Magnetic resonance imaging in neonatal hypoxic-ischemic brain injury
Liauw, L.
Citation
Liauw, L. (2009, March 19). Magnetic resonance imaging in neonatal hypoxic-ischemic brain injury. Retrieved from
https://hdl.handle.net/1887/13690
Version: Corrected Publisher’s Version
License: Licence agreement concerning inclusion of doctoral thesis in the Institutional Repository of the University of Leiden
Downloaded from: https://hdl.handle.net/1887/13690
Note: To cite this publication please use the final published version (if
applicable).
Chapter 9
Summary and conclusions
References in alphabetical order
Summary in Dutch (samenvatting in het Nederlands)
Curriculum vitae List of publications
Chapter 9
Summary and conclusions
Early in gestation the future cerebral hemispheres develop as outpouchings of the telencephalon from the foramina of Monro regions. Within the walls of these outpouchings the germinal matrices develop. Of the hemispheric poles, the occipital poles are the first to develop. The hemispheres start with smooth surfaces. Sulci develop in an orderly sequence: phylogenetically older sulci develop first and the frontobasal, frontopolar, and the anterior temporal regions develop later (97). Gyral development proceeds most rapidly in the region of sensorimotor and visual pathways where myelination can be observed early (97). Myelination progresses from caudal to cephalad and from dorsal to ventral (97). It starts at around eight weeks gestational age in the brain stem, at around 26-30 weeks gestational age in the cerebellum, and at around 26-30 weeks gestational age in the cerebral hemispheres (98). Myelination also progresses more rapidly in functional systems that are used in early life: the vestibular, acoustic, tactile, and proprioceptive systems (56,99,100). On MR images, myelination is associated with T1-shortening (high signal intensity on T1-weighted images) and T2-shortening (low signal intensity on T2-weighted images) (56,99,100). Increase in cholesterol and glycolipids in myelin correlates with T1-shortening in developing white matter.
Galactocerebrosides and water of cholesterol interacting with the surface of myelin bilaminar membranes and the increased amount of bound water and consequently decreased amount of free water are responsible for early T1-shortening (56,100).
T2-shortening may reflect decreasing proton density which is associated with chemical maturation of the myelin sheath (100,101). Several methods have been developed to assess myelination on MR images (100,102-104).
Rapidly maturating and myelinating brain regions have a high metabolic demand (6,16). These metabolically active regions (brainstem, deep grey matter, and peri- Rolandic region) are more susceptible to hypoxia-ischemia than areas with a lower metabolic demand (2,16). Developing brain tissue responds to a hypoxic-ischemic assault with anaerobic glycolysis, resulting in lactic acidosis (16). Oligodendrocyte progenitors are particularly susceptible to lactic acidosis. This susceptibility and the immature vascularisation of the white matter in preterm infants explains the remarkable vulnerability of the white matter and periventricular white matter injury in these patients after moderate hypoxia-ischemia (16,105-111). Injury of the periventricular white matter can be periventricular leucomalacia, both cystic and non-cystic (41), and a more diffuse form of white matter injury (28,46,64,112- 114). Diffuse white matter injury is not easily detected by cranial ultrasonography (45,46,131) but is recognized on MR imaging as signal intensity changes in the periventricular white matter and/or as small punctate lesions (46,112,115,116).
In preterm infants, white matter injury is the predominant hypoxic-ischemic injury pattern (117). In a large number of infants with a gestational younger than 30 weeks, diffuse and excessive high signal intensity (DEHSI) of the white matter on T2-weighted images is seen at term equivalent age (118). DEHSI is commonly associated with signs suggestive of cerebral atrophy, thus it was
pattern in preterm infants, injury to the basal ganglia and brain stem may be seen in these patients after profound hypoxia-ischemia (2,9,10,16,119). Another, frequently encountered form of brain injury in preterm infants is germinal matrix/
intraventricular hemorrhage, being the result of the metabolical high demand and high vacularity of the germinal matrix of the immature brain, often combined with hemodynamic instability and a defective coagulation in these immature patients (21,22). Due to involution of the germinal matrix later in gestation, matrix hemorrhage is unusual after 34 weeks (16).
