Computer separation of multi-unit neuroelectric data: a review

Continuing progress of spike sorting in the era of big data

Engineering efforts are currently attempting to build devices capable of collecting neural activity from one million neurons in the brain. Part of this effort focuses on developing dense multiple-electrode arrays, which require post-processing via 'spike sorting' to extract neural spike trains from the raw signal. Gathering information at this scale will facilitate fascinating science, but these dreams are only realizable if the spike sorting procedure and data pipeline are computationally scalable, at or superior to hand processing, and scientifically reproducible. These challenges are all being amplified as the data scale continues to increase. In this review, recent efforts to attack these challenges are discussed, which have primarily focused on increasing accuracy and reliability while being computationally scalable. These goals are addressed by adding additional stages to the data processing pipeline and using divide-and-conquer algorithmic approaches. These recent developments should prove useful to most research groups regardless of data scale, not just for cutting-edge devices.

Spike sorting based on shape, phase, and distribution features, and K-TOPS clustering with validity and error indices

Spike sorting is one of the most important data analysis problems in neurophysiology. The precision in all steps of the spike-sorting procedure critically affects the accuracy of all subsequent analyses. After data preprocessing and spike detection have been carried out properly, both feature extraction and spike clustering are the most critical subsequent steps of the spike-sorting procedure. The proposed spike sorting approach comprised a new feature extraction method based on shape, phase, and distribution features of each spike (hereinafter SS-SPDF method), which reveal significant information of the neural events under study. In addition, we applied an efficient clustering algorithm based on K-means and template optimization in phase space (hereinafter K-TOPS) that included two integrative clustering measures (validity and error indices) to verify the cohesion-dispersion among spike events during classification and the misclassification of clustering, respectively. The proposed method/algorithm was tested on both simulated data and real neural recordings. The results obtained for these datasets suggest that our spike sorting approach provides an efficient way for sorting both single-unit spikes and overlapping waveforms. By analyzing raw extracellular recordings collected from the rostral-medial prefrontal cortex (rmPFC) of behaving rabbits during classical eyeblink conditioning, we have demonstrated that the present method/algorithm performs better at classifying spikes and neurons and at assessing their modulating properties than other methods currently used in neurophysiology.

Automatic spike sorting for high-density microelectrode arrays

High-density microelectrode arrays can be used to record extracellular action potentials from hundreds to thousands of neurons simultaneously. Efficient spike sorters must be developed to cope with such large data volumes. Most existing spike sorting methods for single electrodes or small multielectrodes, however, suffer from the "curse of dimensionality" and cannot be directly applied to recordings with hundreds of electrodes. This holds particularly true for the standard reference spike sorting algorithm, principal component analysis-based feature extraction, followed by k-means or expectation maximization clustering, against which most spike sorters are evaluated. We present a spike sorting algorithm that circumvents the dimensionality problem by sorting local groups of electrodes independently with classical spike sorting approaches. It is scalable to any number of recording electrodes and well suited for parallel computing. The combination of data prewhitening before the principal component analysis-based extraction and a parameter-free clustering algorithm obviated the need for parameter adjustments. We evaluated its performance using surrogate data in which we systematically varied spike amplitudes and spike rates and that were generated by inserting template spikes into the voltage traces of real recordings. In a direct comparison, our algorithm could compete with existing state-of-the-art spike sorters in terms of sensitivity and precision, while parameter adjustment or manual cluster curation was not required. NEW & NOTEWORTHY We present an automatic spike sorting algorithm that combines three strategies to scale classical spike sorting techniques for high-density microelectrode arrays: 1) splitting the recording electrodes into small groups and sorting them independently; 2) clustering a subset of spikes and classifying the rest to limit computation time; and 3) prewhitening the spike waveforms to enable the use of parameter-free clustering. Finally, we combined these strategies into an automatic spike sorter that is competitive with state-of-the-art spike sorters.

Sub 100 nW Volatile Nano-Metal-Oxide Memristor as Synaptic-Like Encoder of Neuronal Spikes

Advanced neural interfaces mediate a bioelectronic link between the nervous system and microelectronic devices, bearing great potential as innovative therapy for various diseases. Spikes from a large number of neurons are recorded leading to creation of big data that require online processing under most stringent conditions, such as minimal power dissipation and on-chip space occupancy. Here, we present a new concept where the inherent volatile properties of a nano-scale memristive device are used to detect and compress information on neural spikes as recorded by a multielectrode array. Simultaneously, and similarly to a biological synapse, information on spike amplitude and frequency is transduced in metastable resistive state transitions of the device, which is inherently capable of self-resetting and of continuous encoding of spiking activity. Furthermore, operating the memristor in a very high resistive state range reduces its average in-operando power dissipation to less than 100 nW, demonstrating the potential to build highly scalable, yet energy-efficient on-node processors for advanced neural interfaces.

From Maxwell's equations to the theory of current-source density analysis

Despite the widespread use of current‐source density (CSD) analysis of extracellular potential recordings in the brain, the physical mechanisms responsible for the generation of the signal are still debated. While the extracellular potential is thought to be exclusively generated by the transmembrane currents, recent studies suggest that extracellular diffusive, advective and displacement currents—traditionally neglected—may also contribute considerably toward extracellular potential recordings. Here, we first justify the application of the electro‐quasistatic approximation of Maxwell's equations to describe the electromagnetic field of physiological origin. Subsequently, we perform spatial averaging of currents in neural tissue to arrive at the notion of the CSD and derive an equation relating it to the extracellular potential. We show that, in general, the extracellular potential is determined by the CSD of membrane currents as well as the gradients of the putative extracellular diffusion current. The diffusion current can contribute significantly to the extracellular potential at frequencies less than a few Hertz; in which case it must be subtracted to obtain correct CSD estimates. We also show that the advective and displacement currents in the extracellular space are negligible for physiological frequencies while, within cellular membrane, displacement current contributes toward the CSD as a capacitive current. Taken together, these findings elucidate the relationship between electric currents and the extracellular potential in brain tissue and form the necessary foundation for the analysis of extracellular recordings.

