Statistical method for detection of phase-locking episodes in neural oscillations

Area of Science:

• Neuroscience
• Computational Neuroscience
• Signal Processing

Background:

• Neural networks exhibit oscillatory dynamics and synaptic interactions can lead to phase-coupled groups.
• Detecting neural synchrony in experimental data, particularly single-trial recordings, is difficult due to short-duration entrainment, noise, and uncorrelated amplitude fluctuations.

Purpose of the Study:

• To develop and validate methods for detecting neural interactions in pairs of oscillatory signals within a narrow frequency band.
• To address the challenge of amplitude fluctuations interfering with synchrony detection by focusing on phase dynamics.

Main Methods:

• Extracted phase variables from oscillatory signals to circumvent amplitude interference.
• Utilized three distinct phase-locking indices: coherence, entropy, and mutual information.
• Employed sliding analysis windows for temporal resolution and assessed statistical significance using four surrogate data methods.

Main Results:

• Phase-locking indices were calculated over time, allowing analysis at various temporal resolutions and statistical reliabilities.
• The choice of surrogate data method significantly impacted the results.
• Surrogate methods preserving temporal structure of individual phase time series effectively reduced false positives on independent signals.

Conclusions:

• Phase-based analysis with appropriate surrogate data methods offers a robust approach to detecting neural interactions in oscillatory networks.
• Careful selection of surrogate data methods is crucial for accurate synchrony detection and minimizing false positives.