Separating Neural Oscillations from Aperiodic 1/f Activity: Challenges and Recommendations.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Signal Processing

Background:

• Electrophysiological power spectra contain both periodic (neural oscillations) and aperiodic (1/f) components.
• Separating these components is crucial for accurate analysis of neural activity.
• The aperiodic component has gained recent research interest.

Purpose of the Study:

• To scrutinize and compare two common methods, FOOOF and IRASA, for separating periodic and aperiodic spectral components.
• To evaluate the performance of these methods across diverse EEG, MEG, and LFP datasets.
• To identify challenges and spectral features that impede accurate separation.

Main Methods:

• Evaluation of FOOOF and IRASA using real EEG, MEG, and LFP data from three independent datasets.
• Simulation of power spectra to highlight specific features hindering component separation.
• Quantification of parameterization error by comparing method outputs to a priori simulation parameters.

Main Results:

• Both FOOOF and IRASA face distinct challenges in separating spectral components across different datasets.
• Simulations identified specific spectral features that impede accurate separation.
• Performance evaluation included advantages, challenges, computational costs, and recommendations for method usage.

Conclusions:

• Accurate separation of periodic and aperiodic spectral components is essential for robust electrophysiological data analysis.
• FOOOF and IRASA offer different strengths and weaknesses, with performance varying by dataset and spectral characteristics.
• Recommendations are provided to guide the optimal application of these methods in research.