Multineuronal spike classification based on multisite electrode recording, whole-waveform analysis, and hierarchical clustering

We proposed here a method of multineuronal spike classification based on multisite electrode recording, whole-waveform analysis, and hierarchical clustering for studying correlated activities of adjacent neurons in nervous systems. Multineuronal spikes were recorded with a multisite electrode placed in the hippocampal pyramidal cell layer of anesthetized rats. If the impedance of each electrode site is relatively low and the distance between electrode sites is sufficiently small, a spike generated by a neuron is simultaneously recorded at multielectrode sites with different amplitudes. The covariance between the spike waveform at each electrode site and a template was calculated as a damping factor due to the volume conduction of the spike from the neuron to the electrode site. Calculated damping factors were vectorized and analyzed by hierarchical clustering using a multidimensional statistical test. Since a cluster of damping vectors was shown to correspond to an antidromically identified neuron, spikes of different neurons are classified by referring to the distributions of damping vectors. Errors in damping vector calculation due to partially overlapping spikes were minimized by successively subtracting preceding spikes from raw data. Clustering errors due to complex spike bursts (i.e., spikes with variable amplitudes) were avoided by detecting such bursts and then using only the first spike of a burst for clustering. These special procedures produced better cluster separation than conventional methods, and enabled multiple neuronal spikes to be classified automatically. Waveforms of classified spikes were well superimposed. We concluded that this method is particularly useful for separating the activities of adjacent neurons that fire partially overlapping spikes and/or complex spike bursts.     

基于脉冲序列核的脉冲神经元监督学习算法

脉冲神经元应用脉冲时间编码神经信息,监督学习的目标是对于给定的突触输入产生任意的期望脉冲序列.但由于神经元脉冲发放过程的不连续性,构建高效的脉冲神经元监督学习算法非常困难,同时也是该研究领域的重要问题.基于脉冲序列的核函数定义,提出了一种新的脉冲神经元监督学习算法,特点是应用脉冲序列核构造多脉冲误差函数和对应的突触学习规则,并通过神经元的实际脉冲发放频率自适应地调整学习率.将该算法用于脉冲序列的学习任务,期望脉冲序列采用Poisson过程或线性方法编码,并分析了不同的核函数对算法学习性能的影响.实验结果表明该算法具有较高的学习精度和良好的适应能力,在处理复杂的时空脉冲模式学习问题时十分有效.     

压缩感知的重叠神经元峰电位信号分类

随着电生理技术水平的提高,电极可以记录的峰电位信号包含多个神经元峰电位的叠加.本文提出了一种采用压缩感知和最大后验估计的分类算法来解决重叠峰电位分类问题.其中压缩感知算法用于得到稀疏信号,最大后验估计用于搜索出稀疏信号的最优解.在实验中,我们采用仿真和实测的三组数据对本文算法和传统算法进行了测试,实验结果表明,当峰电位波形相似时,相比于k-均值聚类及CBP（Continuous Basis Pursuit）算法,本文算法具有较少的分类错误数.     

一种高效的分布式多传感器多目标跟踪算法

现有的多传感器多目标跟踪算法大都基于马尔科夫-贝叶斯模型,需要诸如目标运动、杂波、传感器检测概率等先验信息,但是在恶劣的环境中,这些先验信息不准确并导致目标跟踪精度下降.为了解决该情况下的多目标跟踪问题,我们提出了一个高效的分布式多目标跟踪算法,该算法通过泛洪(Flooding)共识算法在分布式网络的传感器之间迭代的传...     

Analysis of raw microneurographic recordings based on wavelet de-noising technique and classification algorithm: wavelet analysis in microneurography

We propose a new technique for analyzing the raw neurogram which enables the study of the discharge behavior of individual and group neurons. It utilizes an ideal bandpass filter, a modified wavelet de-noising procedure, an action potential detector, and a waveform classifier. We validated our approach with both simulated data generated from muscle sympathetic neurograms sampled at high rates in five healthy subjects and data recorded from seven healthy subjects during lower body negative pressure suction. The modified wavelet method was superior to the classical discriminator method and the regular wavelet de-noising procedure when applied to simulated neuronal signals. The detected spike rate and spike amplitude rate of the action potentials correlated strongly with number of bursts detected in the integrated neurogram (r = 0.79 and 0.89, respectively, p < 0.001). Eight major action potential waveform classes were found to describe more than 81% of all detected action potentials in all subjects. One class had characteristics similar in shape and in average discharge frequency (27.4 +/- 5.1 spikes/min during resting supine position) to those of reported single vasoconstrictor units. The newly proposed technique allows a precise estimate of sympathetic nerve activity and characterization of individual action potentials in multiunit records.     

