SEED-G: Simulated EEG Data Generator for Testing Connectivity Algorithms.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Signal Processing

Background:

• Electroencephalography (EEG) signals are crucial for understanding brain circuits during cognitive tasks.
• Testing EEG connectivity estimators is challenging due to the lack of ground-truth data in real-world scenarios.
• Current simulation methods have limitations in complexity, signal count, and spectral properties, failing to capture real brain activity nuances.

Purpose of the Study:

• To introduce the SEED-G toolbox for generating high-fidelity pseudo-EEG data with controllable connectivity patterns.
• To overcome limitations of existing EEG data simulation techniques, including non-ideal and non-stationary ground-truth models.
• To provide a robust tool for validating EEG connectivity analysis methods.

Main Methods:

• Development of the SEED-G toolbox, enabling the generation of pseudo-EEG time series with user-defined connectivity patterns.
• Inclusion of guidelines for the effective use of the SEED-G toolbox.
• Performance testing of SEED-G across a wide range of conditions, including datasets with up to 60 time series.

Main Results:

• SEED-G successfully generated datasets with spectral features similar to real EEG data in under 5 seconds.
• The toolbox demonstrated the ability to simulate complex scenarios with imposed connectivity patterns.
• Validation using SEED-G confirmed the robustness of Partial Directed Coherence (PDC) estimates against inter-trial variability.

Conclusions:

• The SEED-G toolbox offers a significant advancement for simulating EEG data with ground-truth connectivity.
• It addresses critical limitations in current data generation methods, facilitating more reliable testing of connectivity estimators.
• SEED-G provides a valuable resource for neuroscience research, particularly for validating analytical techniques like PDC.