Cerebrovascular Accidents in Neonates Treated with Extracorporeal Membrane Oxygenation: Sonographic-Pathologic Correlation

Transcription

1 355 Cerebrovascular Accidents in Neonates Treated with Extracorporeal Membrane Oxygenation: Sonographic-Pathologic Correlation George A. Taylor1 "3 Charles R. Fitz 1-2 Sudesh Kapur2 ' 4 Billie L. Short2-5 Neonates treated with extracorporeal membrane oxygenation are at high risk for the development of intracranial hemorrhage and infarction. The appearance of these lesions on cranial sonography is often unusual and may be confusing. We compared the findings at autopsy with premorbid cranial sonograms in 17 nonsurviving neonates to define better the anatomic basis for the sonographic appearance of these lesions. Macroscopic abnormalities were identified at autopsy in 13 of the 17 neonates. Five neonates had multifocal hemorrhagic white-matterinfarcts, three had large parenchyma! hemorrhages with adjacent areas of parenchyma! necrosis, three had hemorrhagic infarcts of the cerebellum, one had a germinal matrix and intraventricular hemorrhage, and one had bilateral periventricular cysts with surrounding gliosis. All 17 neonates had abnormalities on microscopic examination. Although sonography was accurate in the detection of macroscopic lesions (11 of 13 lesions detected with sonography), the nature and extent of these abnormalities were difficult to judge because of the variable echogenicity of unclotted blood and the presence of focal areas of abnormal echogenicity associated with microscopic calcification and gliosis. Sonography is excellent for the detection of acute cerebrovascular complications during extracorporeal membrane oxygenation, but the appearance of these lesions is variable and nonspecific. AJR 153: , August 1989 Received December 29, 1988; accepted after revision March 21, Department of Radiology, The Children's Hospital National Medical Center, and The George Washington University School of Medicine and Health Sciences, Washington, DC Department of Child Health and Development, The Children's Hospital National Medical Center, and The George Washington University School of Medicine and Health Sciences, Washington, DC 'Present address: Department of Radiology, The Johns Hopkins Hospital, 600 N. Wolfe St., Baltimore, MD Address reprint requests to G. A. Taylor. 4 Department of Pathology, The Children's Hospital National Medical Center, and The George Washington University School of Medicine and Health Sciences, Washington, DC Department of Neonatology, The Children's Hospital National Medical Center, and The George Washington University School of Medicine -and Health Sciences, Washington, DC X/89/ <D American Roentgen Ray Society The increased risk of cerebrovascular injury in neonates treated with extracorporeal membrane oxygenation (ECMO) has been well documented [1, 2]. The severity and distribution of lesions differ from those typically seen in premature neonates or term neonates who have been asphyxiated [3, 4]. Often, the lesions are poorly defined and exhibit unusual patterns of echogenicity [5]. This appearance has been attributed to blood that has clotted poorly because of heparinization [6]. Yet, the nature of the process leading to the atypical sonographic appearance is unclear. This article correlates cranial sonographic findings with autopsy findings in a group of neonates treated with ECMO. Materials and Methods Two hundred one neonates with severe respiratory failure were treated with ECMO at our institution between 1984 and Inclusion criteria for treatment with ECMO have been published elsewhere [7]. All the neonates were treated with venoarterial bypass via the right carotid artery and jugular vein. Both vessels were ligated permanently during the placement of vascular catheters. Cranial, sonography was performed before ECMO and daily during extracorporeal therapy for all neonates. A high-resolution, computerized sonographic unit (Acuson, Mountain View, CA) with 5-MHz sector and linear transducers was used for all studies. Coronal and sagittal images were obtained routinely. Thirty-two neonates (16%) died as a result of discontinuation of ECMO bypass because of severe intracranial complications or cyanotic heart disease. Permission for postmortem