Recognition of multiunit neural signals

An essential step in studying nerve cell interaction during information processing is the extracellular microelectrode recording of the electrical activity of groups of adjacent cells. The recording usually contains the superposition of the spike trains produced by a number of neurons in the vicinity of the electrode. It is therefore necessary to correctly classify the signals generated by these different neurons. This paper considers this problem, and a new classification scheme is developed, which does not require human supervision. A learning stage is first applied on the beginning portion of the recording to estimate the typical spike shapes of the different neurons. As for the classification stage, a method is developed, which specifically considers the case when spikes overlap temporally. The method minimizes the probability of error, taking into account the statistical properties of the discharges of the neurons. The method is tested on a real recording as well as on synthetic data.     

基于克隆算法的网络结构聚类新算法

基于目标函数的聚类算法是目前应用最为广泛的聚类分析方法之一.然而这类算法都需要类别数和聚类原型的先验知识,且只能分析具有相同原型的数值型数据.此外这类算法还存在对初始化敏感,易陷入局部极值点等弱点.为此,本文提出一种基于克隆算法的网络结构聚类新算法以实现聚类分析的自动化.由于新算法将克隆选择与禁忌克隆相结合,使网络既具有免疫的特异性又具有免疫的耐受性,通过分析网络神经元的最小生成树,能够快速准确地获得类别数以及相关的分类信息.对各种类型的数据集的测试结果均表明,本文提出的新算法对于处理具有混和特征的数据集聚类分析问题是相当便捷有效的.     

自适应AP聚类算法及其在入侵检测中的应用

网络数据流量的增大对入侵检测系统的实时性提出了更高的要求,压缩训练数据可加快未知样本的分类处理速度。针对数据量过大造成压缩处理和聚类效率低下的难题,提出了一种改进的自适应AP（affinity propagation）聚类方法,采取直接关联与簇中心距离较近样本的方法,减少聚类样本数量,降低聚类时空消耗,并依据关联结果,不断调整聚类参数,精确聚类结果。2个网络安全数据集的应用结果表明,该方法可从大规模样本中有效聚出代表性子集,在保证准确率的前提下,提高入侵检测的实效性。     

一种改进的道路行车密度峰值模糊聚类算法

为解决城市道路中相邻车辆聚类精度低的问题,本文提出了一种改进的密度峰值模糊聚类算法。首先,该算法使用自适应椭圆距离代替欧式距离,并在决策图中引入指数函数曲线选择密度峰值点,以确定初始聚类中心和聚类数目;接着,将初始信息代入模糊C均值(FCM)聚类算法中,经迭代计算取得一次聚类结果;最后,根据雷达数据中同一辆车的数据点速度差值极小、不同车辆的速度差值相对较大这一特征,引入和速度相关的目标函数,并通过迭代计算取得最终的聚类结果,以对一次聚类结果进行修正。根据真实道路测量数据的实验证明,本文提出的聚类算法精度高、鲁棒性好,能正确聚类城市道路中相邻的车辆目标,具有更好的聚类效果。为道路中车辆的跟踪、交通状态预估等处理提供可靠、准确的目标信息,大大减少后续工程的计算量。     

Encoding of information into neural spike trains in an auditory nerve fiber model with electric stimuli in the presence of a pseudospontaneous activity

This paper presents an information-theoretic analysis of neural spike trains in an auditory nerve fiber (ANF) model stimulated extracellularly with Gaussian or sinusoidal waveforms in the presence of a pseudospontaneous activity of spike firings. In the computer simulation, stimulus current waveforms were applied repeatedly to a stimulating electrode located 1 mm above the 26th node of Ranvier, in an ANF axon model having 50 nodes of Ranvier, each consisting of stochastic sodium and potassium channels. From spike firing times recorded at the 36th node of Ranvier, a post-stimulus time histogram (PSTH) was generated, and raster plots were depicted for 30 stimulus presentations, in order to investigate the temporal precision and reliability of the spike firing times. Also, inter spike intervals were generated and then "total" and "noise" entropies were estimated to obtain the mutual information and the information rate of the spike trains. It was shown in the case of Gaussian electric stimuli that the temporal precision of spike firing times and the reliability of spike firings were found to increase as the standard deviation (SD) of the Gaussian electric stimuli increased. It was also shown in the case of sinusoidal electric stimuli where there was a specific amplitude of sinusoidal waveforms, the information rate being maximized. It was implied that setting the parameters of electric stimuli to the specific values which maximize the information rate might contribute to efficiently encoding information into the spike trains in the presence of a pseudospontaneous activity of spike firings.     

