Cranial sonography (CS) is the most commonly used method for imaging of the newborn brain. The sensitivity and specificity of CS continues to improve with technologic advances, but the limitations of sonography in diagnosing neuropathologic entities such as periventricular leukomalacia (PVL) and intraventricular hemorrhage (IVH) are still being evaluated. A retrospective review of the medical records, sonographic imaging, and brain pathology of infants who died in the Texas Children's Hospital Neonatal Intensive Care Unit from 1 Jan 1993 to 31 Dec 1995 was performed. Of 244 infants who died, 51 had premortal CS as well as brain autopsy. Comparison of sonographic and neuropathologic diagnoses in these 51 infants was made by a neonatologist, a neuropathologist, and a pediatric radiologist experienced in cranial sonography; autopsy results were treated as the ”gold standard“. Most patients had multiple diagnoses. Each diagnosis was considered separately in assigning agreement or disagreement between CS and autopsy. Among the 51 patients there were 132 diagnoses; agreement existed between CS and autopsy in 74 of the 132 diagnoses. Of the 58 disagreements between autopsy and CS diagnosis, 13 could be explained by either the size of the lesion relative to the resolution of sonographic imaging or by timing of onset of the pathology relative to CS. These 13 diagnoses were not used in the statistical evaluation of CS reliability. There were therefore 45 of 132 diagnoses (34%) that were classified as true disagreement. The identification by CS of selected neuropathology was as follows: Table On the whole, agreement between CS and autopsy was good (83-97.5%). Most discrepancies could be explained by identified limitations of CS: posterior fossa location(parenchymal hemorrhage); inconsequential overdiagnosis of mild germinal matrix hemorrhage/gr. 1 IVH; or the diffuse nature of some lesions(HIE/infarction).

Table 1
Full size table