Interobserver variation in computed tomography of the brain.

As part of the University of Western Ontario Dementia Study, the computed tomographic brain scans of 16 patients were reviewed by three neuroradiologists. The size of the ventricles and sulci were rated using a six-point scale. Infarction and white matter changes were assessed in accordance with specified criteria. The interobserver correlations in this small series were statistically significant in 17 of 20 items, and good or acceptable for infarction, leuko-araiosis, and ventricular size. It is suggested that the use of rigid criteria for the definition of abnormality helps to promote interobserver agreement.