INTEROBSERVER AGREEMENT IN ASSESSING EARLY CT SIGNS OF MIDDLE CEREBRAL-ARTERY INFARCTION

PURPOSE: To assess the reliability of detecting signs of hemispheric i nfarction on CT scans obtained within 6 hours of the onset of symptoms . METHODS: A neuroradiologist selected 12 normal and 33 abnormal CT st udies showing the hyperdense middle cerebral artery sign (HMCAS) (n = 10), brain swelling (n = 22), and parenchymal hypodensity (n = 33) fro m two series of 750 patients with recent onset of middle cerebral arte ry stroke, These selections served as the reference source for a nonbl inded analysis of the initial and follow-up CT scans. Six neuroradiolo gists then reviewed the CT scans twice, first blinded then not blinded to clinical symptoms. They assessed the signs of infarction for each hemisphere separately and estimated the volume of abnormal parenchymal hypodensity in increments of 20% within the territory of the middle c erebral artery. RESULTS: Unblinding the reviewers did not change inter observer agreement significantly, The chance adjusted agreement was mo derate to substantial: kappa = .62 (95% confidence interval [Cl],.46 t o .78) and kappa = .57 (95% CI, .33 to .81) for the HMCAS of the right and left hemisphere, respectively; kappa = .59 (95% Cl, .47 to .71) a nd kappa = .56 (95% Cl, .38 to .74) for focal brain swelling of the ri ght and left hemisphere, respectively; and kappa = .58 (95% Cl, .50 to .66) and kappa = .55 (95% Cl, .32 to .67) for parenchymal hypodensity of the right and left hemisphere, respectively. Weighted kappa was .6 5 and .57 for the estimation of the hypodense tissue volume in the rig ht and left hemisphere, respectively. Agreement with the reference sou rce ranged from 73% to 93% for all variables and both hemispheres. CON CLUSION: Even with no clinical information, neuroradiologists can asse ss subtle CT signs of cerebral infarction within the first 6 hours of symptom onset with to substantial interobserver agreement.