2 356 TAYLOR ET AL. AJR:153, August 1989 examination of the brain was obtained for 17 of these neonates. In addition to persistent pulmonary hypertension of the neonate, our patients had the following pulmonary diagnoses: hyaline membrane disease, five; congenital diaphragmatic hernia, five; meconium aspiration syndrome, three; and sepsis, two. Persistent pulmonary hypertension alone was present in two neonates. The mean gestational age was 37.4 weeks (range, weeks). Patients less than 34- weeks gestational age were excluded from ECMO because of a higher risk for intracranial bleeding [8]. The surface of the whole brain was examined at the time of autopsy, and 5-15 mm coronal sections were obtained after formalin fixation. Multiple tissue samples were removed, and routine techniques were used to stain them for microscopic analysis. Sonograms were interpreted without knowledge of autopsy findings. Although all serial sonograms were reviewed, only the final study, obtained within 24 hr of death, was correlated with the pathologic examination. Results Pathologic Findings Macroscopic lesions were identified at autopsy in 13 (76%) of 17 neonates. These results are summarized in Table 1. Five neonates had multifocal hemorrhagic white-matter infarcts, three had hemorrhagic infarcts of the cerebral cortex and white matter, three had hemorrhagic infarcts of the cerebellum, one had a germinal matrix and intraventricular hemorrhage, and one had bilateral periventricular cysts with surrounding gliosis. All 17 neonates had abnormalities on microscopic examination. Multifocal neuronal necrosis with or without gliosis was the lesion seen most often; it was present in 14 neonates. Gliosis only, indicative of previous ischemic injury, was rec- TABLE 1: Sonographic and Autopsy Findings in 13 of 17 Neonates Treated with Extracorporeal Membrane Oxygenation Case Age at K?se Death Sonographic Findings Gross Pathologic Findings Na (days) 1 4 Normal Germinal matrix and intraventricular hemorrhage Bilateral cortical infarcts and petechial white-matter hemorrhage 2 3 Patchy, bilateral, brightly echogenic lesions in frontal, parietal, and occipital lobes 3 3 Diffusely echogenic brain with mass effect in right hemisphere 4 10 Patchy, bilateral, moderately echogenic lesions Hi frontal, parietal, and occipital periventricular white matter 5 2 Large, brightly echogenic, right temporooccipital lesion with "fluid-fluid" interface 6 3 Large, mixed-echogenicity lesion in posterior fossa 7 3 Large, brightly echogenic, right parietooccipital lesion with "fluidfluid" interface 8 4 Large, mixed-echogenicity lesion in left temporal lobe 9 3 Bilateral, subependymal, and intraventricular hemorrhage, and brightly echogenic, occipital periventricular white-matter lesions 10 3 Mixed-echogenicity lesion in posterior fossa with extension into third ventricle; separation and elevation of occipital horns of lateral ventricles Bilateral echogenic basal ganglia Bilateral, poorly defined, moderately echogenic lesions in frontal and occipital white matter; brightly echogenic focal lesions in both basal ganglia Normal _ Large infarct in right parietal and right cerebellar lobes Bilateral hemorrhagic infarcts in periventricular white matter Hemorrhagic infarct in right temporooccipital lobe Hemorrhage in vermis and medial portions of both cerebellar lobes Right parietooccipital hemorrhage Hemorrhagic infarct in left temporal lobe Subependymal, intraventricular hemorrhage and bilateral, occipital white-matter hemorrhagic infarcts 28-g clot in cistema magna, hemorrhagic infarct in right cerebellar hemisphere, and diffuse cerebral edema Normal Bilateral, periventricular white-matter hemorrhagic infarcts; normal basal ganglia Bilateral 2-3 mm periventricular cysts in occipital white matter; subarachnoid hemorrhage in cerebellum Note. Sonograms and gross pathologic examinations of four of the 17 neonates were normal.

