Applied Radiology

Anoxic-ischemic injury is a major cause of morbidity in children, as thosewho survive the initial event are frequently left with residual long-termneurological sequelae. As sophisticated cross-sectional imaging techniques havebeen developed to optimize the early visualization of anoxic-ischemic injury,the radiologist has become more actively involved in the initial evaluation ofthis group of patients. Early manifestations vary in children of differentages; therefore, it is important for the radiologist to recognize the spectrumof findings that can be seen.

In the future, neuronal protective agents may become routinely used to minimizethe amount of damage to the brain following anoxic-ischemic injury inchildhood. When this occurs, imaging will play key roles, not only todemonstrate the presence and location of such injury, but also to monitor theeffects of the interventional agents that have been administered to reduce itsextent.

The mechanisms underlying the development of brain damage from anoxia/ischemiaare complex.1,2 Following a global insult, though initially the whole brain issubjected to the insult, not all areas of the brain are equally injured.Depending on certain factors, such as the nature and duration of the insult andthe level of maturity of the brain at the time of injury, different anddistinct patterns of damage can develop.3,4 For instance, the relative maturityof the brain at the time of the insult will determine the location of watershedinfarction, and whether hemorrhage and subsequent gliosis will occur.5

Term infant

Early imaging findings seen in the term brain following an anoxic-ischemicevent include the presence of focal or diffuse edema (with or withoutinfarction), basal ganglia signal intensity changes, and laminar necrosis.4Hemorrhage also may be identified within areas of infarction.

Because of the inherent high signal intensity on T2-weighted images (T2WI) inthe newborn brain, subtle increases of water content associated with areas ofedema or infarction may be difficult to detect. When focal or diffuse edema ispresent, MR images may demonstrate findings related to mass effect, such asobscuration of sulci and gyri, and the presence of slit-like ventricles. Themost sensitive sign, however, is the loss of gray-white differentiation whichis normally seen on T2WI.

In the newborn period, the signal intensity of the cortical ribbon onT2-weighted images is normally lower than that of the underlying white matter.When cortical edema is present, the signal intensity of the gray matter becomeshigher on T2-weighted images, obscuring its differentiation with the whitematter (figure 1). Cytotoxic (ischemic) edema may be more easily appreciatedusing diffusion-weighted MR imaging sequences than routine T2-weighted spinecho sequences (figure 2). In addition to being more sensitive to the presenceof edema, such sequences also better demonstrate the full extent of the edema.Following episodes of hypoxia, peripherally located wedge-shaped areas ofinfarction may be identified in the parasagittal watershed zones (figures1,2,4,).3,7 In the early stages, such lesions have low signal intensity onT1-weighted images and high signal intensity on T2-weighted images.

Following severe insults, characteristic basal ganglia changes which evolveover time may be seen.3,4,8-12 The earliest finding (occurring as early as thefirst day after an anoxic-ischemic event) is of diffuse high signal intensityin the basal ganglia on T1-weighted images (figure 3); this may be striking orquite subtle. The same areas initially appear normal on T2-weighted images. Inorder not to miss this finding on T1-weighted images, it is helpful to rememberthat in normal infants the signal intensity of the posterior limb of theinternal capsule is always higher than that of the adjacent lentiform nucleusand thalamus. It is not known for certain what this abnormalT1-hypersensitivity represents; possibilities include hemorrhage, calcium,myelin breakdown products, free fatty acids, or free radicals.3,4 This earlypattern of basal ganglia T1-hyperintensity may be seen for up to 7 to 10 daysfollowing the insult. After this time, it transitions into an intermediatepattern of focal or patchy increased T1 signal intensity and decreased T2signal intensity (figure 3). After about day 17, a more "chronic"appearance may be present: T1-weighted images appear relatively normal; areasof gliosis or cystic necrosis may be seen on T2-weighted images. These areas ofcystic necrosis also may be seen on CT images, which are obviously lesssensitive for demonstrating the non-cystic changes and may frequently benegative in such cases.

Cortical laminar necrosis also may be seen following an anoxic-ischemic event.It is first identified on T1-weighted images as gyriform or curvilinear highsignal intensity in the deeper layers of the cortex, and is particularlyprominent at the bases of sulci (figure 4). On follow-up studies, low signalintensity may be seen on T2-weighted sequences in the same areas.13 Thisdistribution of injury is thought to be due to the more precarious corticalblood supply at the base of the sulcus, making it particularly vulnerable toanoxic-ischemic injury.

