Quantifying neural coding of event timing.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Information Theory

Background:

• Single-neuron firing analysis typically uses events to predict neural responses (e.g., perievent time histograms).
• Perievent time histograms lack information on the reliability of single-trial neural responses.
• Understanding single-trial reliability is crucial for assessing a neuron's coding capacity.

Purpose of the Study:

• To propose a novel method for analyzing neural firing by predicting event timing from neural activity.
• To quantify the information a spike train provides about the timing of an external event.
• To assess the factors influencing a neuron's ability to encode event timing.

Main Methods:

• Estimating spike train likelihood using an inhomogeneous gamma process model.
• Applying Bayes' rule to generate a probability distribution for event occurrence.
• Utilizing information theory to derive a quantitative measure of uncertainty reduction.

Main Results:

• The developed method quantifies the reduction in uncertainty about event timing based on observed spike trains.
• The approach is sensitive to response amplitude, baseline firing rate, and trial-to-trial response consistency.
• The technique provides a single value (in bits) reflecting the information conveyed by the spike train.

Conclusions:

• This method offers a new perspective on neural coding by reversing the typical analysis direction (event-to-neuron to neuron-to-event).
• It provides a reliable measure of how well a neuron encodes the timing of external events.
• The technique facilitates objective comparisons between different neuronal populations and their coding capabilities.