3 AJR: 153, August 1989 POST-ECMO CEREBROVASCULAR INJURY IN NEONATES 357 ognized in two, and germinal matrix hemorrhage with microscopic white-matter edema was present in one. The abnormalities were most commonly patchy in distribution and located at the boundary zones of different arterial territories (watershed areas). No preponderance of lesions was found in the distribution of the right middle cerebral artery. Sonographic Findings Sonography detected parenchyma! abnormalities in 11 neonates (85%) that had macroscopic lesions at autopsy. In five neonates, irregular areas of variably increased echogenicity were seen in a patchy, bilateral distribution. At autopsy, these corresponded to areas of focal hemorrhagic infarction in the periventricular white matter (Fig. 1). Hemorrhagic infarction could not be reliably distinguished from nonhemorrhagic infarction or edematous brain on sonography. Focal areas of abnormal brain resulted in increased echogenicity regardless of the presence of blood (Fig. 2). Poorly clotted blood also showed a wide variation in echogenicity, ranging from nearly echo-free (Fig. 1B) to hyperechoic and indistinguishable from adjacent areas of abnormal brain (Fig. 3). The location and true extent of bleeding was particularly difficult to define in Fig. 1. Full-term neonate with meconium aspiration syndrome. A and B, Anterior coronal (A) and left sagittal (B) sonograms show patchy bilateral areas of variably altered echogenicity in periventricular white matter and focal lucent lesion with echogenie rim in left parietal lobe (arrow). C-E, Coronal whole-brain slices confirm bilateral hemorrhagic infarcts and poorly clotted blood within hemorrhagic infarct corresponding to focal lucent lesion in B. Fig week-gestation neonate with sepsis and pulmonary hypertension. A, Coronal sonogram shows lesion in left temporal lobe consisting of moderately echogenic area superiorly and nearly normal-appearing brain parenchyma inferiorly and medially (arrows). B, Coronal whole-brain section through same area shows large hemorrhagic infarct adjacent to dusky zone of ischemic brain. Hemorrhagic infarct could not be distinguished from ischemic, nonhemorrhagic brain on sonography.

4 358 TAYLOR ET AL. AJR: 153, August 1989 Fig week-gestation neonate with congenital diaphragmatic hernia. Sonogram before extracorporeal membrane oxygenation was normal. A and B, Posterior coronal (A) and midline sagittal (B) sonograms show large posterior fossa hemorrhage with heterogeneous echogenicrty. Blood is present in third ventricle (arrow), and temporal horns of lateral ventricles are displaced laterally. Cerebellum, parenchyma! hematoma, and subarachnoid blood cannot be distinguished from surrounding structures. C, Base view of whole brain shows extension of hemorrhage into basal cisterns causing lateral displacement of lateral ventricles on sonography. 0, Posterior view of cerebellum shows patchy areas of parenchymal hemorrhage in left cerebellar lobe (large arrow). Only a thin rim of right cerebellar lobe remains (small arrows) after evacuation of large (28 g) clot from posterior fossa at autopsy. the posterior fossa; indistinct or disappearing cerebellar contours were a sonographic sign of bleeding. As poorly clotted blood with mixed echogenicity accumulated around the cerebellum, it became more difficult to visualize and distinguish the cerebellum from surrounding structures. Sonography showed two hemorrhages with linear interfaces between brightly echogenic and lucent areas that resembled fluid-fluid interfaces between clotted and fluid blood. At autopsy, the echogenic areas were found to be well-defined hematomas adjacent to areas of solid but incompletely clotted blood (Fig. 4). Gross pathologic findings at autopsy were normal in one neonate (case 11 in Table 1), in whom sonography showed diffusely echogenic basal ganglia bilaterally (Fig. 5). Scattered microcalcifications and diffuse neuronal necrosis of the basal ganglia and frontal cortex were identified at microscopy. In a second neonate, sonography showed brightly echogenic lesions in both basal ganglia. These areas were normal macroscopically, but showed microscopic areas of focal gliosis and infarction without calcification (Fig. 6). Tests for viral and parasitic diseases associated with intracranial calcification (e.g., herpes or cytomegalovirus infections, rubella, toxoplasmosis) were negative in both neonates. Two patients had normal sonograms: At autopsy, one neonate had periventricular cysts in the occipital white matter, and the other had macroscopically visible subependymal and intraventricular hemorrhages. Discussion We found excellent correlation between sonographic and gross necropsy findings. Only two of 13 neonates with macroscopic lesions had normal cranial sonograms. Periventricular cysts, considered to be late sequelae of ischemic brain injury, were present in one neonate. These were located peripherally, near the edge of the sonographic beam, and probably were undetected with sonography. The second neonate had subependymal and intraventricular hemorrhages. The last premortem sonogram in this neonate was obtained almost 24 hr before death, and it is likely that these hemorrhages occurred after the last sonogram was obtained.