Spike detection from noisy neural data in linear-probe recordings

Simultaneous recordings of multiple neuron activities with multi-channel extracellular electrodes are widely used for studying information processing by the brain's neural circuits. In this method, the recorded signals containing the spike events of a number of adjacent or distant neurons must be correctly sorted into spike trains of individual neurons, and a variety of methods have been proposed for this spike sorting. However, spike sorting is computationally difficult because the recorded signals are often contaminated by biological noise. Here, we propose a novel method for spike detection, which is the first stage of spike sorting and hence crucially determines overall sorting performance. Our method utilizes a model of extracellular recording data that takes into account variations in spike waveforms, such as the widths and amplitudes of spikes, by detecting the peaks of band-pass-filtered data. We show that the new method significantly improves the cost-performance of multi-channel electrode recordings by increasing the number of cleanly sorted neurons.

Spike sorting by joint probabilistic modeling of neural spike trains and waveforms

This paper details a novel probabilistic method for automatic neural spike sorting which uses stochastic point process models of neural spike trains and parameterized action potential waveforms. A novel likelihood model for observed firing times as the aggregation of hidden neural spike trains is derived, as well as an iterative procedure for clustering the data and finding the parameters that maximize the likelihood. The method is executed and evaluated on both a fully labeled semiartificial dataset and a partially labeled real dataset of extracellular electric traces from rat hippocampus. In conditions of relatively high difficulty (i.e., with additive noise and with similar action potential waveform shapes for distinct neurons) the method achieves significant improvements in clustering performance over a baseline waveform-only Gaussian mixture model (GMM) clustering on the semiartificial set (1.98% reduction in error rate) and outperforms both the GMM and a state-of-the-art method on the real dataset (5.04% reduction in false positive + false negative errors). Finally, an empirical study of two free parameters for our method is performed on the semiartificial dataset.

Neuronal Activity in the Subthalamic Cerebrovasodilator Area under Partial-Gravity Conditions in Rats

The reduced-gravity environment in space is known to cause an upward shift in body fluids and thus require cardiovascular adaptations in astronauts. In this study, we recorded in rats the neuronal activity in the subthalamic cerebrovasodilator area (SVA), a key area that controls cerebral blood flow (CBF), in response to partial gravity. “Partial gravity” is the term that defines the reduced-gravity levels between 1 g (the unit gravity acceleration on Earth) and 0 g (complete weightlessness in space). Neuronal activity was recorded telemetrically through chronically implanted microelectrodes in freely moving rats. Graded levels of partial gravity from 0.4 g to 0.01 g were generated by customized parabolic-flight maneuvers. Electrophysiological signals in each partial-gravity phase were compared to those of the preceding 1 g level-flight. As a result, SVA neuronal activity was significantly inhibited by the partial-gravity levels of 0.15 g and lower, but not by 0.2 g and higher. Gravity levels between 0.2–0.15 g could represent a critical threshold for the inhibition of neurons in the rat SVA. The lunar gravity (0.16 g) might thus trigger neurogenic mechanisms of CBF control. This is the first study to examine brain electrophysiology with partial gravity as an experimental parameter.

A model-based spike sorting algorithm for removing correlation artifacts in multi-neuron recordings

We examine the problem of estimating the spike trains of multiple neurons from voltage traces recorded on one or more extracellular electrodes. Traditional spike-sorting methods rely on thresholding or clustering of recorded signals to identify spikes. While these methods can detect a large fraction of the spikes from a recording, they generally fail to identify synchronous or near-synchronous spikes: cases in which multiple spikes overlap. Here we investigate the geometry of failures in traditional sorting algorithms, and document the prevalence of such errors in multi-electrode recordings from primate retina. We then develop a method for multi-neuron spike sorting using a model that explicitly accounts for the superposition of spike waveforms. We model the recorded voltage traces as a linear combination of spike waveforms plus a stochastic background component of correlated Gaussian noise. Combining this measurement model with a Bernoulli prior over binary spike trains yields a posterior distribution for spikes given the recorded data. We introduce a greedy algorithm to maximize this posterior that we call "binary pursuit". The algorithm allows modest variability in spike waveforms and recovers spike times with higher precision than the voltage sampling rate. This method substantially corrects cross-correlation artifacts that arise with conventional methods, and substantially outperforms clustering methods on both real and simulated data. Finally, we develop diagnostic tools that can be used to assess errors in spike sorting in the absence of ground truth.

Semi-supervised spike sorting using pattern matching and a scaled Mahalanobis distance metric

Sorting action potentials (spikes) from tetrode recordings can be time consuming, labor intensive, and inconsistent, depending on the methods used and the experience of the operator. The techniques presented here were designed to address these issues. A feature related to the slope of the spike during repolarization is computed. A small subsample of the features obtained from the tetrode (ca. 10,000-20,000 events) is clustered using a modified version of k-means that uses Mahalanobis distance and a scaling factor related to the cluster size. The cluster-size-based scaling improves the clustering by increasing the separability of close clusters, especially when they are of disparate size. The full data set is then classified from the statistics of the clusters. The technique yields consistent results for a chosen number of clusters. A MATLAB implementation is able to classify more than 5000 spikes per second on a modern workstation.

