The PREP pipeline: standardized preprocessing for large-scale EEG analysis.

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Portable brain imaging technology enables large-scale real-world data collection.
• Large and complex EEG data necessitate automated processing.
• Early preprocessing stages significantly impact signal-to-noise ratio and artifact introduction.

Purpose of the Study:

• To demonstrate how early EEG preprocessing choices affect data quality.
• To introduce a robust, multi-stage referencing scheme.
• To propose a standardized early-stage EEG processing pipeline (PREP).

Main Methods:

• Investigated the impact of average referencing on signal-to-noise ratio and artifact contamination.
• Examined the dependency of noisy channel identification on the reference.
• Developed and applied a multi-stage robust referencing scheme within the PREP pipeline.

Main Results:

• Ordinary average referencing improves signal-to-noise ratio but can be contaminated by noisy channels.
• Noisy channel identification is reference-dependent, highlighting complex interactions.
• The PREP pipeline was successfully applied to over 600 EEG datasets, generating automated reports.

Conclusions:

• Standardized early-stage EEG preprocessing is essential for reliable large-scale analysis.
• The proposed PREP pipeline effectively addresses challenges in referencing and noisy channel identification.
• The PREP pipeline is available as a free MATLAB library for researchers.