5 AJR:153, August 1989 POST-ECMO CEREBROVASCULAR INJURY IN NEONATES 359 Fig. 4. Full-term neonate with meconium aspiration. A and B, Coronal (A) and right sagittal (B) sonograms show sharp linear demarcation between two areas of abnormal echogenicity in right parietooccipital lobe. Although this interface resembled a fluid-fluid interface, it was oriented almost perpendicular to the horizon in this supine neonate. C, Coronal whole-brain section through the same area shows well-defined hematoma adjacent to area of solid but poorly clotted blood forming sharp interface depicted on sonogram (arrows). No free fluid was seen at autopsy. Fig. 5. Full-term neonate with persistent pulmonary hypertension. A, Left sagittal sonogram shows diffusely echogenic basal ganglia (arrow). This finding was present bilaterally. No other abnormalities were detected on cranial sonography. B-D, Postmortem CT scans show symmetrically radiodense putamina, globi pallidus (B), and frontal cortex (C and D). Histologic examination showed these imaging abnormalities were associated with microscopic calcifications and neuronal necrosis without hemorrhage. The abnormalities present at autopsy typically consisted of larger, focal lesions within areas of more diffuse microscopic neuronal injury. Although the detection of lesions with sonography was excellent in this group, the variable echogenicity of abnormal areas, together with distortion of anatomic landmarks, made it difficult to define prospectively the extent and exact location of parenchyma! lesions. This was particularly so in the posterior fossa. When the normally anechoic CSF

6 360 TAYLOR ET AL. AJR: 153, August 1989 Fig week-gestation neonate with respiratory distress syndrome. Coronal image shows bright echogenic lesions in both basal ganglia (arrows). Focal gliosis and infarction of basal ganglia were seen on microscopic examination. No hemorrhage was found. spaces were filled with poorly clotted blood of moderate echogenicity, normal cerebellum, parenchymal hemorrhage, and-subarachnoid blood became indistinguishable. The unusual and variable sonographic appearance exhith. ited by these lesions may be related to preexisting brain ischemia caused by perinatal asphyxia, or to systemic heparinization during ECMO. Hernanz-Schulman et al. [3] recently described a wide variability in the echogenicity of nonhemorrhagic infarcts, ranging from inhomogeneously echogenic parenchyma to brightly echogenic focal lesions. Edema, cellular disruption, and debris occur in the ischemic areas to a variable degree and are likely to increase the number of reflected echoes on sonography. In addition, parenchymal hemorrhages in premature neonates often represent secondary bleeding within previously infarcted or ischemic brain [9-12]. It is likely that the systemic heparinization currently necessary for the ECMO procedure precipitates or exacerbates the degree of bleeding within areas of ischemic brain, also contributing to the unusual sonographic patterns seen. Lesions not related to bleeding or acute infarction were seen in the basal ganglia of two neonates who died. One neonate (case 11 in Table 1) with diffusely echogenic basal ganglia was studied by using postmortem CT and MR imaging. Although both studies initially were interpreted as consistent with hemorrhage, none was found on either gross or microscopic examination. Neuronal necrosis with scattered microcalcification was identified on histologic examination. Another neonate had bilateral, brightly echogenic lesions on sonography that were associated with areas of focal gliosis. Gliosis with and without calcification represents the end stage of cerebral infarction [12, 13], beginning as early as 48 hr after the acute event and becoming more evident after 1 week. Calcification and the dense cellular nature of glial scars are probably responsible for the increased echogenicity of these areas. No evidence of intracranial infection was present in these neonates. Neonates treated with ECMO are at high risk for cerebrovascular injury for several reasons. Perinatal hypoxia, asphyxia, and high ventilator settings are associated with intracranial bleeding and infarction in severely ill neonates and are present in nearly all candidates for ECMO [14-17]. During ECMO therapy, systemic heparinization, alterations in cerebral blood flow, and unilateral ligation of the right carotid artery and jugular vein all may contribute in unique ways to this risk [18-20]. The exact mechanisms by which these factors