Fuzzy logic-based spike sorting system

We present a new method for autonomous real-time spike sorting using a fuzzy logic inference engine. The engine assigns each detected event a 'spikiness index' from zero to one that quantifies the extent to which the detected event is like an ideal spike. Spikes can then be sorted by simply clustering the spikiness indices. The sorter is defined in terms of natural language rules that, once defined, are static and thus require no user intervention or calibration. The sorter was tested using extracellular recordings from three animals: a macaque, an owl monkey and a rat. Simulation results show that the fuzzy sorter performed equal to or better than the benchmark principal component analysis (PCA) based sorter. Importantly, there was no degradation in fuzzy sorter performance when the spikes were not temporally aligned prior to sorting. In contrast, PCA sorter performance dropped by 27% when sorting unaligned spikes. Since the fuzzy sorter is computationally trivial and requires no spike alignment, it is suitable for scaling into large numbers of parallel channels where computational overhead and the need for operator intervention would preclude other spike sorters.

Spike sorting of muscle spindle afferent nerve activity recorded with thin-film intrafascicular electrodes

Afferent muscle spindle activity in response to passive muscle stretch was recorded in vivo using thin-film longitudinal intrafascicular electrodes. A neural spike detection and classification scheme was developed for the purpose of separating activity of primary and secondary muscle spindle afferents. The algorithm is based on the multiscale continuous wavelet transform using complex wavelets. The detection scheme outperforms the commonly used threshold detection, especially with recordings having low signal-to-noise ratio. Results of classification of units indicate that the developed classifier is able to isolate activity having linear relationship with muscle length, which is a step towards online model-based estimation of muscle length that can be used in a closed-loop functional electrical stimulation system with natural sensory feedback.

Fast automatic template matching for spike sorting based on Davies-Bouldin validation indices

The present study introduces an approach to detecting and classifying the extracellular action potentials of neurons, a process usually referred to as spike sorting. Our approach is based on template matching which is an optimal filter under Gaussian noise. However, this is usually expensive in terms of computational time, and constructing appropriate templates can be also problematic. Despite its theoretical consistency, only a few algorithms have been proposed to efficiently solve this problem. To speed up the filter, it is important to curtail the matching process when the distance between the template and waveform exceeds some threshold. We approach this aspect of the problem using Davies-Bouldin validation indices (DBVIs), which are a function of the ratio of the sum of within-cluster scatter to between-cluster separation to prioritize point-by-point calculation. The templates are also constructed automatically by combining principle component analysis (PCA) and k-means clustering. This matching process performed well, with a shorter computational time and fewer incorrect classifications than other ordering methods.

Spike sorting based on multi-class support vector machine with superposition resolution

A new spike sorting method based on the support vector machine (SVM) is proposed to resolve the superposition problem. The spike superposition is generally resolved by the template matching. Previous template matching methods separate the spikes through linear classifiers. The classification performance is severely influenced by the background noise included in spike trains. The nonlinear classifiers with high generation ability are required to deal with the task. A multi-class SVM classifier is therefore applied to separate the spikes, which contains several binary SVM classifiers. Every binary SVM classifier corresponding to one spike class is used to identify the single and superposition spikes. The superposition spikes are decomposed through template extraction. The experimental results on the simulated and real data demonstrate the utility of the proposed method.

EEG features extraction using neuro-fuzzy systems and shift-invariant wavelet transforms for epileptic seizures diagnosing

Electro-encephalogram Spikes Classification and latency computing is one of the important tools in epilepsy diagnosing. However, overlapped spikes cause complexity in problem solving. We use neuro-fuzzy systems and shift-invariant wavelet transforms to solve this problem. It has been shown that our suggested procedures have high-resolution and are able to classify and perform latency computing of overlapped spikes.

Evaluation of the stability of intracortical microelectrode arrays

In order to use recorded neural activities from the brain as control signals for neuroprosthesis devices, it is important to maintain a stable interface between chronically implanted microelectrodes and neural tissue. Our previous paper introduced a method to quantify the stability of the recording microelectrodes. In this paper, the method is refined 1) by incorporating stereotypical behavioral patterns into the spike sorting program and 2) by using a classifier based on Bayes theorem for assigning the recorded action potentials to the underlying neural generators. An improved method for calculating stability index is proposed. The results for the stability of microelectrode arrays that differ in structure are presented.

Nonauditory events of a behavioral procedure activate auditory cortex of highly trained monkeys

A central tenet in brain research is that early sensory cortex is modality specific, and, only in exceptional cases, such as deaf and blind subjects or professional musicians, is influenced by other modalities. Here we describe extensive cross-modal activation in the auditory cortex of two monkeys while they performed a demanding auditory categorization task: after a cue light was turned on, monkeys could initiate a tone sequence by touching a bar and then earn a reward by releasing the bar on occurrence of a falling frequency contour in the sequence. In their primary auditory cortex and posterior belt areas, we found many acoustically responsive neurons whose firing was synchronized to the cue light or to the touch or release of the bar. Of 315 multiunits, 45 exhibited cue light-related firing, 194 exhibited firing that was related to bar touch, and 268 exhibited firing that was related to bar release. Among 60 single units, we found one neuron with cue light-related firing, 21 with bar touch-related firing, and 36 with release-related firing. This firing disappeared at individual sites when the monkeys performed a visual detection task. Our findings corroborate and extend recent findings on cross-modal activation in the auditory cortex and suggests that the auditory cortex can be activated by visual and somatosensory stimulation and by movements. We speculate that the multimodal corepresentation in the auditory cortex has arisen from the intensive practice of the subjects with the behavioral procedure and that it facilitates the performance of audiomotor tasks in proficient subjects.

