First spike latency code for interaural phase difference and binaural correlation discrimination in the guinea pig inferior colliculus

Certain computations in the brain, such as identifying the direction of an approaching predator, must be completedfast and reliably. However, traditional neural coding theory focused on rate codes that extract information
from the neural spike count over long periods of time.
Recently, it was suggested that first spike latency is a
feature of the neural response that enables fast transmission
of information in the brain. Here we study a latency
based readout model that allows for fast computation of
sound source location in the guinea pig auditory system.
The accuracy of the latency model for discrimination
between different interaural phase differences and binaural
correlation values was compared with a conventional
rate-based model and found to improve response speed at
small cost to the decision accuracy. We find that the
spontaneous firing of neurons limits the capacity of the
latency model to accumulate information from large
populations. We study two possible solutions that allow
the latency model to overcome the detrimental effect of
spontaneous firing and achieve fast and reliable readout.
To utilize a latency based model, the brain must first
identify stimulus onset. We demonstrate how stimulus
onset time can be estimated from the responses of
neurons that are less selective to the interaural phase
difference of the stimulus and study the accuracy of this
scheme. Combining the onset estimation and the latency
models improves the accuracy of the model, especially
for readouts based on neurons with high spontaneous
firing rate. We conclude that use of latency codes has the
potential to increase processing speed and decrease
transmission time, with very little detrimental effect on
processing accuracy.