Study On The High Performance Neural Spike Detection

As the most mysterious organ in the organism,the brain not only dominates various biological behaviors,but also controls advanced cognitive functions such as learning,memory,thinking and so on.In the past century,a great deal of researches have been done to explore the mysteries of the brain and reveal the way how the brain works.It is found that in the complex nervous system,spikes generated by neurons are the main carrier to take and transfer information.Through studying and monitoring the neural spike activity,one can understand the neuronal coding mechanism and the interactions between single neurons,with which various complex brain functions can be further analyzed.Therefore,neural spike activity has attracted more and more attention and become an important research hotspot in the field of neuroscience.Nowadays neural spikes can be collected by extracellular recording with microelectrodes,which is one of the most widely used techniques for neuroscience research.Generally,it is a multi-stage process to analysis and acquisition of the neuronal activity information hidden in the recorded signal,including spike detection,feature extraction and spike sorting.As the first and the most important stage,spike detection aims to detect and find the occurrence times of individual spikes,which promises to be able to extract necessary spike data from the recorded signal for subsequent analyses.In view of the fact that the accuracy of spike detection will directly affect the reliability of all subsequent analyses,the research on effective spike detection method is of great significance.According to the way of data processing,the existing spike detection methods can be roughly divided into two types: online spike detection and offline spike detection.Online spike detection methods are simple to calculate and mainly used in real-time applications,such as brain disorder detection,brain-computer interface(BCI)and so on.Differently,offline spike detection methods have no requirements on computational complexity.In order to obtain a higher detection rate,this kind of methods are usually at the expense of running speed.Although the great progress has been made,the existing methods still cannot fully meet the practical demands.Aiming to overcome the shortcomings of existing spike detection methods,we have made some improvements in this paper.The main contributions are summarized as follows.(1)An online spike detection method based on improved differential operator is proposed.The existing spike detection approaches usually confront the contradiction between the run-ning speed and accuracy,and it is very difficult to achieve a tradeoff between these two factors.Focus on this issue,the authors propose a high efficiency algorithm for real-time spike detection.Researches on neuroscience discover that neural activity results in a rapid sharp rise.Inspired by this,we attempt to accentuate spikes in the signal using the famous differential operator,which exhibits a strong capacity to detect significant changes with slight computational burden.To this end,we exploit the structural features of spikes,and propose an improved differential operator to realize spike enhancement.Furthermore,considering the influence of strong background noise,a simple and effective measure is introduced to further suppress background noise with respect to spike events,which makes spike detection more accurate.Experimental results show that the proposed method is able to achieve better performance while maintaining low computational complexity.(2)An object-dependent sparse representation framework for high-accuracy and robust spike detection is proposed.Specifically,by exploiting the structural similarities of spikes,we construct an object-dependent dictionary to achieve a sparse and comprehensive representation of the recorded signals.Thus,the problem of spike detection can be formulated as a sparse representation model,which is a convex sparse optimization problem.Through systematically analyzing the optimal solution,the number and location of spikes in the recorded signal can be determined.In addition,singular value decomposition is introduced to further improve the adaptivity of the proposed method.Experimental results show that the proposed method outperforms the existing methods.(3)A spike detection method based on the theory of sparse representation and morphological component analysis is proposed.A high-accuracy detection method should satisfy two demanding requirements at the same time.On the one hand,it should be robust to the background noise.On the other hand,as the spikes from different neurons have different shapes,the detector should have strong adaptivity in the spike shape.To this end,we present a novel spike detection method based on the theory of sparse representation and morphological component analysis.Firstly,we construct a discriminative and universal dictionary to give a sparse representation to various spike signals and background noise.Then,the sparse coefficients can be extracted as features for the identification of spikes.By comprehensively analyzing the sparse coefficient features,the location and number of spikes in the signal can be obtained.Experimental results validate that the proposed method achieves much better results than the existing methods in terms of both robustness and adaptivity.