A rate and history-preserving resampling algorithm for neural spike trains

Area of Science:

• Computational Neuroscience
• Statistical Signal Processing

Background:

• Resampling methods are crucial for analyzing neural spike train data.
• Existing methods often fail to preserve important non-Poisson features like refractory periods and bursting.
• Trial-to-trial variability poses a challenge for accurate statistical analysis of neural activity.

Purpose of the Study:

• To introduce a novel resampling technique, pattern jitter, for neural spike trains.
• To develop a method that preserves spike train structure and is robust to variability.
• To enable more accurate statistical exploration of neural coding.

Main Methods:

• Pattern jitter resampling algorithm developed.
• Preserves recent spiking history of all spikes.
• Constrains resampled spikes to remain near original positions.
• Utilizes dynamic programming for algorithmic efficiency.

Main Results:

• Resampled spike trains maintain non-Poisson properties (refractory periods, bursting).
• Method demonstrates robustness against trial-to-trial variability.
• Achieves maximal randomness within imposed constraints.
• Efficient implementation via dynamic programming.

Conclusions:

• Pattern jitter is an effective resampling method for neural spike trains.
• Preserves key statistical properties and handles variability.
• Offers a robust tool for analyzing complex neural data.