谱聚类划分随机森林算法在图像识别中的应用

随机森林是近些年发展起来的新集成学习算法,具有较好的分类准确率。针对该算法计算复杂度较高的不足,提出了一种基于谱聚类划分的随机森林算法。首先,利用聚类效果较好的谱聚类算法对原始样本集的每一类进行聚类处理。然后,在每一聚类簇中随机选取一个样本作为代表,组成新训练样本集合。最后,在新训练样本集上训练随机森林分类器。该算法通过谱聚类技术对原始样本进行了初步划分,将位置相近的多个样本用簇内的一个样本代表,较大程度地减少了训练样本的个数。在Corel Image图像识别数据集上的实验表明,算法可以用较少的分类时间达到较高的分类精度。     

On the variability of manual spike sorting

The analysis of action potentials, or "spikes," is central to systems neuroscience research. Spikes are typically identified from raw waveforms manually for off-line analysis or automatically by human-configured algorithms for on-line applications. The variability of manual spike "sorting" is studied and its implications for neural prostheses discussed. Waveforms were recorded using a micro-electrode array and were used to construct a statistically similar synthetic dataset. Results showed wide variability in the number of neurons and spikes detected in real data. Additionally, average error rates of 23% false positive and 30% false negative were found for synthetic data.     

Detection, classification, and superposition resolution of actionpotentials in multiunit single-channel recordings by an on-linereal-time neural network

Determination of single-unit spike trains from multiunit recordings obtained during extracellular recording has been the focus of many studies over the last two decades. In multiunit recordings, superpositions can occur with high frequency if the firing rates of the neurons are high or correlated, making superposition resolution imperative for accurate spike train determination. In this work, a connectionist neural network (NN) was applied to the spike sorting challenge. A novel training scheme was developed which enabled the NN to resolve some superpositions using single-channel recordings. Simulated multiunit spike trains were constructed from templates and noise segments that were extracted from real extracellular recordings. The simulations were used to determine the performances of the NN and a simple matched template filter (MTF), which was used as a basis for comparison. The network performed as well as the MTF in identifying nonoverlapping spikes, and was significantly better in resolving superpositions and rejecting noise. An on-line, real-time implementation of the NN discriminator, using a high-speed digital signal processor mounted inside an IBM-PC, is now in use in six laboratories     

Spike Recognition and On-Line Classification by Unsupervised Learning System

An on-line spike recognition system allows separation of multiple spikes present on a single channel, in up to six different classes. The learning phase is unsupervised, and uses the data samples of the waveform as coordinates in a multidimensional feature space. Additional signal characteristics may improve the system performance in special cases. Using the well known nearest neighbor technique, all possible cluster configurations are determined. From this analysis, the investigator selects the physiologically best suited duster layout, primary based on a curve showing the number of clusters versus the maximum distance of two neighboring spikes in the same cluster. This procedure is supported by visual examination of the spikes of each cluster. Statistics are calculated for inter-and intracluster distances, yielding confidence limits for the cluster bounds, and estimates for the quality of separation. During the classification phase, a separate graphic display processor permits continuous control without delay. Each classified spike is projected over its cluster, identifying mean waveform.     

Neural spike train synchronization indices: definitions, interpretations, and applications

A comparison of previously defined spike train syncrhonization indices is undertaken within a stochastic point process framework. The second-order cumulant density (covariance density) is shown to be common to all the indices. Simulation studies were used to investigate the sampling variability of a single index based on the second-order cumulant. The simulations used a paired motoneurone model and a paired regular spiking cortical neurone model. The sampling variability of spike trains generated under identical conditions from the paired motoneurone model varied from 50% to 160% of the estimated value. On theoretical grounds, and on the basis of simulated data a rate dependence is present in all synchronization indices. The application of coherence and pooled coherence estimates to the issue of synchronization indices is considered. This alternative frequency domain approach allows an arbitrary number of spike train pairs to be evaluated for statistically significant differences, and combined into a single population measure. The pooled coherence framework allows pooled time domain measures to be derived, application of this to the simulated data is illustrated. Data from the cortical neurone model is generated over a wide range of firing rates (1-250 spikes/s). The pooled coherence framework correctly characterizes the sampling variability as not significant over this wide operating range. The broader applicability of this approach to multielectrode array data is briefly discussed.     

