Approaches to characterizing oscillatory burst detection algorithms for electrophysiological recordings

Area of Science:

• Neuroscience
• Computational Neuroscience

Background:

• Fast neural oscillations in local field potentials and electroencephalograms are vital for cognitive processes.
• These oscillations are disrupted in aging and disease, making them a target for therapeutic interventions.
• Analyzing these oscillations is challenging due to their short-lived bursts and interference from aperiodic neural activity.

Purpose of the Study:

• To develop a robust, open-source toolkit for analyzing and detecting transient oscillatory bursts in neural data.
• To provide a method for evaluating the reliability of oscillatory burst detection algorithms.

Main Methods:

• A four-step processing pipeline was created, including power spectrum decomposition into periodic and aperiodic components.
• Properties of transient oscillatory bursts were derived and optimized to account for aperiodic contamination.
• Surrogate neural signals were synthesized to evaluate oscillatory burst detection algorithms using receiver operating characteristic analysis.

Main Results:

• The developed algorithm characterizes oscillatory burst features across various frequency bands and brain regions.
• It enables recording-specific performance evaluation of detection algorithms.
• For the analyzed dataset, the optimal detection threshold for gamma bursts was found to be lower than commonly used values.

Conclusions:

• The developed pipeline facilitates the evaluation of detection algorithm thresholds for individual neural recordings.
• This approach enhances the ability to reliably study neural oscillations disrupted by aging and disease.