Infants demonstrating these early findings are at high risk for the developmentof long-term sequelae and should be followed closely. It should be emphasizedthat the demonstration of focal or diffuse edema or of early basal gangliachanges following an insult does not necessarily correlate with poor outcome.In some instances the edema seen on early images may be reversible, and infantsdemonstrating this finding can go on to develop into neurologically normalchildren.4 Poor outcome is much more likely, however, in infants who developthe intermediate type of basal ganglia changes and/or show evidence of laminarnecrosis on follow-up studies.4

Preterm infant

Imaging findings seen in the preterm brain following an anoxic-ischemicevent differ from those seen in the term brain and include germinal matrixhemorrhage, periventricular venous infarction, and periventricularleukomalacia.14 Hemorrhage and edema may be identified in the first weekfollowing insult. Because of the risks associated with moving premature infantsout of the neonatal intensive care unit, ultrasonography usually is the initialimaging technique used to evaluate such infants. This modality is particularlyuseful in demonstrating and screening for hemorrhage adjacent to or within theventricular system.14

Germinal matrix-intraventricular hemorrhage (GMIH) is the most common type ofhemorrhage found in the premature infant. It is found in 35 to 55% of infantsof less than 32 weeks gestation and less than 1500 gm in weight; 90% of thesehemorrhages occur during the first week of life. By contrast, GMIH is rarelyseen in term infants. The vessels of the germinal matrix, which are thin-walledand lack connective tissue, are vulnerable to hemorrhage when fluctuations inarterial pressure occur. This type of hemorrhage may be seen in associationwith many varied conditions that occur during infancy including respiratorydistress syndrome, pneumothorax, patent ductus arteriosus, noxious stimuli, andseizures.

When hemorrhage occurs, it destroys the germinal matrix and may burst throughthe subependymal layer into the lateral ventricles. Blood products may then mixwith CSF and pass through the foramina of Lushka and Magendie into thesubarachnoid space. Obstructive or communicating forms of hydrocephalus candevelop if the blood obstructs the flow of CSF through the aqueduct or preventsits absorption by the arachnoid villi.15

On imaging studies, germinal matrix hemorrhage usually exhibits mass effect. Asit resolves, it undergoes liquifaction and eventual cyst formation. It is seenadjacent to the head of the caudate nucleus in the floor of the lateralventricles (grade I). When hemorrhage is present within the ventricles (gradesII and III), the blood clots fill all or part of the ventricle and may alsodistend it. Serial imaging studies should be performed in infants with knowngerminal matrix intraventricular hemorrhage in order to detect the developmentof hydrocephalus.

Grade IV hemorrhages are now known to occur by a different mechanism thangrades I through III. Although previously thought to be extensions into theadjacent parenchyma of germinal matrix-intraventricular hemorrhage, these arenow known to represent periventricular hemorrhagic venous infarctions.14 Volpebelieves grade IV hemorrhages develop when germinal matrix-intraventricularhemorrhage causes compression of the terminal vein as it passes through thesubependymal region of the caudate nucleus, leading to venous infarction.14,15Infants with periventricular venous infarction have a much poorer prognosisthan those with type I, type II, or type III GMIV hemorrhage: 90% develop majorlong-term neurological sequelae compared to only 30 to 40% of infants withgrade III hemorrhage.

Hemorrhage in any location may be easily identified using magnetic resonanceimaging (figures 5,6).16 The presence of peripherally located parenchymal orextra-axial hemorrhage, and of posterior fossa hemorrhage may be seen on MRIdespite negative ultrasound and CT studies. Hemorrhage within the ventricularsystem also can be detected for a longer period of time. On FLAIR sequences,the normally low CSF signal intensity is altered due to the T1-shorteningeffects of protein in the serum and blood breakdown products. Later,hemosiderin may be seen staining the ependyma (figure 5).

The age of the hemorrhage and its location and extent can be assessed usingroutine spin-echo T1- and T2-weighted images. In the newborn infant, the signalintensity of parenchymal hematomas follows similar sequential changes to thosedescribed in adult patients despite the presence of fetal hemoglobin in thisage group, which has a stronger oxygen affinity. When there is a strongclinical suspicion that hemorrhage has occurred but none is identified onroutine sequences, increased sensitivity may be obtained using gradient- echoT2*-weighted images.

Periventricular leukomalacia (PVL) is seen primarily in ventilator-dependantpremature infants who survive more than a few days. Although it is generallythought to be due to a combination of lack of cerebrovascular autoregulationand systemic hypotension in an acutely ill premature infant, its mechanism isnot well understood. The most common locations of PVL are in theperiventricular white matter, at the trigone of the lateral ventricles, andadjacent to the foramen of Monro.

Pathologically, these lesions are characterized by areas of coagulationnecrosis with subsequent macrophage activity, liquifaction, and cyst formation.By 3 to 4 weeks, the cystic cavities frequently have coalesced and communicatedwith the ventricles. This leads to a reduction in white matter volume andenlargement of the ventricles. Hemorrhage is reported to occur in 25% of casesof PVL.

