Characterizing neural coding performance for populations of sensory neurons: comparing a weighted spike distance metrics to other analytical methods

ABSTRACTThe identity of sensory stimuli is encoded in the spatio-temporal patterns of responses of the neural population. For stimuli to be discriminated reliably, differences in population responses must be accurately decoded by downstream networks. Several methods to compare the pattern of responses and their differences have been used by neurophysiologist to characterize the accuracy of the sensory responses studied. Among the most widely used analysis, we note methods based on Euclidian distances or on spike metric distance such as the one proposed by van Rossum. Methods based on artificial neural network and machine learning (such as self-organizing maps) have also gain popularity to recognize and/or classify specific input patterns. In this brief report, we first compare these three strategies using dataset from 3 different sensory systems. We show that the input-weighting procedure inherent to artificial neural network allows the extraction of the information most relevant to the discrimination task and thus the method performs particularly well. To combine the ease of use and rapidity of methods such as spike metric distances and the advantage of weighting the inputs, we propose a measure based on geometric distances were each dimension is weighted proportionally to how informative it is. In each dimension, the overlap between the distributions of responses to the two stimuli is quantified using the Kullback-Leibler divergence measure. We show that the result of this Kullback-Leibler-weighted spike train distance (KLW distance) analysis performs as well or better than the artificial neural network we tested and outperforms the more traditional spike distance metrics. We applied information theoretic analysis to Leaky-Integrate-and-Fire model neuron responses and compare their encoding accuracy with the discrimination accuracy quantified through these distance metrics to show the high degree of correlation between the results of the two approaches for quantifying coding performance. We argue that our proposed measure provides the flexibility, ease of use sought by neurophysiologist while providing a more powerful way to extract the relevant information than more traditional methods.