Comparing the Signal-to-Noise Ratio Estimates for Event-Related Potentials: A Simulation Study

Event-related potential (ERP) is one of the commonly used electrophysiologic measures for brain activity with millisecond time resolution, which has been widely applied to psychology and neuroscience research. Conventionally, ERP is obtained by grand-averaging EEG recordings across multiple trials to improve the signal-to-noise ratio (SNR). Reliable quantitative analysis of the amplitude or latency of ERP requires sufficient SNR. Estimating SNR thus offers a criterion for selecting the trial number in designing experiments and the ERP analysis. Unfortunately, most researchers miss assessing SNR, which leads to the reliability of the results being unchecked, particularly under a low SNR. Although a few SNR estimates for ERP have been proposed, their performances have not yet been well compared. As a result, researchers are still left without a guideline quantifying the quality of their ERP signals. An SNR estimate is considered superior if it more successfully differentiates the difference in noises. Using both simulated and actual ERP components, in this study, we aimed to compare the performances of four SNR estimates. The area under the curve (AUC) was computed from the receiver operating characteristics (ROC) curves to quantify the performances of the SNR estimates in Task I: classifying ERP and spontaneous EEG and Task II: classifying the ERP with different levels of noises. Our results showed that the SNR estimate by calculating the ratio of the highest amplitude in the ERPs to the standard deviation in the baseline time interval (SNR
Maidhof
) was outstanding in Task I. While the SNR estimate by dividing the mean root square of the signal by the variance of the baseline (SNR
M&P
) was the best SNR estimate in Task II. These results provided a guideline for assessing the quality of the ERP, excluding experimental subjects, or designing the number of required trials before the quantitative analysis.Clinical Relevance— This study provides the rules of thumb for quantifying the ERP data quality, screening the subjects and designing the number of trials in ERP experiments.