Algorithms and architectures for low power spike detection and alignment

We introduce algorithms and architectures for automatic spike detection and alignment that are designed for low power. Some of the algorithms are based on principal component analysis (PCA). Others employ a novel integral transform analysis and achieve 99% of the precision of a PCA detector, while requiring only 0.05% of the computational complexity. The algorithms execute autonomously, but require off-line training and setting of computational parameters. We employ pre-recorded neuronal signals to evaluate the accuracy of the proposed algorithms and architectures: the recorded data are processed by a standard PCA spike detection and alignment software algorithm, as well as by the several hardware algorithms, and the outcomes are compared.

Unsupervised Spike Sorting of extracellular electrophysiological recording in subthalamic nucleus of Parkinsonian patients

The present study demonstrates the application of the Unsupervised Spike Sorting algorithm (USS) to separation of multi-unit recordings and investigation of neuronal activity patterns in the subthalamic nucleus (STN). This nucleus is the main target for deep brain stimulation (DBS) in Parkinsonian patients. The USS comprises a fast unsupervised learning procedure and allows sorting of multiple single units, if any, out of a bioelectric signal. The algorithm was tested on a simulated signal with different levels of noise and with application of Time and Spatial Adaptation (TSA) algorithm for denoising. The results of the test showed a good quality of spike separation and allow its application to investigation of neuronal activity patterns in a medical application. One hundred twenty-four single channel multi-unit records from STN of 6 Parkinsonian patients were separated with USS into 492 single unit trains. Auto- and crosscorrellograms for each unit were analyzed in order to reveal oscillatory, bursting and synchronized activity patterns. We analyzed separately two brain hemispheres. For each hemisphere the percentage of units of each activity pattern were calculated. The results were compared for the first and the second operated hemispheres of each patient and in total.

Spike detection, characterization, and discrimination using feature analysis software written in LabVIEW

Rapid and accurate discrimination of single units from extracellular recordings is a fundamental process for the analysis and interpretation of electrophysiological recordings. We present an algorithm that performs detection, characterization, discrimination, and analysis of action potentials from extracellular recording sessions. The program was entirely written in LabVIEW (National Instruments), and requires no external hardware devices or a priori information about action potential shapes. Waveform events are detected by scanning the digital record for voltages that exceed a user-adjustable trigger. Detected events are characterized to determine nine different time and voltage levels for each event. Various algebraic combinations of these waveform features are used as axis choices for 2-D Cartesian plots of events. The user selects axis choices that generate distinct clusters. Multiple clusters may be defined as action potentials by manually generating boundaries of arbitrary shape. Events defined as action potentials are validated by visual inspection of overlain waveforms. Stimulus-response relationships may be identified by selecting any recorded channel for comparison to continuous and average cycle histograms of binned unit data. The algorithm includes novel aspects of feature analysis and acquisition, including higher acquisition rates for electrophysiological data compared to other channels. The program confirms that electrophysiological data may be discriminated with high-speed and efficiency using algebraic combinations of waveform features derived from high-speed digital records.

Robust, automatic spike sorting using mixtures of multivariate t-distributions

A number of recent methods developed for automatic classification of multiunit neural activity rely on a Gaussian model of the variability of individual waveforms and the statistical methods of Gaussian mixture decomposition. Recent evidence has shown that the Gaussian model does not accurately capture the multivariate statistics of the waveform samples' distribution. We present further data demonstrating non-Gaussian statistics, and show that the multivariate t-distribution, a wide-tailed family of distributions, provides a significantly better fit to the true statistics. We introduce an adaptation of a new expectation-maximization based competitive mixture decomposition algorithm and show that it efficiently and reliably performs mixture decomposition of t-distributions. Our algorithm determines the number of units in multiunit neural recordings, even in the presence of significant noise contamination resulting from random threshold crossings and overlapping spikes.

An unsupervised automatic method for sorting neuronal spike waveforms in awake and freely moving animals

The present study introduces an approach to automatic classification of extracellularly recorded action potentials of neurons. The classification of spike waveform is considered a pattern recognition problem of special segments of signal that correspond to the appearance of spikes. The spikes generated by one neuron should be recognized as members of the same class. The spike waveforms are described by the nonlinear oscillating model as an ordinary differential equation with perturbation, thus characterizing the signal distortions in both amplitude and phase. It is shown that the use of local variables reduces the problem of spike recognition to the separation of a mixture of normal distributions in the transformed feature space. We have developed an unsupervised iteration-learning algorithm that estimates the number of classes and their centers according to the distance between spike trajectories in phase space. This algorithm scans the learning set to evaluate spike trajectories with maximal probability density in their neighborhood. Following the learning, the procedure of minimal distance is used to perform spike recognition. Estimation of trajectories in phase space requires calculation of the first- and second-order derivatives, and integral operators with piecewise polynomial kernels were used. This provided the computational efficiency of the developed approach for real-time application as required by recordings in behaving animals and in human neurosurgical operations. The new method of spike sorting was tested on simulated and real data and performed better than other approaches currently used in neurophysiology.