Thus, the site, extent, and severity of neonatal hypoxic-ischemic brain injury depends, amongst others, on maturity and on severity of the hypoxic-ischemic incident (profound, mild, or moderate events) (1,2,6,10,16,23,119,120). In (near) term infants, grey matter injury is common in profound hypoxia-ischemia (2,16). Injury is seen in the lateral thalami, globus pallidus, posterior putamina, hippocampi, dorsal brainstem, and peri-Rolandic cortex (the sensorimotor cortex), often combined with cortical and/or white matter injury (2,16) (6,16,121). In less severe cases of hypoxia-ischemia in term infants, injury is mainly confined to the cortex and subcortical white matter in the intervascular boundary zones (2,16) (1,6,16,23,105,109,110,121), which are the most susceptible regions in full-term infants with moderate hypoxia-ischemia (6,16). Classically, this injury pattern has been attributed to the location of these intervascular boundary zones (105,109,110). However, this theory has also been disputed (122,123). In chronic repetitive hypoxia-ischemia in (near) term infants, predominantly cortical and white matter injury (classically located in the parasagittal boundary zones) is seen (2). Symptoms may be discordant with the extent of the brain injury seen with imaging (2). Possibly, repetitive antenatal incidents prime the white matter.
Thus, even after a relatively minor perinatal incident extensive brain damage may occur (2).
Imaging findings in hypoxic-ischemic brain injury
Using ultrasonography, so-called flares are regions with increased echogenicity located in the periventricular white matter. The echogenicity of flares is greater than or equal to that of the choroid plexus (4,16,38,39,41,124). In periventricular white matter injury, cranial ultrasonography can be normal in the first two days after the hypoxic-ischemic incident (4,16). Thereafter it may show hyperechoic changes in the periventricular regions (4,16), which may later evolve into cystic lesions (41,125). Formation of cavities may typically occur two to six weeks after injury and is due to liquefaction (4,16). Cavitation is recognized as localized anechoic or hypoechoic lesions. Although serial, high quality cranial ultrasonography may also demonstrate noncystic white matter injury (131), it seems to be less sensitive for noncavitary white matter injury than for cystic lesions (4,16,28,30,45,46). Cranial ultrasonography demonstrates hypoxic-ischemic injury of the deep grey matter as hyperechogenicity, mostly subtle during the first days after the incident, but more prominent several days after the hypoxic-ischemic incident (2,16,39). In severe hypoxic-ischemic brain injury, Duplex Doppler measurements of cerebral blood flow may demonstrate abnormal resistive indices. Watershed injury may be hard to detect with ultrasonography due to it’s location at the brain’s convexity (16).
Chapter 9
On MR images in young infants with hypoxic-ischemic brain injury, brain swelling, abnormal signal intensity in the brainstem, basal ganglia, thalamus, and/or PLIC, and restricted diffusion can be seen (2-15,30,126). Other findings can be loss of the grey/white matter differentiation, white matter signal intensity changes, and cortical highlighting (high signal intensity of the cortex on T1-weighted images due to laminar necrosis) (2). In cases with cortical infarctions, local loss of grey/white matter differentiation can be seen (2,16). Hypoxic-ischemic injury to grey matter is characteristically seen as high signal on T1-weighted images and variable signal intensity on T2-weighted images. Injury to white matter is characteristically seen as low signal intensity on T1-weighted images and high signal intensity on T2-weighted images. Conventional MR imaging may not be able to demonstrate abnormalities in the first few days after the hypoxic-ischemic incident (61-64,77,94,127-130). In end-stage white matter injury, wide ventricles with irregular outline, loss of white matter, and thinning of the corpus callosum, deep sulci, delayed myelination, and periventricular high signal intensity on T2- weighted images are seen (4,16,24,26). It may be difficult to detect pathology on neonatal brain MR images due to the high water content of the neonatal brain and the sometimes subtle findings of hypoxic-ischemic lesions. Furthermore, dynamic physiologic processes, such as the process of myelination reflected as reversed signal intensity patterns on T1- and T2-weighted images compared to older children and adults, should not be mistaken for (subtle) pathology.
This thesis
The main purpose of this thesis was to study MR imaging features of hypoxic- ischemic brain injury.
In Chapter 2 we studied the contribution of several MR pulse sequences (T1-, T2-weighted imaging, FLAIR, contrast enhanced imaging, and diffusion-weighted imaging) in detection of hypoxic-ischemic brain damage after perinatal asphyxia.