Early periventricular leukomalacia may be more easily seen using MRI.17Initially, areas of hemorrhage are seen in the periventricular white matterparalleling the borders of the lateral ventricles (figure 6). Later, as thehemorrhage resorbs, these areas become cystic. Serial MR studies maydramatically demonstrate incorporation of these lesions into the lateralventricles, which subsequently enlarge and develop the classic"ragged" borders (figure 6).

When preterm infants suffer catastrophic anoxic-ischemic events, either in theperinatal period or in utero, they may demonstrate basal ganglia changessimilar to those described above in term infants. The extent of basal gangliainvolvement usually is less than that seen in the term infant, with the mostsevere damage occurring in the thalami and the posterior aspects of themidbrain and brainstem.3,4,18 The relatively decreased area of involvement inthis group is due to the fact that the metabolically active areas of the brainand those areas that are actively myelinating are less extensive in the preterminfant than at term. Because of this, the volume of brain most vulnerable toanoxic-ischemic damage is also smaller in preterm infants.

The older child

Older children who have suffered severe anoxic-ischemic insults demonstrateyet another pattern of injury.12,19 Typically, this type of injury is seenclinically in older children who have suffered a near drowning episode,cardiorespiratory arrest, an anesthesia accident, or attempted strangulation.Images obtained soon after the insult or in the subacute stage may show thepresence of edema, or signal intensity changes in the basal ganglia and cortex.Edema may be focal (frequently seen in the occipital lobes) or generalized,with relative sparing of the perirolandic cortex. Affected children demonstratea characteristic evolution of signal intensity changes in the basal gangliawhich differ from those seen in infants. Initially, T1-weighted images may showindistinctness of the borders of the basal ganglia ("fuzzy basalganglia") (figure 7). At the same time, T2-weighted images show highsignal intensity in the lentiform nuclei which progresses from involving onlythe periphery of the nuclei to appearing patchy, and later to involving theentire nucleus diffusely (figure 7).19 T2-weighted images also may showcortical changes of focal high signal intensity or the presence of high signalintensity subcortical lines (figure 8).3,19 Follow-up studies may show thepresence of diffuse atrophy, gliosis and, occasionally, iron deposition.4Although some reversibility may be seen in early focal edema and indistinctbasal ganglia margins, once high signal intensity changes have developed onT2-weighted images, reversibility is unlikely and the ultimate clinical outcomewill be poor.

Causes of focal infarction in the pediatric age group are diverse and includeembolus, thrombosis, vasospasm (e.g., migraine), vascular malformations,tumors, and metabolic causes.20 Embolus may be due to several causes, includingcyanotic heart disease (i.e., right to left shunt), cardiomyopathy, carotiddissection, and mitral valve prolapse. Thrombosis may form as a result ofpolycythemia, trauma leading to dissection, viral infection, meningitis,coagulopathies, maternal drug use (especially cocaine), and a host ofvasculopathies.

Focal infarction has a different presentation in newborns and infants than itdoes in older children and adults. Clinically, it may be much more difficult todetect at this age. In the term newborn, infarction may present as seizures, orit may present with nonspecific, non-focal findings such as hypotonia. Thedemonstration of hand preference in a child of less than a year can be thefirst indication of the presence of a unilateral hemiparesis due to a previousinfarction. Focal infarction is rare in premature infants, but it has beenreported. In this age group, seizures usually do not occur as a result of theinfarction. Older children, like adults, more frequently present with evidenceof a focal neurological deficit.

Although MRI is sensitive for the demonstration of infarction, the findings maybe subtle, especially in the neonatal period when the water content of thebrain is still very high. Areas of infarction may demonstrate focal masseffect, effacement of sulci and gyri, and loss of the distinct gray-whitematter border. When hemorrhage is present, the signal intensity of the involvedarea will change depending on the age of the hemorrhage. Recently, the use ofdiffusion MRI in the detection of acute stroke in infants has been described.In the acute situation, diffusion imaging demonstrates evidence of stroke thatcan not be seen or is difficult to identify on routine spin-echo images (figure9).

In older children, as in adults, focal infarction is easier to identify andmost frequently is seen as a wedge-shaped area of low signal intensity onT1-weighted images and high signal intensity on T2-weighted images in avascular distribution (figure 10). Infarction involving the basal ganglia mayshow heterogeneous signal intensity of the deep gray matter and results fromobstruction of the lenticulostriate and thalamoperforating vessels. Loss ofsignal void within intracranial vessels may be seen in the blood vesselssupplying the affected area, and this sign should be actively sought in allcases of suspected infarction. Cerebral angiography or MR angiography22,23 mayboth be useful in identifying the cause of infarction in children with no knownunderlying etiology. AR