Method for unsupervised classification of multiunit neural signal recording under low signal-to-noise ratio

Neural spike sorting is an indispensable step in the analysis of multiunit extracellular neural signal recording. The applicability of spike sorting systems has been limited, mainly to the recording of sufficiently high signal-to-noise ratios, or to the cases where supervised classification can be utilized. We present a novel unsupervised method that shows satisfactory performance even under high background noise. The system consists of an efficient spike detector, a feature extractor that utilizes projection pursuit based on negentropy maximization (Huber, 1985 and Hyvarinen et al, 1999), and an unsupervised classifier based on probability density modeling using mixture of Gaussians (Jain et al., 2000). Our classifier is based on the mixture model with a roughly approximated number of Gaussians and subsequent mode-seeking. It does not require accurate estimation of the number of units present in the recording and, thus, is better suited for use in fully automated systems. The feature extraction stage leads to better performance than those utilizing principal component analysis and two nonlinear mappings for the recordings from the somatosensory cortex of rat and the abdominal ganglion of Aplysia. The classification method yielded correct classification ratio as high as 95%, for data where it was only 66% when a kappa-means-type algorithm was used for the classification stage.

Detection of single unit activity from the rat vagus using cluster analysis of principal components

In vivo recordings from subdiaphragmatic vagal afferent nerves generally lack the resolution to distinguish single unit activity. Several methods for data acquisition and analysis were combined to produce a high degree of reliability in recording electrophysiological signals from gastrointestinal and hepatic afferent fibers in the rat. Recordings with low noise were achieved by paralysis of the respiratory muscles and by pinning the nerve to a recording platform. Single unit activity was isolated using principal component (PC) analysis and cluster cutting of data in multi-dimensional space (1-3 PCs). Cluster assignments were determined by a semi-automated approach using the k-means algorithm. The accuracy of single unit classification was assessed by checking inter-spike intervals (ISIs) to determine the length of the refractory period, and by cross-correlation analysis to assess whether single units were mistakenly split into more than one cluster. These analyses produced up to four isolated single units from each nerve filament (a bundle of nerve fibers), and typically it was possible to further increase yield by recording from several nerve filaments simultaneously using an array of electrodes.

Temporal transformation of multiunit activity improves identification of single motor units

This report describes a temporally based method for identifying repetitive firing of motor units. This approach is ideally suited to spike trains with negative serially correlated inter-spike intervals (ISIs). It can also be applied to spike trains in which ISIs exhibit little serial correlation if their coefficient of variation (COV) is sufficiently low. Using a novel application of the Hough transform, this method (i.e. the modified Hough transform (MHT)) maps motor unit action potential (MUAP) firing times into a feature space with ISI and offset (defined as the latency from an arbitrary starting time to the first MUAP in the train) as dimensions. Each MUAP firing time corresponds to a pattern in the feature space that represents all possible MUAP trains with a firing at that time. Trains with stable ISIs produce clusters in the feature space, whereas randomly firing trains do not. The MHT provides a direct estimate of mean firing rate and its variability for the entire data segment, even if several individual MUAPs are obscured by firings from other motor units. Addition of this method to a shape-based classification approach markedly improved rejection of false positives using simulated data and identified spike trains in whole muscle electromyographic recordings from rats. The relative independence of the MHT from the need to correctly classify individual firings permits a global description of stable repetitive firing behavior that is complementary to shape-based approaches to MUAP classification.

Spike sorting based on discrete wavelet transform coefficients

Using the novel mathematical technique known as wavelet analysis, a new method (WSC) is presented to sort spikes according to a decomposition of neural signals in the time-frequency space. The WSC method is implemented by a pyramidal algorithm that acts upon neural signals as a bank of quadrature mirror filters. This algorithm is clearly explained and an overview of the mathematical background of wavelet analysis is given. An artificial spike train, especially designed to test the specificity and sensibility of sorting procedures, was used to assess the performance of the WSC method as well as of methods based on principal component analysis (PCA) and reduced feature set (RFS). The WSC method outperformed the other two methods. Its superior performance was largely due to the fact that spike profiles that could not be separated by previous methods (because of the similarity of their temporal profile and the masking action of noise) were separable by the WSC method. The WSC method is particularly noise resistant, as it implicitly eliminates the irrelevant information contained in the noise frequency range. But the main advantage of the WSC method is its use of parameters that describe the joint time-frequency localization of spike features to build a fast and unspecialized pattern recognition procedure.

A technique to prevent dural adhesions to chronically implanted microelectrode arrays

Minimizing relative movements between neural tissues and arrays of microelectrodes chronically implanted into them is expected to greatly enhance the capacity of the microelectrodes to record from single cortical neurons on a long-term basis. We describe a new surgical technique to minimize the formation of adhesions between the dura and an implanted electrode array using a 12 microm (0.5 mil) thick sheet of Teflon film positioned between the array and the dura. A total of 15 cats were implanted using this technique. Gross examination of 12 implant sites at the time of sacrifice failed to find evidence of adhesions between the arrays and the dura when the Teflon(R) film remained in its initial position. In six implants from which recordings were made, an average of nine of the 11 (81%) connected electrodes in each array recorded evoked neural activity after 180 days post implantation. Further, on average, two separable units were identified on each of the implanted electrodes in these arrays. No significant change was found in the density of cell bodies around implanted electrodes of four of the implanted electrode arrays. However, histological evaluation of the implant sites revealed evidence of meningeal proliferation beneath the arrays. The technique described is shown to be effective at preventing adhesions between implanted electrode arrays and improve the characteristics of chronic recordings obtained with these structures.

