Research On Spike Sorting And Noise To Frequency Synchronization Of Neuron Model

The release and transfer of the neuron action potential (spike) is the basis of which the brainnervous system communicating and processing. Spike sorting is an important premise for theresearch about information encoding mechanism of the brain neural system. With the developmentof multi-electrode arrays technology (MEA), it provides necessary experimental foundation for theresearch on neural information encoding mechanism. However, signals collected through MEA arethe overlay of spikes emitted by several neurons around electrodes and noise of inside and outsidein neural system. Therefore, it is necessary to detect valid spikes from original signals collectedthrough MEA and sort spikes fast and exactly in order to complete the reconstruction of emittedsequence of corresponding neurons.For the nonlinearity and nonstationarity of spikes and the correlation dimension being anapproach of measuring the irregularity of signal waveforms, this paper proposed a new method ofextracting spike features which based on correlation dimension and combined K-means to realizethe spike sorting. Because of the distinguish ability of correlation dimension extraction, this paperproposed a new algorithm framework which was based on the multi-correlation dimension featureextraction and combined with the fuzzy C-means sorting algorithm. The results of these two sortingalgorithms are perfect which have been verified through simulation data and real data. Because ofthe uncontrollability of noise inside and outside Neural System in true experiments, it is difficult toconfirm the effects of noise on the relationship between the input and output of neuron system.However, the mathematical model of neural system can help people to understand problems andphenomena appeared in experiments, and it is able to make up the deficiencies of true experimentsthrough quantitative analysis. Therefore, this paper studied the effects about noise on the frequencysynchronization of input and output of neurons by Hodgkin-Huxley (HH) neuron model.The main work and research results of this paper are as follows:(1) This paper proposed a new sorting algorithm based on correlation dimension featureextraction and showed the basis of selecting the delay time、embedding dimension and other keyparameters in the calculation progress of correlation dimension .Besides that, it was also found thatfeatures extracted based on correlation dimension have a hierarchy phenomenon. The result of newalgorithm is good which have been verified through simulation data and real data. The experimentalresults show that features extracted based on correlation dimension are able to express anddistinguish the nonlinearity and nonstationarity of nonhomologous spikes. Through the comparisonwith other sorting means, it indicates the advantages of the new sorting algorithm such as high sorting accuracy、good reliability and so on.(2) This paper proposed a new algorithm framework based on multi-correlation dimensionfeature extraction and combined with fuzzy C-means. It gave the minimum embedding dimensionmethod used in reconstructed phase space to show the nonlinear dynamic characteristics of spikewaveforms. Multi-correlation dimension feature vectors can be obtained through adjusting theembedding dimensions. Those feature vectors that have high discrimination can be picked outthrough KS inspection. Simulation data and real data indicate that new algorithm frame has highsorting accuracy that makes feasibility of manual sorting being replaced. At the same time,F-measure is introduced as the evaluation index about the function of new algorithm frame.Through comparison with one-dimensional correlation dimension features and sorting algorithm, itis confirmed that the function of new algorithm framework is better.(3) Based on the Hodgkin-Huxley (HH) Neuron model, this paper constructed a double-layerneuron network model and studied the effects of noise on frequency synchronization of the inputand output, which is difficult to be analyzed of the uncontrollability of noise in true experiment. Theresults of experiment show that appropriate noise intensity can enhance the rhythmicity of theHH-model response and increase the frequency synchronization of input and output. But the rangeof favorable noise intensity is narrow and is different for diverse input objects. These phenomenawill help people find further information on the neurons encoding mechanism.