The study also assessed the influence of the length of the time interval between the hypoxic-ischemic event and the imaging examination on the contribution of these individual techniques. The individual MR sequences and techniques, performed within ten days after birth in 40 term born neonates, were retrospectively and separately evaluated by two investigators. Analysis was performed for the whole group and subsequently separately for a younger (infants imaged ≤ 4 days) and older group (infants imaged > 4 days after birth). Interobserver agreement was calculated for the individual MR techniques. Subsequently individual assessments were compared to the consensus reading of the complete MR imaging examinations, considered to be the golden standard. Last, it was evaluated which combination of pulse sequences performed best in visualizing presence of injury patterns. Related to the gold standard, T1-weighted imaging performed best in the younger and older groups for lesions in the basal ganglia, thalamus, and posterior limb of internal capsule (PLIC), and for punctate white matter lesions. For infarctions, diffusion-weighted imaging scored best in both age groups. For non-punctate more
ischemic brain injury. This was true for both age groups. Adding FLAIR and/or contrast- enhanced imaging to the combination of T1-, T2-, and diffusion-weighted imaging did not contribute to detection of hypoxic-ischemic brain damage.
In Chapter 3 we hypothesized that comparing signal intensity of brain structures on T1-weighted images enables differentiation of normal myelination from signal intensity increase due to hypoxic-ischemic brain damage. T1-weighted images obtained in 57 infants born after a gestational age of 35 weeks or more were retrospectively evaluated. Subjects were assigned to a patient group with perinatal hypoxic-ischemic encephalopathy grades 2 or 3 (n=23) (29) or a control group consisting of infants in whom for other reasons than hypoxic-ischemic encephalopathy MR imaging had been performed (n=34). In each subject a signal intensity score was assigned to 19 brain structures, which was based on comparisons (in pairs) with the other 18 structures. Two comparisons (PLIC versus corona radiata, and posterolateral putamen versus peri-Rolandic cortex) were best to distinguish between patients and controls and to predict absence or presence of hypoxic-ischemic encephalopathy. The comparison posterolateral putamen versus the PLIC was most discriminative to distinguish infants of the hypoxic- ischemic encephalopathy group from those of the control group (likelihood ratio test= 31.54, p <0.0001): A higher or equal signal intensity of the posterolateral putamen compared with the signal intensity of the PLIC was seen significantly more often in infants from the hypoxic-ischemic encephalopathy group than in infants from the control group. The pairwise comparison peri-Rolandic cortex versus corona radiata had the most additional discriminating power (p <0.0001).
Thus, we concluded that in (near) term infants signal intensity changes due to hypoxia-ischemia can be differentiated from changes due to normal myelination, comparing signal intensity of two pairs of brain structures on T1-weighted images.
Presence of hypoxic-ischemic brain injury is very likely if the signal intensity of the posterolaterale putamen is equal to or higher than the signal intensity of the PLIC. Absence of hypoxic-ischemic brain injury is very likely if the signal intensity of the PLIC is higher than the signal intensity of the posterolaterale putamen and if the signal intensity of the corona radiata is higher than the signal intensity of the peri-Rolandic cortex.
In Chapter 4 we assessed the diagnostic value of several morphological features on MR imaging to differentiate between ‘terminal zones’ (a normal variant due to delayed myelination of fibre tracts in the associative regions of the inferoposterior parietal cortex and the posterior temporal cortex) and hypoxic-ischemic white matter injury. Seventy-five brain MR examinations performed in subjects up to 20 years of age (range 0.65-16.01 years) showing increased signal intensity on T2-weighted images in the peritrigonal areas were retrospectively selected.
Aspect, location, extent, shape, and borders of signal intensity changes in the peritrigonal areas were compared between subjects with evidence of perinatal episodes of hypoxia-ischemia (n=28) and control subjects without a history of hypoxia-ischemia (n=47). Presence of Virchow Robin spaces, hypoxic-ischemic abnormalities, and local atrophy were also recorded. Very high signal intensity
Chapter 9
of the peritrigonal areas on FLAIR images (Odds Ratio 25, confidence interval:
5-142.9) and presence of local atrophy (Odds Ratio 14.3, confidence interval:
1.4-166.7) were the best independent predictors to discriminate between the two groups and to differentiate ‘terminal zones’ from pathological signal intensity changes in the peritrigonal areas.
