC values for the thalami were taken from both medial thalami and lateral thalami in the region of the ventrolateral nuclei (VLN) when readily visualized. In all 36 infants (patients and control subjects) in whom the VLN of the thalami was measurable, the ADC values were significantly lower than for more medial areas of the thalamus (P = .0003). There was no significant difference in the ADCs for the VLN in moderate BGT infants whether performed in week 1 or later, but for infants with severe BGT lesions, the VLN were significantly lower than in control subjects in the first week (P = .023) and then increased during the second week.
Thirty-two infants had a normal PLIC (15 control subjects and 17 patients), 12 had an equivocal PLIC, and 34 had an abnormal PLIC (bilaterally in 30 and unilaterally in 4). ADC values in infants with abnormal PLIC were significantly reduced (P < .0001) in the first week (median: 0.6; range: 0.48–0.9) when compared with the control subjects and increased to above normal during the second week (median: 1.2; range: 0.6–1.25), although this did not reach significance (P = .62; Fig 5). There was no significant difference between ADC values in PLICs that seemed equivocal as compared with those with a normal PLIC.
Infants with severe BGT lesions had reduced ADC values in the PLIC compared with those with normal BGT. The ADC values in the PLIC in those with moderate BGT lesions were not significantly different from normal.
Twenty-four infants had normal WM (15 control subjects and 9 patients), 27 had moderate WM abnormalities, and 27 had severe abnormalities on conventional imaging. In 18 of those with severe changes, these were combined with severe BGT lesions. The ADC values in infants with severe WM abnormalities were significantly reduced in week 1: CSO (median: 0.76; range: 0.5–1.55; P = .0004), anterior WM (median: 1; 0.6–1.75; P = .003), and posterior WM (median: 0.8; 1–1.65; P < .0001). During week 2, ADC values increased, but this did not reach significance within any of the WM areas (Fig 6) when compared with the control subjects. In infants with moderate WM abnormalities, the ADC values were not significantly different from control infants at any time in the CSO or anterior or posterior WM (Table 3).
ADC values measured within the crus of the mesencephalon were a median of 0.98 (range: 0.86–1.1) in the control subjects and a median of 0.92 × 10−3/mm2/second (range: 0.5–1.25) in the patients. This was not significantly different. However, in infants with BGT lesions, the brainstem ADC was a median of 0.85 × 10−3/mm2/second (range: 0.5–1.25), which was significantly different from control subjects (P = .028) and highly significant (P = .0004) when first week values alone were used (median: 0.78 × 10−3/mm2/second; range: 0.5–1.1).
There was no difference in ADC values within the cerebellar hemispheres between the control subjects (median: 1.1 × 10−3/mm2/second; range: 1–1.25) when compared with all patients (median: 1.1 × 10−3/mm2/second; range: 0.8–1.3) or just with infants with BGT lesions (median: 1.05 × 10−3/mm2/second; range: 0.8–1.25). In addition, there was no significant difference in ADC values within the cerebellar vermis in the control subjects (median: 0.97 × 10−3/mm2/second; range: 0.8–1.2) when compared with all patients (median: 0.98 × 10−3/mm2/second; range: 0.7–1.2) or just with the patients with BGT lesions (median: 0.97 × 10−3/mm2/second; range: 0.77–1.11).
Delayed Changes in ADC
In control infants, there was a decrease in ADC values within all WM areas with increasing age at time of scan, but this did not reach statistical significance. In infants with moderate or severe BGT lesions but apparently normal WM, the ADC values in all areas of WM were increased significantly in infants who were scanned after the first week when compared with those who were scanned in the first week (CSO, P = .019; anterior WM, P = .02; posterior WM, P = .006) and compared with week 2 control subjects (anterior WM, P = .025; and in posterior WM, P = .02; but not in the CSO, P = .13; Fig 4).