Detection and classification of sensory information from acute spinal cord recordings

One avenue of research for partial restoration of function following spinal cord injury is the use of neural prostheses, an example of which is functional electrical stimulation (FES) devices for motor functions. Neural prostheses may also be useful for the extraction of sensory information directly from the nervous system. We suggest the spinal cord as a possible site for the detection of peripheral sensory information from neural activity alone. Acute multichannel extracellular recordings were used to extract neural spike activity elicited from peripheral sensations from the spinal cords of rats. To test the recording method and classification potential, eight classes of sensory events were recorded consisting of electrical stimulation of seven locations on rat forepaws, and another class of data during which no stimulus was present. A dual-stage classification scheme using principal component analysis and k-Means clustering was devised to classify the sensory events during single trials. The eight tasks were correctly identified at a mean accuracy of 96%. Thus, we have shown the methodology to detect and classify peripheral sensory information from multichannel recordings of the spinal cord. These methods may be useful, for example, in a closed-loop FES for restoration of hand grasp.     

基于形态匹配聚类的近邻扩展目标跟踪算法

针对传统扩展目标跟踪（Extended Target Tracking, ETT）算法在处理近邻目标时面临的计算效率低下和跟踪不准确的问题,提出了一种形态匹配聚类量测集划分与高斯逆威沙特概率假设密度（Gaussian Inverse Wishart Probability Hypothesis Density, GIW-PHD）滤波器相结合的跟踪处理方法。该方法首先由GIW-PHD滤波器得到预测的目标状态,其次使用DBSCAN(Density-Based Spatial Clustering of Applications with Noise, DBSCAN)算法完成量测集的初步划分,在此基础上利用较高权重的预测分量实现对多个近邻目标混合量测簇的判断,进而使用椭圆形状约束（Elliptic Shape Constraint, ESC）的FCM(Fuzzy C-Means, FCM)算法（ESC-FCM）对混合簇进行二次划分得到更精确的划分结果,最后将划分结果合并后送入GIW-PHD滤波器完成目标状态的更新。仿真结果表明,本文所提量测集划分方法能够充分利用GIW-PHD滤波器预测步获取...     

Methods for robust clustering of epileptic EEG spikes

We investigate algorithms for clustering of epileptic electroencephalogram (EEG) spikes. Such a method is useful prior to averaging and inverse computations since the spikes of a patient often belong to a few distinct classes. Data sets often contain outliers, which makes algorithms with robust performance desirable. We compare the fuzzy C-means (FCM) algorithm and a graph-theoretic algorithm. We give criteria for determination of the correct level of outlier contamination. The performance is then studied by aid of simulations, which show good results for a range of circumstances, for both algorithms. The graph-theoretic method gave better results than FCM for simulated signals. Also, when evaluating the methods on seven real-life data sets, the graph-theoretic method was the better method, in terms of closeness to the manual assessment by a neurophysiologist. However, there was some discrepancy between manual and automatic clustering and we suggest as an alternative method a human choice among a limited set of automatically obtained clusterings. Furthermore, we evaluate geometrically weighted feature extraction and conclude that it is useful as a supplementary dimension for clustering.     

基于聚类支持向量机的入侵检测算法

针对支持向量机应用到入侵检测中训练时间长的特点,提出了一种基于聚类的支持向量机的入侵检测算法。该方法可以对训练数据进行剪枝,以靠近判别边界的聚类中心集合作为有效的训练样本集合对支持向量机进行训练,减少了样本的训练时间,提高了算法的效率。实验结果表明该方法对入侵检测是有效的。     

Blind image clustering based on the Normalized Cuts criterion for camera identification

Camera identification is a well known problem in image forensics, addressing the issue to identify the camera a digital image has been shot by. In this paper, we pose our attention to the task of clustering images, belonging to a heterogenous set, in groups coming from the same camera and of doing this in a blind manner; this means that side information neither about the sources nor, above all, about the number of expected clusters is requested. A novel methodology based on Normalized Cuts (NC) criterion is presented and evaluated in comparison with other state-of-the-art techniques, such as Multi-Class Spectral Clustering (MCSC) and Hierarchical Agglomerative Clustering (HAC). The proposed method well fits the problem of blind image clustering because it does not a priori require the knowledge of the amount of classes in which the dataset has to be divided but it needs only a stop threshold; such a threshold has been properly defined by means of a ROC curves approach by relying on the goodness of cluster aggregation. Several experimental tests have been carried out in different operative conditions and the proposed methodology globally presents superior performances in terms of clustering accuracy and robustness as well as a reduced computational burden.