In Chapter 5 we assessed whether the comparisons in pairs of signal intensities, as described in Chapter 3, enable prediction of outcome. Fifty-seven infants with neonatal hypoxic-ischemic encephalopathy grades 2 or 3 (29) and control subjects, consisting of infants in whom for other reasons than hypoxic-ischemic encephalopathy MR imaging has been performed, all of whom underwent neonatal MR imaging examinations, were retrospectively assigned to one of four outcome groups. Outcome was graded with a four-point classification system (normal, mildly abnormal, or definitely abnormal at age five years, and death). For statistical analysis the ‘definitely abnormal’ and ‘death’ categories were combined as adverse outcome, and the ‘normal’ and ‘mildly abnormal’ categories as favourable outcome.
We assessed which signal intensity comparison in pairs scored best for outcome prediction. Predictive values of that signal intensity comparison for outcome were calculated for the entire group (hypoxic-ischemic encephalopathy grades 2 or 3 and controls) and for hypoxic-ischemic encephalopathy grade 2 only, a group with highly variable outcome. The comparison PLIC versus posterolateral putamen scored best for outcome prediction (likelihood ratio test=37.05, p=0.001), with a positive predictive value for adverse outcome of 69% and a negative predictive value of 98%. The chance to have an adverse outcome was 69% if the signal intensity in posterolaterale putamen ≥ PLIC (positive predictive value), the chance to have a favourable outcome 98% if the signal intensity in PLIC > posterolaterale putamen (negative predictive value). In case of a favourable neonatal clinical classification (grades 1 or 2 according to Sarnat), the chance to have a favourable outcome was 100%. In children with hypoxic-ischemic encephalopathy grade 2 only, the chance to have an adverse outcome was 45% (positive predictive value) and the negative predictive value ‘0’%. Using the comparison signal intensity in posterolateral putamen ≥ PLIC in this group only, the positive predictive value for adverse outcome was 67% and negative predictive value 88%. This study shows that comparison of signal intensity in certain brain structures on conventional T1-weighted images is a valuable method to predict outcome in (near) term infants with hypoxic-ischemic encephalopathy. In infants with hypoxic-ischemic encephalopathy grade 2, prediction of outcome is improved by using signal intensity comparisons on T1-weighted images.
In Chapter 6 we assessed the predictive value for outcome of diffusion-weighted imaging and apparent diffusion coefficient (ADC) measurements in 24 full-term neonates who underwent MR imaging within ten days after birth because of perinatal aphyxia. The MR examinations were retrospectively evaluated for hypoxic- ischemic brain injury. ADC measurements were performed in 30 standardized
abnormal, and death). For statistical analysis the ‘definitely abnormal’ and ‘death’
categories were combined as adverse outcome, the ‘normal’ and ‘mildly abnormal’
categories as favourable outcome, and all categories except the ‘normal’ category as abnormal outcome. We found 1) no difference in outcome (range 2.08- 5.75 years, mean 3.75) between infants with and without visible abnormalities on diffusion-weighted imaging, and 2) no predictive value for outcome of ADC values in visibly abnormal regions on diffusion-weighted imaging. Of ADC values obtained in normal appearing brain tissue (displaying no visible abnormalities on conventional MR imaging, DWI, and ADC images), those in the basal ganglia and brain stem correlated with outcome (p respectively 0.03 and 0.006 for abnormal; p respectively 0.01 and 0.03 for adverse outcome). There was a negative correlation between ADC values in the basal ganglia and brain stem and outcome, meaning that low ADC values in the basal ganglia or brain stem correlated with abnormal/
adverse outcome and higher ADC with normal/favourable outcome. This was independent of patterns of hypoxic-ischemic brain injury on conventional MRI and diffusion-weighted imaging.