influence risk remain poorly understood and require further study. Daily screening with cranial sonography has become a routine part of ECMO therapy, as clinical symptoms of intracranial hemorrhage and infarction are often absent or difficult to recognize in the neonate. Early detection of intracranial complications is essential in the management of these neonates. Extension of small parenchymal hemorrhages can be prevented by lowering the amount of systemic heparin used and increasing the amount of blood bypassed through the circuit (higher flow rates help maintain the extracorporeal circuit from clotting). If a parenchymal lesion is already extensive at the time of discovery, a change in the amount of heparin is seldom effective, and discontinuation of ECMO bypass is strongly considered. Although sonography is an accurate tool for the detection of intracranial complications, it is important to consider its limitations. Small or peripherally located lesions may not be detected, and the sonographic appearance of larger, easily detectable lesions is variable and nonspecific. Familiarity with the limitations of sonography and the unusual spectrum of appearances of these lesions is important for the accurate and early diagnosis of life-threatening hemorrhage that may occur during ECMO. ACKNOWLEDGMENTS We thank Cynthia Pointdexter, Linda M. Catena, Debra B. Garin, and John Hilton for help with this project. REFERENCES 1. Ciltey RE, Zwischenberger JB, Andrews AF, Bowerman RA, Roloff DW, Bartlett RH. Intracranial hemorrhage during extracorporeal membrane oxygenation in neonates. Pediatrics 1986:78: Sell LL, Cullen ML, Whittlesey GC, et al. Hemorrhagic complications during extracorporeal membrane oxygenation: prevention and treatment. J Pediatr Surg 1986;21: Hemanz-Schulman M, Conn W, Genieser NB. Sonography of cerebral infarction in infancy. AJR 1988:150: Schellinger 0, Grant EG, Manz HJ, Patronas NJ. Intraparenchymal hemorrhage in preterm neonates: a broadening spectrum. AJR 1988;150: Taylor GA, Fitz CR, Miller MK, Garin DB, Catena LM, Short BL. Intracranial abnormalities in infants treated with extracorporeal membrane oxygenation: imaging with US and CT. Radiology 1987:165: Bowerman RA, Zwischenberger JB, Andrews AF, Bartlett RH. Cranial sonography of the infant treated with extracorporeal membrane oxygenation. AJR 1985:145: Beck R, Anderson KD, Pearson GD, Cronin J, Miller MK, Short BL. Criteria for extracorporeal membrane oxygenation in a population of infants with persistent pulmonary hypertension of the newborn. J Pediatr Surg 1986; 21:

7 AJR: 153, August 1989 POST-ECMO CEREBROVASCULAR INJURY IN NEONATES Andrews AF, Roloff DN, Bartlett RH. Use of extracorporeal membrane oxygenation in persistent pulmonary hypertension of the newborn. Clin Perinatal 1984;11: Voipe JJ, Perlman JM, Herscovitch P, Raichle ME. Positron emission tomography in the assessment of regional cerebral blood flow in the newborn. Ann A/euro/1982; 12: Pape KE, Wigglesworth JS. Pathologic anatomy of intraventricular hemorrhage. In: Hemorrhage, Ischemia and the perinatal brain. Philadelphia: Lippincott, 1979: Lavene Ml, Wigglesworth JS, Dubowitz V. Hemorrhagic periventricular leukomalacia in the neonate: a real-time ultrasound study. Pediatrics 1983; 71: Duyckaerts C, Hauw JJ. Pathology and pathophysiology of brain ischemia. Neuroradiology 1985;27: Rushton Dl, Preston PR, Durbin GM. Structure and evolution of echo dense lesions in the neonatal brain. Arch Dis Child 1985;60: Tarby TJ, Voipe JJ. Intraventricular hemorrhage in the premature infant. Pediatr Clin North Am 1982;29: Klesh KW, Murphy TF, Scher MS, Buchanan DE, Maxwell EP, Guthrie RD. Cerebral infarction in persistent pulmonary hypertension of the newborn. Am J Dis Child 1987;141: Scher MS, Klesh KW, Murphy TF, Guthrie RD. Seizures and infarction in neonates with persistent pulmonary hypertension. Pediatr Neurol 1986,2: Lou HC, Niels AL, Friis-Hansen B. Impaired autoregulation of cerebral blood flow in the distressed newborn infant. J Pediatr 1979;94: Schumacher RE, Barks JDE, Johnson MV, et al. Right-sided brain lesions in infants following extracorporeal membrane oxygenation. Pediatrics 1988:82: Taylor GA, Short BL, Glass P, Ichord Rl. Cerebral hemodynamics in infants undergoing extracorporeal membrane oxygenation: further observations. Radiology 1988;168: Mitchell DG, Merton D, Desai H, et al. Neonatal brain: color Doppler imaging. Part II. Altered flow patterns from extracorporeal membrane oxygenation. Radiology 1988;167: