Estimating the Information Extracted by a Single Spiking Neuron from a Continuous Input Time Series

Area of Science:

• Neuroscience
• Computational Neuroscience
• Information Theory

Background:

• Understanding the neural code, the relationship between sensory stimuli and neuronal activity, is crucial for deciphering brain computation.
• Quantifying information transfer from stimulus to neural spike trains remains a significant challenge in neuroscience.

Purpose of the Study:

• To introduce a novel in vitro method for measuring information transfer from input stimuli to a single neuron's output spike train.
• To analyze the non-linear and amplifying characteristics of information processing in hippocampal pyramidal neurons.

Main Methods:

• Utilizing an artificial neural network to generate time-varying input currents based on a hidden state stimulus.
• Injecting this current into recorded neurons and estimating mutual information between the hidden state and spike trains.
• Independently varying input current characteristics like time constant and amplitude.

Main Results:

• The developed method allows for robust estimation of mutual information due to the binary nature of the hidden state.
• Information transfer in hippocampal pyramidal cells was found to be non-linear and amplifying, similar to a Bayesian neuron model.
• Information loss decreased significantly when the input to the neuron was more informative about the hidden state.

Conclusions:

• Hippocampal neurons exhibit non-linear information processing, amplifying signals when input is informative.
• Neurons dynamically adjust firing rates to maintain a consistent relative information transfer rate, even with changing stimulus dynamics.