Deconvolution of Mixtures in Analysis of Neural Synaptic Transmission
Nov 30 1991
Neurophysiologists investigating mechanisms underlying neural responses to stimuli have, in recent years, developed substantial interest in modeling certain types of neural response data suing simple mixture distributions. Techniques of mixture deconvolution using likelihood based techniques have become popular. This paper reports on novel Bayesian approaches using (uncertain) mixtures of (uncertain numbers of) noise distributions to model data measuring maximum levels of evoked neural responses following various levels of electrical stimulus of nerve tissue. We discuss some of the key scientific issues, including physiological hypotheses of 'quantal' levels of neuronal transmissions, together with technical aspects of data analysis, modeling, and the use of prior information in addressing these issues within an appropriate Bayesian framework. Illustration of neural response deconvolution analysis using this approach is presented.