Method for stationarity-segmentation of spike train data with application to the Pearson cross-correlation

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Neural correlations are crucial for brain computation and information coding.
• Cross-correlation functions are standard for assessing neuronal functional interactions.
• Existing methods often assume stationarity, which is frequently violated in real neural data.

Purpose of the Study:

• To develop and validate a method for analyzing neural spike trains under nonstationary conditions.
• To improve the accuracy of correlation analysis in neuroscience.
• To provide a tool for detecting significant events in neural activity.

Main Methods:

• Introduced a method to empirically assess stationarity in spike trains.
• Segmented spike trains into stationary segments based on weak-sense stationarity.
• Developed a corrected pairwise Pearson cross-correlation (PCC) to account for segmentation effects.
• Validated methods using simulated data and in vivo recordings from rat prefrontal cortex.

Main Results:

• Identified firing rate covariance as a source of error in nonstationary data.
• The corrected PCC effectively accounts for segmentation-induced covariance.
• The method successfully analyzed both simulated and real in vivo neural data.
• Demonstrated the utility of the method for event detection in spike trains.

Conclusions:

• The developed method effectively handles nonstationarity in neural recordings.
• The corrected PCC offers a more accurate measure of neuronal interactions.
• This approach enhances the analysis of neural coding and can aid in event detection.