A quality measure for repeating multiple-unit spike patterns

Area of Science:

• Computational Neuroscience
• Systems Neuroscience
• Data Analysis

Background:

• Multiple-neuron recordings often reveal repeating spatio-temporal spike patterns (STPs).
• Distinguishing genuine STPs from chance occurrences is crucial for understanding neural coding.
• Existing statistical tests face challenges due to multiple quality criteria like pattern size and occurrence count.

Purpose of the Study:

• To propose a unified quality measure for spatio-temporal spike patterns (STPs).
• To develop a method for combining multiple pattern criteria into a single, statistically robust measure.
• To facilitate reliable comparison of STPs and enhance statistical testing.

Main Methods:

• Defined a canonical quality measure based on the 'unlikeliness' of a pattern.
• Derived a mathematical formula for computing this measure in stationary spike trains.
• Evaluated the correlation between the stationary formula and simulated quality measures on experimental data.

Main Results:

• The proposed quality measure provides a single metric to assess the significance of STPs.
• A derived formula for stationary spike trains can serve as a proxy for quality in non-stationary data.
• Simulations showed that stationary-derived formulas can be useful, though sometimes imprecise, for comparing STPs.

Conclusions:

• The proposed quality measure offers a robust way to evaluate STPs in multi-neuron recordings.
• Simplified formulas can act as effective proxies for STP quality assessment and statistical testing.
• New test statistics are proposed to reduce computational and sampling demands in STP evaluation.