Neuron Spike Detection And Spike Sorting Algorithm

The nervous system controls the activities of biological muscles,coordinates various tissues and organs,establishes and receives outside intelligence and coordinates them.The nervous system is the most important animal's liaison and control system.Nervous system distributes frontal potential through neurons to transmit information.Analysis of spike sequence is an important prerequisite for analyzing neuroscience.Extracellular recording is a technique that inserts electrodes into brain tissue to record individual neuronal activity and is a common method used by neuroscientists to study how the brain works.These electrodes record information about the activity of multiple neurons around them.The spike sorting is the process of decomposing this signal into individual neuronal activity information.Since the neural signal is susceptible to external noise,the signal-to-noise ratio of the collected neural signal is relatively low.Multiple neurons existing around the electrodes send out spike signals at the same time to form overlapping spikes.These have increased the difficulty of the spike sorting.In order to analyze the neural signal better,it is very important to improve the robustness and accuracy of the spike sorting.In this paper,aiming at low signal-to-noise ratio signal and a large number of overlapped spike signals,the corresponding algorithms are proposed from three aspects: spike detection,spike feature representation and spike classification strategy.Specific algorithm improvements are as follows:1.The traditional threshold detection method has a lot of problems of missing detection and false detection when dealing with low signal-to-noise ratio signals.The window detection method is easy to falsely detect the overlapping spike in the window.In order to improve signal-to-noise ratio and identify overlapping spike signal,this paper proposes an improved window detection algorithm,adding energy detection at the time of overlapping signal distribution as a second detection.2.Dictionary learning methods can always try to learn the most rustic features behind the sample.In order to enhance the feature ability of the dictionary learning method for the spike signal,the resulting features are easier to distinguish the different types of spikes accurately.In this paper,a dictionary learning method similar to neural network structure is used,which increases the parsing layer of dictionary learning and enhances the feature ability of dictionary learning.3.In the clustering process,in order to reduce the impact of the boundary points on the clustering,this paper uses fuzzy clustering method.By calculating the fuzzy membership of the boundary point,to improve the classification effect.At the same time,combined with the density of characteristic sample points,the position of outliers is judged.Overlapping spike signals are often treated as outliers,and the signals are superimposed by several spikes using template matching,increasing the accuracy of the overlapping spike classification.Based on the simulation data set and the real data set,the paper tests the algorithm.The test results show that the classification accuracy of both Easy and Difficult data is above 90%,and the signal with different signal-to-noise ratio has good robustness.For overlapping spikes,the classification results improve accuracy by nearly 20% over the traditional method.Finally,the real biological data test,the method of this paper can identify overlapping spikes,has a good practical application.