In Chapter 7 we studied the predictive value of electroencephalography (EEG) and two neuroimaging techniques (cranial ultrasonography and MR imaging) for outcome at two years of age in 23 (near) term infants with hypoxic-ischemic brain injury. EEG, cranial ultrasonography findings, and MR imaging findings were retrospectively studied. Outcome was graded with a four-point classification system (normal, moderately abnormal, severely abnormal at age two years, and death). For statistical analysis the ‘definitely abnormal’ and ‘death’ categories were combined as adverse outcome. Abnormal EEG background was most predictive of adverse outcome (positive predictive value of 0.88 and negative predictive value of 0.86). The predictive value increased if abnormal EEG background coexisted with presence of diffuse white and deep and/or cortical grey matter changes on ultrasonography and/ or MR imaging (for ultrasonography and MR imaging a positive predictive value of 1 ). Abnormal neuroimaging findings alone were also highly predictive of adverse outcome, especially abnormal signal intensity of the PLIC and diffuse cortical grey matter injury (for ultrasonography a positive predictive value of 1 and a negative predictive value of 0.50; for MR imaging a positive predictive value of 1 and a negative predictive value of 0.67). MR imaging showed deep grey matter changes more frequently than ultrasound. Severely abnormal neuroimaging findings were always associated with an abnormal EEG background pattern.
In Chapter 8 we assessed the predictive value of white matter abnormalities on sequential cranial ultrasonography performed at a young age for white matter abnormalities on MR imaging performed in 40 preterm infants. We also assessed the predictive values of sequential ultrasonography and MR imaging for outcome at two years’ corrected age. Mean postnatal age and corrected gestational age at MR imaging were, respectively, 34.4 (4-111) days and 33.2 (27.3-45.1) weeks.
The mean number of ultrasound scans performed in each infant during admission was seven (3-16). Abnormal ultrasonography was highly predictive of white matter changes on MR imaging. Severely abnormal white matter on ultrasonography
Chapter 9
and/or MR imaging was highly predictive of adverse outcome (negative predictive value of 1) and normal-mildly abnormal white matter on ultrasonography and/
or MR imaging of favourable outcome (positive predictive value of 0.88, negative predictive value of 0.64). Moderately abnormal white matter on ultrasonography and/or MR imaging was associated with variable outcome. MR imaging only slightly increased the predictive value of sequential ultrasonography. Thus, MR imaging performed at variable age in very preterm infants has limited additional value to sequential cranial ultrasonography for predicting outcome at two years’ corrected age.
Conclusion
To identify all cases of neonatal hypoxic-ischemic brain injury and for optimal outcome prediction, a combination of diagnostic modalities is mandatory. For the diagnosis and outcome prediction of hypoxic-ischemic brain injury in young infants, clinical assessment of the infant, neuroimaging (cranial ultrasonography and MR imaging), and EEG are complementary.
Specific conclusions
1) In neonates with perinatal asphyxia, the combination of conventional T1-, T2- and diffusion-weighted imaging is optimal to detect hypoxic-ischemic brain injury.
2) In (near) term infants comparison of signal intensity on T1-weighted images between the PLIC and posteroloateral putamen and between the corona radiata and peri-Rolandic cortex permits detection of hypoxic- ischemic brain injury.
3) In young subjects, distinction between ‘terminal zones’ and periventricular white matter injury is possible. If signal intensity in the peritrigonal areas is very high on FLAIR images and if local atrophy is present, presence of white matter injury is likely.
4) The signal intensity comparisons described in Chapter 2 can be used for prediction of outcome in young infants with hypoxic-ischemic brain injury. If the signal intensity of the posterolateral putamen exceeds or equals that of the PLIC on T1-weighted images, adverse outcome is very likely. If the signal intensity of the PLIC exceeds that of the posterolateral putamen, normal outcome is very likely. In infants with hypoxic-ischemic encephalopathy grade 2 outcome prediction is improved using these signal intensity comparisons on T1-weighted images.
5) In neonates with hypoxic-ischemic brain injury, low ADC values in the basal ganglia and brain stem are predictive of abnormal outcome.
6) In infants with hypoxic-ischemic brain injury, abnormal EEG background activity is highly predictive and more predictive than abnormal neuroimaging findings for abnormal outcome at two years of age. If abnormal EEG background coexists with abnormal signal intensity in the PLIC and with diffuse cortical abnormalities, predictive value further increases.
7) In preterm infants, sequential cranial ultrasonography at a young age reliably predicts white matter abnormalities on MR imaging. Severely abnormal white matter on ultrasonography and/or MR imaging is highly predictive of abnormal outcome at two years’ corrected age and normal/
mildly abnormal white matter of favourable outcome. Moderately abnormal white matter on ultrasonography and/or MR imaging is associated with variable outcome.