Predictive Abilities of ADC Values
Infants with very early MRI scans had follow-up scans when they survived. The exact pattern of brain lesions therefore could be confirmed. The best predictor for abnormal tissue, as identified on conventional imaging, was the ADC value within the medial thalamus. An ADC ≤0.8 or ≥1.2 × 10−3/mm2/second was 100% specific but only 80% sensitive for detecting severe BGT lesions on conventional imaging. This sensitivity increased to 85% when only values that were obtained in the first week after birth were used. When moderate BGT lesions are included as an outcome, the sensitivity of the thalamic ADCs falls to 60%, and this was not improved by taking only first-week values.
An ADC ≤1.1 × 10−3/mm2/second in the WM were always associated with infarction. However, within an area of infarction, the ADC values could be variable.
We obtained DWI in a large group of term-born neonates who presented with acute neonatal neurologic abnormality of hypoxic-ischemic origin and in 15 term born control infants. We compared the ADC values to the sites of lesion as detected on the conventional images and showed a significant reduction in ADC values in infants with severe BGT and severe WM lesions during the first 7 days after delivery. WM ADC values below 1.1 × 10−3/mm2/second were always associated with infarction often when the conventional images were not overtly abnormal. Thalamic ADC values below 0.8 × 10−3/mm2/second were always associated with severe BGT lesions. Infants with BGT lesions also showed significantly reduced ADC values within the mesencephalon.
In the presence of BGT lesions, the PLIC had similar diffusion characteristics to WM elsewhere in the brain when severely injured. This is in keeping with its appearances on conventional imaging and its role in predicting outcome.4 A reduction in the ADC values in the PLIC has been noted by other groups in infants with NE.19 This study has shown, therefore, that DWI can be used during the first week after birth to identify severe WM and severe BGT lesions, which are known to be associated with significant neurodevelopmental impairment.
However, ADC values in both severe BGT and severe WM increased after the first week from delivery and therefore cease to be valuable for prediction, but at this time, lesion pattern on conventional imaging becomes clearer. It is important to note that in infants with moderate BGT and with moderate WM abnormality on conventional imaging, the ADC values were not significantly different from control subjects and therefore are not as clinically useful.
We used conventional images to identify the sites of injury, and we tried to establish an additional, more objective method for establishing tissue abnormality. Once established, the pattern of abnormality can be used to predict outcome. In a few infants, the conventional scanning was performed early, a time when we have said that the brain appearances may seem relatively normal in the presence of infarction. In those cases with very early scans, we repeated the imaging to confirm the sites of injury when the child survived. In the few nonsurvivors without repeat imaging, clinical investigation with ultrasound, cerebral function monitoring, and neurologic examination all were in keeping with the severe injury suspected with conventional imaging.
Visual analysis of DWI in infants with widespread abnormality can be misleading as they lack normal tissue for comparison and the images may look unremarkable, although measurements of the ADC values confirm the presence of infarcted tissue. The cerebellum seemed to have been spared in all of our infants, and this may be used as a visual comparison, although this was not always reliable. The so-called white cerebrum on DWI with cerebellar sparing in the presence of widespread supratentorial involvement has been reported previously.27 Our data show that ADC values should always be measured in combination with visual analysis of both conventional and diffusion-weighed images for maximum detection of pathologic tissue, and the timing of the scan needs to be taken into account when interpreting the results.
In this study, we used an echoplanar imaging single-shot DWI technique to obtain 15 slices across the brain in 3 directions. The examination has the advantage of taking only 37 seconds and imaging the whole brain. This allows repeat acquisitions to be performed during an examination if infant motion results in poor-quality data. It was not possible to use diffusion tensor imaging techniques in this study, but these would be likely to improve the detection of abnormal tissue and fiber disruption. Serial measurement using diffusion tensor imaging would provide measures of fractional anisotropy and data for WM tracking, which could give detailed insights into the response of the immature brain to injury.21,28 In addition, diffusion sequences that use higher b values may enhance the detection of abnormal tissue, although their clinical role still needs to be explored.29,30
We were able to image only 15 control infants as it is always difficult to recruit newborn control infants. Other groups have published values for some areas of the brain in normal term-born neonates, and their results are comparable to our own.10,12 Both studies obtained DWI at 1.5 Tesla with 2 b values (0 and 1000 mm2/second). In Forbes's study of 40 children from birth to 1 year, 14 infants were term born, but all were scanned for clinical reasons and therefore were not true control subjects. Their ADC values for subcortical frontal WM were higher than ours (median: 1.88 vs 1.6 × 10−3/mm2/second, respectively), but for the PLIC, ADC values were comparable (median: 1.09 vs 1.0 × 10−3/mm2/second). Forbes et al also noted an increase in ADC values within anterior WM compared with posterior WM. The median ADC in our control subjects was 1.6 for anterior WM, which was not significantly different from the 1.55 for posterior WM. Variation in the site of the ROIs to be measured within the WM may explain some differences between studies. In Tanner's study, they measured ADC values in 10 term-born, neurologically normal neonates who were <43 days old, and their values for anterior WM were comparable to ours (1.62 vs 1.6 × 10−3/mm2/second).
