Spike-sorting Based On Genetic Algorithm-support Vector Machine And Characteristic Analysis

The visual system of higher animals is the most effective way of receiving external information as the most important sensory system. The visual information is processed and handled firstly in Primary visual cortex, almost 90% visual information from the retina is dealt in this area. The information about neurons activities of nervous system can be carried by visual cortex neurons spikes, which are the carrier of neurons information. So the research of neurons spikes is significant important for exploring the visual information transmission and coding and explaining the working mechanisms of the brain.Based on this background, neurons spikes were collected from the primary visual cortex of the rats. Then spike-sorting algorithm based on genetic algorithm-support vector machine was proposed, the characteristics of neurons from the aspects of synchrony and orientations selective were analyzed lastly.Specific content as follows:(1) Data collection. The primary visual cortex of rat was determined as the best position for visual signal acquisition through understanding the passing access and transmission mechanism of animal visual information and the feature of information transmission of primary visual cortex. On the existing experimental platform basis, neurons spikes of rat's primary visual cortex were acquisited.(2)Spike-sorting. The algorithm based on genetic algorithm-support vector machine(GA-SVM) was proposed for spike-sorting, aiming at the collected neurons signal. Support vector machine is suitable for finite sample classification and has the characteristic of the global optimal solution, combined with genetic algorithm's stochastic optimization ability. Spike-sorting can be realized by GA-SVM, and higher classification accuracy was obtained. It is demonstrated that the results of this classification algorithm are stable through contrasting several test data. Template matching method was used to spike-sorting further by extracting typical spike templates from spike-sorting results which based on genetic algorithm-support vector machine, higher classification accuracy was achieved when the signal-noise rate is lower.(3)Research for the characteristics of neurons. The correlation analysis method was applied respectively to the same channel among different classes and the various channels to realize synchrony analysis. The conclusion of existing synchronicity when neurons fire obtaining can be got; For the orientations selective of the primary visual cortex neurons receive field, the best stimulation orientations of a part of neurons can be found, but others not.