Identification of reliable spike templates in multi-unit extracellular recordings using fuzzy clustering

A method for extracting single-unit spike trains from extracellular recordings containing the activity of several simultaneously active cells is presented. The technique is particularly effective when spikes overlap temporally. It is capable of identifying the exact number of neurons contributing to a recording and of creating reliable spike templates. The procedure is based on fuzzy clustering and its performance is controlled by minimizing a cluster-validity index which optimizes the compactness and separation of the identified clusters. Application examples with synthetic spike trains generated from real spikes and segments of background noise show the advantage of the fuzzy method over conventional template-creation approaches in a wide range of signal-to-noise ratios.

A 100-channel system for real time detection and storage of extracellular spike waveforms

As extracellular electrode arrays with 100 or more active recording sites become more widely used for simultaneous recording of neural ensembles, practical data acquisition systems that can efficiently accommodate high electrode counts are needed. To reduce the high data rates associated with extracellular recordings from these arrays, various algorithms and systems have been designed to provide complete online detection and classification of extracellular spike waveforms. However, many of these algorithms require significant user supervision to ensure accurate performance. In this paper, we discuss the design and validation of a 100-channel PC-based system that can be used with arrays of extracellular electrodes such as the Utah Electrode Array. Instead of comprehensive online spike analysis, the system performs online detection and storage of the spike waveforms for offline classification. This strategy preserves the data of interest, reduces system complexity, and requires less user supervision during experiments.

Processing of sound sequences in macaque auditory cortex: response enhancement

It is well established that the tone-evoked response of neurons in auditory cortex can be attenuated if another tone is presented several hundred milliseconds before. The present study explores in detail a complementary phenomenon in which the tone-evoked response is enhanced by a preceding tone. Action potentials from multiunit groups and single units were recorded from primary and caudomedial auditory cortical fields in lightly anesthetized macaque monkeys. Stimuli were two suprathreshold tones of 100-ms duration, presented in succession. The frequency of the first tone and the stimulus onset asynchrony (SOA) between the two tones were varied systematically, whereas the second tone was fixed. Compared with presenting the second tone in isolation, the response to the second tone was enhanced significantly when it was preceded by the first tone. This was observed in 87 of 130 multiunit groups and in 29 of 69 single units with no obvious difference between different auditory fields. Response enhancement occurred for a wide range of SOA (110-329 ms) and for a wide range of frequencies of the first tone. Most of the first tones that enhanced the response to the second tone evoked responses themselves. The stimulus, which on average produced maximal enhancement, was a pair with a SOA of 120 ms and with a frequency separation of about one octave. The frequency/SOA combinations that induced response enhancement were mostly different from the ones that induced response attenuation. Results suggest that response enhancement, in addition to response attenuation, provides a basic neural mechanism involved in the cortical processing of the temporal structure of sounds.

Branching and/or collateral projections of spinal dorsal horn neurons

Branching and/or collateral projections of spinal dorsal horn neurons is a common phenomenon. Evidence is presented for the existence of STTm/STTl, STTc/STTi, STT/SMT, STT/SRT, SCT/DCPS, SST/DCPS, SCT/SST, STT/SHT, STeT/SHT, STeTs and other doubly or multiply projecting spinal neurons that have been anatomically and physiologically identified and named based on the locations of the cells of origin and their terminations in the brain. These newly discovered spinal projection neurons are characterized by a single cell body and branched axons and/or collaterals that project to two or more target areas in the brain. These novel populations of neurons seem to be a fuzzy set of spinal projection neurons that function as an intersection set of the corresponding single projection spinal neurons and to be at an intermediate stage phylogenetically. Identification strategies are discussed, and general concluding remarks are made in this review.

Digital video: a tool for correlating neuronal firing patterns with hand motor behavior

This report describes the use of multimedia technology for simultaneous recording of single unit responses in cerebral cortex, and imaging of hand kinematics as monkeys grasp and manipulate objects. These imaging methods allow direct correlation of full-frame, full-field video images with the actual spike trains recorded with microelectrodes. Our implementation of digital video provides high-resolution snapshots of the hand motor behavior every 33.3 ms, and a precise calibration and display of the synchronously recorded electrophysiological activity digitized at rates up to 44.5 kHz on the same platform. These imaging methods permit non-invasive, non-traumatic monitoring of both trained and spontaneous activity in experimental animals, while providing synchronized digitized records of neuronal spike trains. We also describe software instruments that quantify and analyze the digitized spike trains. One instrument employs user-selectable objective criteria for distinguishing spikes from noise, separates individual action potential waveforms by their amplitude and duration, and compiles time stamps for each spike train. A second instrument constructs rasters and histograms of repeated behavioral trials using the timing of the corresponding video frame for alignment. These analyses reveal functional classes of cortical neurons signaling specific stages of prehension.