Although there was some spread of the WM values in our ADC data for normal term-born control subjects, there was no relation to exact GA at birth or postnatal age at scanning. The variation within WM in our group may be related to developmental variation or to specific changes occurring with respect to labor and delivery. We have noticed a similar variation in the WM appearances on visual analysis of conventional imaging. We believe that our small group of control subjects represent the normal values and range for this age group and are not surprised that there is some variation. We accept that it would be desirable to obtain larger numbers of control subjects, but this is difficult to achieve.
We made the assumption that the cause of the symptoms and imaging findings in all infants was hypoxic-ischemic, as overt sepsis, recognized metabolic disorders, congenital infections, or malformations all were excluded. Human neonatal studies have shown that an acute global perinatal insult in the context of fetal distress and acidosis results in central gray matter lesions, and chronic or partial repetitive insults may result in WM injury.2,3,5,31–35 Although the exact timing of either a single injury or repeated injuries cannot usually be determined, conventional imaging findings were consistent with a perinatal onset of the lesions in all of our infants.23 Widespread WM infarction in the term infant is often thought to be evidence of a chronic or repeated insult, but this and previous imaging studies confirm that the process of infarction does not start until the time of labor and delivery, suggesting that a degree of asphyxia or trauma or possibly even a “normal labor and delivery” is necessary to establish infarction in “primed” tissue. One half of the infants in this study with severe BGT lesions also had severe or moderate WM lesions, suggesting a mixed cause.
We documented a reduction in ADC in acute WM infarction with a “pseudonormalization”: and increased values after 7 days of age. This pattern of ADC has been well documented with adult stroke patients, although our “normalization” may occur slightly earlier. This evolution has been reported by others after perinatal stroke.16
Of interest is that in our moderate WM group, ADC values were within the normal range on the earlier scans but higher after 7 days, although this difference did not reach significance. It is unlikely that these infants have overt WM infarction. Increased ADC values could result from extracellular edema, but after perinatal asphyxia, edema is more obvious in the first few days and has decreased by the end of the first week.1 Moderate WM abnormality on conventional imaging may result in later atrophy of the tissue and may represent a combination of tissue responses that result in an overall normal or increased ADC. Apoptosis has been associated with normal DWI36 and with an increase in ADC values.37 More sophisticated diffusion tensor imaging may allow improved assessment of these moderate abnormalities. The difficulty in detecting or confirming moderate WM lesions from the ADC values during the first week is less concerning than the inability to detect moderate BGT as the outcome for moderate WM lesions may be surprisingly good.
The outcome for infants with WM lesions is determined largely by any additional involvement of central gray matter. The outcome with moderate WM infarction in the presence of normal BGT may be good. Severe widespread WM infarction gives rise to a secondary microcephaly, cognitive and behavioral impairment, and sometimes a mild spastic diplegia.8,38 This illustrates the need to combine the degree of ADC abnormality with the extent and the sites involved to predict outcome.
