Reliability of fully automated versus visually controlled pre- and post-processing of resting-state EEG

Objective To compare the reliability of a newly developed Matlab® toolbox for the fully automated, pre- and post-processing of resting state EEG (automated analysis, AA) with the reliability of analysis involving visually controlled pre- and post-processing (VA). Methods 34 healthy volunteers (age: median 38.2 (20–49), 82% female) had three consecutive 256-channel resting-state EEG at one year intervals. Results of frequency analysis of AA and VA were compared with Pearson correlation coefficients, and reliability over time was assessed with intraclass correlation coefficients (ICC). Results Mean correlation coefficient between AA and VA was 0.94 ± 0.07, mean ICC for AA 0.83 ± 0.05 and for VA 0.84 ± 0.07. Conclusion AA and VA yield very similar results for spectral EEG analysis and are equally reliable. AA is less time-consuming, completely standardized, and independent of raters and their training. Significance Automated processing of EEG facilitates workflow in quantitative EEG analysis.