Unsupervised classification of neural spikes with a hybrid multilayer artificial neural network

The understanding of the brain structure and function and its computational style is one of the biggest challenges both in Neuroscience and Neural Computation. In order to reach this and to test the predictions of neural network modeling, it is necessary to observe the activity of neural populations. In this paper we propose a hybrid modular computational system for the spike classification of multiunits recordings. It works with no knowledge about the waveform, and it consists of two moduli: a Preprocessing (Segmentation) module, which performs the detection and centering of spike vectors using programmed computation; and a Processing (Classification) module, which implements the general approach of neural classification: feature extraction, clustering and discrimination, by means of a hybrid unsupervised multilayer artificial neural network (HUMANN). The operations of this artificial neural network on the spike vectors are: (i) compression with a Sanger Layer from 70 points vector to five principal component vector; (ii) their waveform is analyzed by a Kohonen layer; (iii) the electrical noise and overlapping spikes are rejected by a previously unreported artificial neural network named Tolerance layer; and (iv) finally the spikes are labeled into spike classes by a Labeling layer. Each layer of the system has a specific unsupervised learning rule that progressively modifies itself until the performance of the layer has been automatically optimized. The procedure showed a high sensitivity and specificity also when working with signals containing four spike types.

Protocol for microneurography with concentric needle electrodes

In 1968, the method of human percutaneous microneurography with solid tungsten electrodes was introduced. Since then many investigators used this technique to study peripheral mechanisms in the somatosensory, motor and autonomic systems of conscious humans. Although some modifications of the method were described, the basic construction of the recording electrode has remained the same over the years. In the present protocol we describe in detail the procedures of microneurography using a thin diameter concentric needle electrode. There are some advantages with the concentric electrodes in comparison with the tungsten needles: (1) the electrical and mechanical properties of the electrode are stable which allows repeated use, (2) its restricted and one-dimensionally directed recording area provides the possibility to study topographical aspects within even a part of a peripheral nerve fascicle, and (3) multi-channel recordings can be achieved by adding more recording surfaces to the electrode. Based on recent investigations evaluating the recording properties of concentric electrodes we propose a novel procedure for signal analysis where template matching is incorporated. The analyses described in this protocol might also be applicable for extracellular recordings from muscle or elsewhere within the nervous system.

Detection, classification, and superposition resolution of action potentials in multiunit single-channel recordings by an on-line real-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.

The Utah intracortical Electrode Array: a recording structure for potential brain-computer interfaces

We investigated the potential of the Utah Intracortical Electrode Array (UIEA) to provide signals for a brain-computer interface (BCI). The UIEA records from small populations of neurons which have an average signal-to-noise ratio (SNR) of 6:1. We provide specific examples that show the activities of these populations of neurons contain sufficient information to perform control tasks. Results from a simple stimulus detection task using these signals as inputs confirm that the number of neurons present in a recording is significant in determining task performance. Increasing the number of units in a recording decreases the sensitivity of the response to the stimulus; decreasing the number of units in the recording, however, increases the variability of the response to the stimulus. We conclude that recordings from small populations of neurons, not single units, provide a reliable source of sufficiently stimulus selective signals which should be suitable for a BCI. In addition, the potential for simultaneous and proportional control of a large number of external devices may be realized through the ability of an array of microelectrodes such as the UIEA to record both spatial and temporal patterns of neuronal activation.

Multi-unit spike discrimination using wavelet transforms

A new spike discrimination procedure addressing the specific problem of spike superposition is described. The method, based on a shift-invariant wavelet transform and its amplitude-and-phase representation, has the advantage of both reducing the effect of noise present in the data and correcting the latency of specific components in a waveform. When spikes overlap and produce unknown patterns, the procedure extracts the constituent spikes and also estimates their exact time of occurrence. Fast implementation algorithms, having complexity of at most O (N log N), allow the use of the method in real-time applications.

Analysis of neuro-behavioral Discovery data on the Macintosh computer

This paper describes a suite of routines using IgorPro, a powerful analysis and graphing software package for the Macintosh computer, to enhance the ability to analyze, manipulate, and display data recorded with the Discovery acquisition software marketed by DataWave Technologies. The routines are able to time-align fast and slow data channels, and are especially useful for analyses that involve both neural and behavioral data. The software was designed for eyeblink conditioning and vocalization experiments, but it can easily be used for analyzing other types of neurobehavioral data. The data are first prepared on the PC with routines that inspect the header of the data file and translate the data file into a compact binary format that can be read by IgorPro. An option is also available to splice out data from unnecessary portions of an intertrial interval. The new file is then put on the Macintosh computer for display and analysis by IgorPro. These routines enable both neural and behavioral data to be quickly and easily reduced, manipulated, and statistically and graphically summarized.

Application of non-linear filters based on the median filter to experimental and simulated multiunit neural recordings

Two non-linear, high-pass filters based on the median filter are proposed and tested as substitutes for linear filtering in applications involving multiunit neural recordings. The first, the median-based high-pass (MH) filter, operates by subtracting the output from the input of the median filter; it is aimed at preserving the shape of the impulses. The second, the negative median-based high-pass (NMH) filter, sets at zero the positive values in the output of the MH filter; it is aimed at transforming the impulses into monophasic waves placed on a flat baseline. When applied to experimental recordings and to a template action potential, the two median-based filters clearly outperformed two corresponding procedures based on a linear filter (moving-average filter). They did not produce appreciable distortions of the impulses, whereas their two counterparts induced or enlarged lateral lobes, as is the rule for linear high-pass filters. The recording display was much improved and impulse identification was made easier. When the two filters were applied to simulated recordings and the mean output was estimated by averaging and cross-correlation, a certain degree of performance deterioration was assessed in conditions of sustained activity and/or noise, with a resulting growing similarity to the mean output of the two corresponding, moving-average-based filters.

