Using response models to estimate channel capacity for neuronal classification of stationary visual stimuli using...

Area of Science:

• Computational Neuroscience
• Neural Coding
• Information Theory

Background:

• Both spike count and temporal modulation of neural responses convey stimulus information.
• The precise contribution of temporal modulation to information encoding remains debated.
• Previous studies often depend on specific experimental datasets, limiting generalizability.

Purpose of the Study:

• To develop a response model to quantify the additional information provided by temporal modulation.
• To compare the information capacity of spike count alone versus combined spike count and temporal modulation.
• To assess the contribution of temporal modulation at different time scales to stimulus identification.

Main Methods:

• Constructed a response model based on principal component analysis (PCA) of neural responses.
• Incorporated both response strength (spike count) and coarse temporal modulation (+/-10 ms) into the model.
• Modeled responses using truncated Gaussian distributions and a linear mean-variance relationship.

Main Results:

• Stimulus-related information from spike count was 0.35 bits (V1) and 0.31 bits (IT).
• Using a two-principal-component code (including temporal modulation) increased information to 0.52 bits (V1) and 0.37 bits (IT).
• Model estimates show channel capacity increased from 1.1-1.4 bits (spike count only) to 2.0-2.2 bits (two principal components).

Conclusions:

• Incorporating temporal modulation, even at coarse time scales, significantly increases neural information capacity.
• The information gained from temporal modulation is comparable to adding a second independent neuron.
• Finer temporal modulations do not substantially contribute to identifying stationary visual stimuli.