Reliability and validity of ratings of signal hyperintensities on MRI by visual inspection and computerised measurement

Abstract

Brain magnetic resonance imaging (MRI) scans on 98 elderly subjects, 62 with a diagnosis of schizophrenia and 36 healthy controls, were independently and blindly rated by two investigators using the visual rating methods of Fazekas et al. (Fazekas, F., Chawluk, J.B., Alavi, A., Hurtig, H.I., Zimmerman, R.A., 1987. MR signal abnormalities at 1.5 T in Alzheimer's dementia and normal aging. American Journal of Neuroradiology 8, 421-426) and Victoroff et al. (Victoroff, J., Mack, W.J., Grafton, S.T., Schreiber, S.S., Chui, H.C., 1994. A method to improve interrater reliability of visual inspection of brain MRI scans in dementia. Neurology 44, 2267-2276) for periventricular, deep white matter and subcortical gray matter signal hyperintensities (SHs) on T2-weighted images. The hyperintense signal volumes were measured by manual delineation of the signals on a workstation using Analyze software (computerised method). The subjects also underwent a detailed neuropsychological assessment. There was a high correlation between two experienced raters for both visual ratings, and the correspondence between the two methods was high. The inter-rater reliability for the computerised method was modest but significant, and the association between the visual and computerised methods was good except for ratings for SHs in subcortical nuclei. The Fazekas and computerised methods, and to a lesser degree the Victoroff method, had modest but significant correlations with some neuropsychological test measures. In conclusion, we did not demonstrate a clear superiority in reliability or validity for one demanding computerised method of rating SHs. Visual ratings should therefore be considered adequate for most clinical and research purposes, but such ratings should be accompanied by adequate training and the provision of standard reference images.