We noted previously that in infants with moderate or severe BGT lesions but initially normal WM, there is a progressive increase in T1 and T2 of the WM on conventional imaging over the first month after birth and that this WM subsequently atrophies. In marked cases, the child develops a secondary microcephaly.6 In this study, we documented that the WM ADC values in this group of infants increased with postnatal age compared with a relative decrease seen in normal control subjects. This injury may represent small foci of necrosis within the WM, which we have been unable to detect with our diffusion technique. The late increase in ADC values in the WM could occur as a consequence of the BGT damage. This may be a form of Wallerian degeneration, which has been reported as showing reduced ADC in the acute stages,39 which we were unable to detect. It is possible that this observation represents a late effect on oligodendrocytes possibly mediated via the axons. Once again, if oligodendrocytes were dying by necrosis, then one might expect a reduced ADC. Increased ADC values may be the hallmark of apoptosis, during which cells shrink and extracellular water increases, which would result in an increased ADC value.37 Delayed apoptotic death of oligodendroglial cells has been reported after excitotoxic lesions in rat thalamus, and it was postulated that this death resulted from toxic microglia/macrophage effects.40 This suggests a secondary or delayed injury in WM as a consequence of the initial central gray matter insult and may represent a window of opportunity for interventions in infants with NE. Infants with isolated basal ganglia lesions develop a motor impairment, and those with associated WM atrophy, perhaps reflecting the severity of the central gray matter lesions, develop additional cognitive impairment.6 If the secondary WM atrophy could be prevented, then perhaps with antiapoptotic measures, it may be possible to improve the cognitive outcome in these children.
The evolution of the ADC values in severe BGT injury is similar to the evolution for WM injury seen in adult stroke studies. Pseudonormalization occurs ∼7 days from the injury, but, fortunately, by 1 week of age, conventional imaging will clearly show abnormal SI, lessening the difficulty in image interpretation. These data show that DWI is useful in the first few days after a perinatal ischemic insult to the BGT, reduced values being consistent with severe lesions. Levels of ≤0.8 × 10−3/mm2/second were associated with severe BGT lesions with a sensitivity of 85%. The remaining 15% may represent lesions that occurred at a slightly earlier time before delivery or that are evolving differently. However, severe BGT lesions were associated with less reduction in ADC than severe WM lesions. This difference in ADC reduction between gray matter and WM has been noted by others.41
The normal values for ADC in the infants with moderate lesions within the BGT has been documented previously,16 coupled with the fact that these lesions may also have normal visual appearances on DWI. These lesions may not show a reduced ADC because of their relatively small size. If the lesions are surrounded by some extracellular edema, then the net effect may be to have a relatively normal ADC within the ROI. These lesions may arise from different mechanisms; if apoptotic cell death predominates, then the ADC value may show a relative increase.37,42 The lack of detection of these moderate BGT lesions by DWI is worrying as these infants are likely to have significant neurodevelopmental impairments. Conventional MRI toward the end of the first week remains the best imaging method for detecting these lesions.
We measured ADC values in the cerebellar hemispheres and within the vermis and found them to be comparable with control subjects. This is additional evidence that the cerebellum seems to be resistant to perinatal hypoxic-ischemic injury. Although animal studies show a high incidence of injury to the dentate nuclei, it is unusual to detect any visual abnormalities on conventional MRI.43 We measured our ADC values within the cerebellar hemispheres but not within the dentate nuclei, which are not always well visualized with the slices that we obtained. It is possible that more sophisticated diffusion studies might detect acute injury in the dentate nuclei. We have previously shown that cerebellar growth is reduced after perinatally acquired BGT lesions and that this is most marked within the vermis.44 It is likely that reduced cerebellar growth occurs as a secondary phenomenon in infants with perinatally acquired BGT lesions.
A continuing difficulty with any study of human neonates is the inability to time the onset, duration, or precise nature of the injury to the brain. In rare cases, the event is obvious (eg, uterine rupture, placental abruption), but these reports still assume that the fetus was completely normal before this sentinel event. Although conventional imaging patterns in the first days after delivery are highly suggestive of the perinatal timing of an insult, we still have not established whether ADC values will provide data for timing the onset of injury more accurately than is currently possible. Unfortunately, in medicolegal cases, what is required often is timing to within fractions of an hour of an action or an event, and it is probable that we will not be able to achieve those levels of accuracy. This data show that DWI is a powerful and clinically useful technique for assessing the neonatal brain after suspected perinatal hypoxic-ischemic injury. A larger serial study, performing more sophisticated diffusion tensor studies on several occasions during the first 2 weeks after delivery, may allow the issue of timing and evolution of perinatal brain injury to be addressed in more detail.
