Interoperator variability in quantitative electroencephalography.

Abstract

The purpose of the study was to determine whether quantitative or discriminant analysis of the electroencephalograph (EEG) would vary significantly when the same EEG was analysed by 3 different operators. EEGs on 10 healthy volunteers were recorded on the Cadwell Spectrum AT 386, using the Electrocap (10-20 system). The EEGs were analysed independently, with each operator selecting the first 48 artifact-free epochs. The results were analysed using the non-parametric Friedman two-way analysis of variance (ANOVA) for the discrimination analysis and a one-way ANOVA for the monopolar and bipolar Absolute Power raw measures. Statistical analysis of the discriminant data showed no significant differences between operators, with 7 of 10 studies yielding the same results. The remaining 3 studies were classified either as borderline or normal when analysed by different operators. Although a series of "t" tests comparing 2 operators showed most variability occurring in Absolute Power as compared with Relative Power, Power Asymmetry and Coherence, ANOVA of the raw mono- and bipolar Absolute Power measures showed no significant differences between the operators at the P = 0.05 level. Thus the differences between the operators were non-significant when comparing quantitative EEG analyses with respect to both the raw measures and the discriminant analyses.