Cross-correlation measures of unresolved multi-neuron recordings

Area of Science:

• Neuroscience
• Computational Neuroscience
• Signal Processing

Background:

• Multi-electrode recordings in neuroscience often capture unresolved spike activity from multiple neurons.
• Previous work simplified cross-correlation analysis under restrictive assumptions about individual neuron correlations.
• Accurate interpretation of neural population activity requires understanding distortions in unresolved data.

Purpose of the Study:

• To investigate the impact of unresolved multi-neuron spike activity on inter-electrode cross-correlation measurements.
• To analyze how close and distant correlations between neurons affect measured cross-correlations.
• To evaluate the consequences of imperfect spike sorting on correlation analyses.

Main Methods:

• Mathematical modeling of cross-correlation coefficients for two or three unresolved neurons per electrode.
• Analysis of the relationship between measured inter-electrode correlation and underlying single-neuron correlations.
• Examination of the effects of waveform sorting errors on correlation metrics.

Main Results:

• The measured cross-correlation coefficient is a linear sum of distant correlations divided by a non-linear function of close correlations.
• Under most conditions, close correlations significantly reduce the measured distant correlation.
• Poor waveform sorting introduces substantial distortions, particularly with dynamic neural activity.

Conclusions:

• Cross-correlation analysis of unresolved neural recordings can lead to significant misinterpretations of neuronal organization and coding.
• Researchers must account for distortions caused by multi-unit activity and imperfect spike sorting.
• This study serves as a cautionary note for applying cross-correlation methods to complex neural data.