The Medical Research Council, the Academy of Medical Sciences, the Health Foundation, and Philips Medical Systems supported this study.
- ↵Barkovich AJ. MR and CT evaluation of profound neonatal and infantile asphyxia. AJNR Am J Neuroradiol.1992;13 :959– 972
- ↵Mercuri E, Cowan F, Rutherford M, Acolet D, Pennock J, Dubowitz L. Ischaemic and haemorrhagic brain lesions in newborns with seizures and normal Apgar scores. Arch Dis Child Fetal Neonatal Ed.1995;73 :F67– F74
- ↵Rutherford MA, Pennock JM, Counsell SJ, et al. Abnormal magnetic resonance signal in the internal capsule predicts poor neurodevelopmental outcome in infants with hypoxic-ischemic encephalopathy. Pediatrics.1998;102 :323– 328
- ↵Mercuri E, Ricci D, Cowan F, et al. Head growth in infants with hypoxic-ischaemic encephalopathy: correlation with neonatal magnetic resonance imaging. Pediatrics.2000;106 :235– 243
- ↵Cowan F, Dubowitz L, Mercuri E, Counsell S, Rutherford M. White matter injury can lead to cognitive without major motor deficits following perinatal asphyxia and early encephalopathy. Dev Med Child Neurol Suppl.2003;93 :45:14
- ↵Mercuri E, Rutherford M, Cowan F, et al. Early prognostic indicators of outcome in infants with neonatal cerebral infarction: a clinical, electroencephalogram, and magnetic resonance imaging study. Pediatrics.1999;103 :39– 46
- ↵Krishnamoorthy KS, Soman TB, Takeoka M, Schaefer PW. Diffusion-weighted imaging in neonatal cerebral infarction: clinical utility and follow-up. J Child Neurol.2000;15 :592– 602
- ↵Mader I, Schoning M, Klose U, Kuker W. Neonatal cerebral infarction diagnosed by diffusion-weighted MRI: pseudonormalization occurs early. Stroke.2002;33 :1142– 1145
- ↵Forbes KP, Pipe JG, Bird R. Neonatal hypoxic-ischemic encephalopathy: detection with diffusion-weighted MR imaging. AJNR Am J Neuroradiol.2000;21 :1490– 1496
- ↵McKinstry RC, Miller JH, Snyder AZ, et al. A prospective, longitudinal diffusion tensor imaging study of brain injury in newborns. Neurology.2002;59 :824– 833
- ↵Barkovich AJ, Westmark KD, Bedi HS, Partridge JC, Ferriero DM, Vigneron DB. Proton spectroscopy and diffusion imaging on the first day of life after perinatal asphyxia: preliminary report. AJNR Am J Neuroradiol.2001;22 :1786– 1794
- ↵D'Arceuil H, Rhine W, de Crespigny A, Yenari M, et al. 99mTc annexin V imaging of neonatal hypoxic brain injury. Stroke.2000;31 :2692– 2700
- ↵Mazumdar A, Mukherjee P, Miller JH, Malde H, McKinstry RC. Diffusion-weighted imaging of acute corticospinal tract injury preceding Wallerian degeneration in the maturing human brain. AJNR Am J Neuroradiol.2003;24 :1057– 1066
- ↵Mukherjee P, Miller JH, Shimony JS, et al. Diffusion-tensor MR imaging of gray and white matter development during normal human brain maturation. AJNR Am J Neuroradiol.2002;23 :1445– 1456
- ↵Jouvet P, Cowan F, Cox P, et al. Reproducibility and accuracy of magnetic resonance imaging studies of the brain after birth asphyxia. AJNR Am J Neuroradiol.1999;20 :1343– 1348
- ↵Le Strange E, Saeed N, Cowan F, Edwards D, Rutherford M. Magnetic resonance quantification of the cerebellum following hypoxic ischaemic injury to the neonatal brain. AJNR Am J Neuroradiol.2004;25 :463– 468
- Copyright © 2004 by the American Academy of Pediatrics