Multiple action potential waveforms of single units in man as signs of variability in conductivity of their myelinated fibres

Percutaneous microneurography was performed with concentric needle electrodes to record neural activity from myelinated fibres in human peripheral nerves. Template matching techniques were used together with interspike interval analysis and studies on functional class, receptive field characteristics, conduction velocities and other single fibre properties to classify single units. Sometimes the same fibres exhibited different action potentials at the same time. The potentials had some common features, but differed either in their waveform types or only in duration. There was a correlation between the occurrence of the different potential shapes and firing frequency of the studied unit. The outcome of the studies suggested that there was a common denominator which could explain the observations. Most likely, momentary fluctuations in excitability of the myelinated fibres occurring during the relative refractory period or the supernormal period were responsible for the variations in complexity of the studied units due to a partial block of fibre propagation probably caused by the recording electrode. Thus, action potentials deriving from the same axon may not always have the same shapes. Methods for unit classification, such as template matching, are discussed in the light of our findings.

Automatic sorting of multiple unit neuronal signals in the presence of anisotropic and non-Gaussian variability

Neuronal noise sources and systematic variability in the shape of a spike limit the ability to sort multiple unit waveforms recorded from nervous tissue into their single neuron constituents. Here we present a procedure to efficiently sort spikes in the presence of noise that is anisotropic, i.e., dominated by particular frequencies, and whose amplitude distribution may be non-Gaussian, such as occurs when spike waveforms are a function of interspike interval. Our algorithm uses a hierarchical clustering scheme. First, multiple unit records are sorted into an overly large number of clusters by recursive bisection. Second, these clusters are progressively aggregated into a minimal set of putative single units based on both similarities of spike shape as well as the statistics of spike arrival times, such as imposed by the refractory period. We apply the algorithm to waveforms recorded with chronically implanted micro-wire stereotrodes from neocortex of behaving rat. Natural extension of the algorithm may be used to cluster spike waveforms from records with many input channels, such as those obtained with tetrodes and multiple site optical techniques.

A neural network approach to real-time spike discrimination during simultaneous recording from several multi-unit nerve filaments

A multi-channel, real-time, unsupervised spike discriminator was developed in order to reconstruct single spike trains from several simultaneously recorded multi-unit nerve filaments. The program uses a Self Organising Map (SOM) algorithm for the classification of the spikes. In contrast to previous similar techniques, the described method is made for use on a PC, and the method may thus be implemented at relatively low cost. In order to test the accuracy of the program, a robustness test was performed, where noise with different RMS levels was superimposed on the spikes. Furthermore, the maximal classification rate was determined. The program is easy to use, since the only manual inputs needed are the voltage threshold for spike detection, and the number of units present in each recorded nerve filament.

Real-time separation of multineuron recordings with a DSP32C signal processor

We have developed a hardware and software package for real-time discrimination of multiple-unit activities recorded simultaneously from multiple microelectrodes using a VME-Bus system. Compared with other systems cited in literature or commercially available, our system has the following advantages. (1) Each electrode is served by its own preprocessor (DSP32C); (2) On-line spike discrimination is performed independently for each electrode. (3) The VME-bus allows processing of data received from 16 electrodes. The digitized (62.5 kHz) spike form is itself used as the model spike; the algorithm allows for comparing and sorting complete wave forms in real time into 8 different models per electrode.

Method for determining individual neuron size in simultaneous single-unit recordings

A technique for estimating the size of neurons is based on extracellular recordings with paired-electrode sets. Simultaneous single-unit recordings are obtained from the dragonfly mesothoracic ganglion. It is assumed that the ganglion is a passive electrical environment, where spike amplitudes decrease with the inverse of distance squared, and spike angles (widths) increase linearly with distance from the cellular source to the recording electrodes. Starting with the recorded spike amplitudes and angles for each cell, a numerical algorithm is iterated to estimate the true value of the amplitude and angle minus these passive electrical distance effects. The resolved amplitude is a direct, consistent estimate of the size of each recorded neuron. The results indicate that a dichotomy of small and large cells is recorded in roughly a 2:1 ratio. The dichotomy of cell sizes is consistent with the available histological data, although a larger ratio of small to large cells (approximately 10:1) would be expected. Thus, a sampling bias for large cells is apparent, which may be reflective of the larger soma/proximal geometries of such cells. As the technique determines the size of each individual neuron, such biases are eliminated from population studies of the neural tissue. Furthermore, knowledge about the size of each individual neuron permits more detailed analyses of the interactions and contributions of single cells within a network of cells based upon size.

A neural network-based spike discriminator

A software routine to reconstruct individual spike trains from multi-neuron, single-channel extracellular recordings was designed. Using a neural network algorithm that automatically clusters and sorts the spikes, the only user input needed is the threshold level for spike detection and the number of unit types present in the recording. Adaptive features are included in the algorithm to allow for tracking of spike trains during periods of amplitude variation and also to identify noise spikes. The routine will operate on-line during extracellular studies of the cochlear nucleus in cats.

Action potential classification with dual channel intrafascicular electrodes

Using recordings of peripheral nerve activity made with carbon fiber intrafascicular electrodes, we compared the performance of three different recording techniques (single channel, differential, and dual channel) and four different unit classification methods (linear discriminant analysis, template matching, a novel time amplitude windowing technique, and neural networks) in terms of errors in waveform classification and artifact rejection. Dual channel recording provided uniformly superior unit separability, neural networks gave the lowest classification error rates, and template matching had the best artifact rejection performance.