Spike Sorting

In neural system, as to spike, due to the all or none property of spike generation and no attenuation during spike transmission, which is very similar to digital signal, neurons encode information about environment into spike trains. For example, the study on the spike rate and inter-spike interval of spike trains may shed light on unveiling the mechanism of neural signal processing. It has been discovered that the firing patter of action potential relates closely to the behavior. Moreover, the spike firing malfunction would cause several serious brain disease, e.g. Epilepsy and Depression. For this reason, research on the firing patter of the spike is very crucial for us to understand the way how the neural system process information. Neuron as a fundamental unit for structure and function, its firing pattern (i.e., single unit) get considerable interest of neurophysiologist. Appropriately, the classification of spikes and assignment of spikes to corresponding neuron will play an important role in the study of neural information processing.As a traditional method to acquire neural signal, extracellular single electrode recording is still widely used in current neurophysiology. Because the extracellular signal is generally contaminated by high level of background noise and Single electrode probably acquires signal of more than one neuron which is depending on the resistance and shape of the tip of electrode, How to extract credible spike and classify it accurately has always been the research focus. Recent years, multi-electrode array become more and more popular for electrophysiological experiment. Multi-electrode array with various arrangements of record sites and for variable research purposes has already been developed. Tetrode, a stereo multi-electrode array enables us to obtain the signal of single neuron by at least four electrodes simultaneously. Thus through combining all information from four channels, it can significantly improve the correction of classification. In this paper, I study the algorithm for spike detection and classification for single channel. Then I give some improvement for traditional algorithm. After improvement, the algorithm effectively decreases the background noise to a low level and prevents the interference of signal superposition. Consequently, it enhances the detection rate and accuracy of classification. In the last part, using a detailed algorithm, I interpret how to perform detection and classification of spikes from multi electrode array. The main work and characteristic are as follows:1. Improved a signal channel spike detection algorithm with mathematical morphology.The main problem of spike detection is the indeterminacy noise. In order to reduce the noise disturbance, we introduced the mathematical morphology into the step of signal channel spike detection.2. Research an improved a feature extraction method of spike waveforms withwavelet transform and kernel principal components analysis.Review several techniques of the feature extraction of spike waveforms, and design a new method of it with wavelet transform and kernel principal components analysis.3. Improve an automatic spike clusting algorithm with hierarchical clustering. Review several techniques of the step of spike clusting, and improve an automatic spike clusting algorithm with hierarchical clustering.4. Introduced a multi-channel spike-sorting algorithm. Illustrate the basic problems and issues involved in multi-channel spike sorting, and address these problems with a specific multi-channel spike-sorting algorithm.The simulation signal results indicate that the spike sorting algorithm with the above steps can improve the accuracy rates of spike detection and spike clusting. It is very important in the relevant experiments.