World Congress on Neuroinformatics location:Vienna, Austria date:Sep 24 29/09/2001
Introduction:Neural spike trains of acoustically stimulated auditory nerve fibers have properties of fractal stochastic processes. These properties could be beneficial to the processing of the spike trains for three reasons. First, there exists a kind of long lasting memory because of the long-time correlations between the pulses. Second, since a lot of natural sounds are fractal, coding them by fractal spike trains might be a good choice. And third, fractal rates are assumed to be fault tolerant. It has therefore been suggested that auditory prostheses, such as cochlear implants, should mimic these fractal effects in their electric stimulation approach .Our aim is to show that these fractal effects are caused by processes inherent to the neurons themselves and do not need to be mimicked by the stimulation strategy of the neural prosthesis.Methods:We developed a simple 'integrate and fire' neuron model with a noisy threshold. It has been shown that the cross-membrane voltage fluctuations have a Gaussian amplitude distribution and a 1/f power spectrum over a wide range of frequencies and depolarizing current intensities . We modelled the threshold as a stochastic process with fixed mean and the same characteristics as the cross-membrane voltage fluctuations. We also added absolute and relative refractory effects.The inputs to our model were both deterministic and stochastic pulse trains. Pulses were biphasic and symmetrical square current pulses. Deterministic pulse trains were periodic pulse trains with constant amplitude. Stochastic inputs were trains of pulses generated by poisson processes.Results:The responses to the stochastic pulse trains exhibited the same fractal characteristics as the spike trains of an acoustically stimulated nerve fiber. For long counting times Fano-factors increased as a power-law function of the counting time. On the other hand, the responses to the deterministic pulse trains did not exhibit the same properties as the acoustically stimulated nerve fiber.Conclusion and discussion:From our simulations it appears to be unnecessary to include fractal fluctuations in the stimulation in auditory prostheses, as long as the stimulation is stochastic.Previously, it has been suggested that the fractal effects in spike trains were due to random fluctuations in the release of neurotransmitter. Our results suggest that these effects are due to the random fluctuations of the cross-membrane voltage. Yet the former does not exclude the latter.References:1. Lowen SB and Teich MC.: Toward fractal coding in auditory prostheses. Cochlear Implants, edited by S. B. Waltzman and N. Cohen. 2000. Thieme Medical Publishers, New York. 57-59. 2. Verveen AA. et al.: Fluctuation Phenomena in Nerve Membrane. Proc. IEEE 1968; 56: 906-916.