Bleeck, S. and Winter, I.M.
A model of vental cochlear nucleus units based on first order intervals
At BSA Short Papers Meeting on Experimental Studies of Hearing and Deafness.
14 - 15 Sep 2006.
Full text not available from this repository.
Significant parallel processing occurs at the level of the ventral cochlear nucleus (VCN). It is believed that underlying this processing are discrete cellular response types which have a unique signature in their temporal adaptation patterns (TAPs). However, in reality it is difficult to define these TAPS as discrete. Here we demonstrate that a large sample (n=874) of VCN neurons can be accurately modelled as belonging to a continuous parameter space, even if they were not classifiable using traditional methods (Blackburn and Sachs, 1988).
We constructed first order interspike interval distributions (FOIDs) from the responses to 50ms best-frequency tone bursts. FOIDs were then plotted on a logarithmic time axis and fitted with an “extreme value distribution” (EVD). The EVD is a variation of the statistical “central limit theorem” that states that the distribution of the extreme values of arbitrary populations has a defined EVD. EVDs predict the probability of the occurrence of rare events independently of the underlying probability function. The validity of the fit was checked by a Kolgomorov-Smirnov test. EVDs fitted more units than other models (e.g. normal, gamma, weibull, exponential). Therefore the probability for spike generation can be described statistically without knowing the exact underlying mechanism. The two parameters of the fitted FOID (mean and variance) uniformly fill a well defined area of the parameter space. Single unit types recorded from the VCN occupy different, slightly overlapping areas of this parameter space. An “automatic classification” scheme (measuring the distance to population means) closely matches the classification schemes traditionally used by physiologists.
Actions (login required)