We examined the correlation between cranial sonogram (CS) and autopsy neurodiagnoses in 51 infants. As shown in the data table, of 21 cases of periventricular leukomalacia(PVL), 7 had been diagnosed by CS prior to infant demise; 12 of 14 cases of false negative (FN) diagnoses could be explained by either time interval (TI) between CS and demise or the microscopic nature (M) of the PVL at autopsy. Other diagnoses are similarly outlined: IVH--intraventricular hemorrhage; PH--parenchymal hemorrhage; FP--false positive; FN--false negative; TI--time interval (# of cases); M--microscopic lesions (# of cases).

Table 1 No caption available.
Full size table

The data were analyzed by considering one primary diagnosis for either CS or autopsy in each infant, the most clinically-significant lesion as defined by 2 neonatologists. When the disagreement between autopsy and CS primary diagnosis could be explained by TI, the case was not included in the calculation of CS FNs. The accuracy of the CS in relation to autopsy findings for the primary diagnosis was 59%. Positive predictive value of the CS was 77%, while negative predictive value was 19%. Of 21 false negative (FN) CS primary diagnoses, 4 cases were explained by TI, and M explained 5 cases; 8 of these FN cases were significant to patient management and outcome. Factors involved in the failure of CS to identify primary diagnosis neuropathology were not defined in 12 infants. Two significant false positive (FP) primary diagnoses were made by CS: PVL in one infant and grade 2-3 IVH in another infant. For CS to be used in clinical decision-making, the clinician must recognize this tools' limitations.