EEG might be better left alone, but ERPs must be attended to: Optimizing the late positive potential preprocessing pipeline

Abstract

The late positive potential (LPP) is an ERP component commonly used to study emotional processes and has been proposed as a neuroaffective biomarker for research and clinical uses. These applications, however, require standardized procedures for elicitation and ERP data processing.

We evaluated the impact of different EEG preprocessing steps on the LPP's data quality and statistical power. Using a diverse sample of 158 adults, we implemented a multiverse analytical approach to compare preprocessing pipelines that progressively incorporated more steps: artifact detection and rejection, bad channel interpolation, and bad segment deletion. We assessed each pipeline's effectiveness by computing the standardized measurement error (SME) and conducting simulated experiments to estimate statistical power in detecting significant LPP differences between emotional and neutral images.

Our findings highlighted that artifact rejection is crucial for enhancing data quality and statistical power. Voltage thresholds to reject trials contaminated by artifacts significantly affected SME and statistical power. Once artifact detection was optimized, further steps provided minor improvements in data quality and statistical power. Importantly, different preprocessing pipelines yielded similar outcomes.

These results underscore the robustness of the LPP's affective modulation to preprocessing choices and the critical role of effective artifact management. By refining and standardizing preprocessing procedures, the LPP can become a reliable neuroaffective biomarker, supporting personalized clinical interventions for affective disorders.