Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
Review
. 2012 Feb;22(1):11-7.
doi: 10.1016/j.conb.2011.10.001. Epub 2011 Oct 22.

Towards reliable spike-train recordings from thousands of neurons with multielectrodes

Gaute T Einevoll  1 , Felix Franke, Espen Hagen, Christophe Pouzat, Kenneth D Harris
Affiliations
Review

Towards reliable spike-train recordings from thousands of neurons with multielectrodes

Gaute T Einevoll et al. Curr Opin Neurobiol. 2012 Feb.

Abstract

The new generation of silicon-based multielectrodes comprising hundreds or more electrode contacts offers unprecedented possibilities for simultaneous recordings of spike trains from thousands of neurons. Such data will not only be invaluable for finding out how neural networks in the brain work, but will likely be important also for neural prosthesis applications. This opportunity can only be realized if efficient, accurate and validated methods for automatic spike sorting are provided. In this review we describe some of the challenges that must be met to achieve this goal, and in particular argue for the critical need of realistic model data to be used as ground truth in the validation of spike-sorting algorithms.

PubMed Disclaimer

Figures

Figure 1
Figure 1
Overview of spike-sorting process. Electrodes record changes in the extracellular electrical potential (b) caused by action potentials of neurons in its vicinity (a) [intracellular potentials of three spiking example neurons (N1, N2, N3) are depicted in lower left panel]. For the detection and analysis a highpass or bandpass filter is usually used to remove the low-frequency part of the potential (c). The optimal procedure for detection of spikes, especially in multielectrode recordings, is still is an unsolved problem but simple voltage thresholding is commonly used (d). For most spike-sorting procedures the extracted spike waveforms need to be temporally aligned on a common feature like the position of the voltage peak (e) before features are extracted from every waveform (f). The feature extraction is crucial for decreasing the dimensionality of the data to the most informative dimensions. This can be done by, for example, using principal component analysis [4] or wavelets [49]. Clustering, that is, finding the number of clusters and their position in the feature space (g), is highly susceptible to the choice of those features. Individual clusters should contain all spikes of one putative neuron only and are commonly assumed to exhibit multivariate normal [48] or t-distributions [50]. The average waveform of all spikes belonging to one cluster is called the ‘template’ of that neuron (h). The outcome of the clustering is often a statistical model of the data (e.g., number of neurons, templates, covariance matrices) that can be used for quality estimation of the sorting result (see Figure 2 and [30••,48]) and derivation of a classifier for yet unclassified spikes (e.g., template matching).
Figure 2
Figure 2
Examples of spike-sorting validation techniques applicable without knowledge of the true spike trains (ground truth) [30••]. After spike sorting of the raw recorded signals (a) the resulting spike trains (b) can be checked for biological plausibility by, for example, computing the interspike interval (ISI) (c) and inspected for violation of a refractory period. Using a dynamic multidimensional data visualization software like GGobi (www.ggobi.org), cluster separation can be judged visually (d). The assumptions of model-based sorters, for example, assuming a mixture of Gaussian or t-distribution for the clusters, can be checked by finding for each pair of clusters a direction along which the separation of the spikes of these two clusters is optimal, that is, determining the optimal linear discriminant. The overlap of the densities of spikes from each pair of clusters projected onto this optimal direction can then be theoretically obtained, compared to the empirical distribution to validate the model assumptions, and used as a measure of the probability of misclassification [48] (e) and (f). The linear discriminant approach applies only if the different clusters have a common covariance matrix. Otherwise, the quadratic discriminant can be used instead.
Figure 3
Figure 3
Validation of spike-sorting methods by use of ground-truth data obtained from biophysically detailed modeling of extracellular potentials accompanying action-potential firing. (a1) A population of (in this case three) spiking model neurons is placed around a virtual multielectrode (in this case a tetrode). Ground-truth somatic membrane potentials (a2) and the neuronal transmembrane currents needed to evaluate the corresponding extracellular traces using an electrostatic forward-modeling formula [40,17], are calculated with standard multicompartmental modeling of spiking neurons [–43,17]. Alternatively, recorded intracellular action potentials can be imposed directly as an electric boundary condition in the soma of the multicompartmental model neuron [17]. Noise can be added to the calculated extracellular traces as desired (b). The model ‘raw data’ are then used to reconstruct the spike trains (c) following the standard spike-sorting procedure depicted in Figure 1. These estimated spike trains can then be compared with the known ground-truth spike trains read directly out from the somatic membrane potentials (a2).
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Adrian ED, Moruzzi DD. Impulses in the pyramidal tract. J Physiol. 1939;97:153–199. - PMC - PubMed
    1. Haykin S, Chen Z. The cocktail party problem. Neural Comp. 2005;17:1875–1902. - PubMed
    1. Abeles M, Goldstein M. Multispike train analysis. Proc IEEE. 1977;65:762–773.
    1. Lewicki MS. A review of methods for spike sorting: the detection and classification of neural action potentials. Network: Comp Neur Syst. 1998;9:R53–R78. - PubMed
    1. Quiroga RQ. Spike sorting. Scholarpedia. 2007;2:3583.

Publication types