Keywords

1 Introduction

Recent technological advancements allowed scientists to have unprecedented access to biological data. These data come from different organs in the body and have been effectively utilized in screening and diagnosis of diseases and their treatment (Mahmud et al. 2018). Brain is the most complex organ in the mammals. Though it is analog, its capability in decision-making and pattern recognition is higher than any existing computing machines (Mahmud et al. 2017). To understand brain’s functionality, diagnose disease, and devise treatments, scientists have been investigating it using different approaches (Mahmud and Vassanelli 2016). In last two decades, micro- and nanotechnology underwent an exponential growth in terms of developing novel miniaturized devices and this allowed neuroscientists to target large populations of neurons and record from them to decode the activities of the brain cells (Mahmud et al. 2017; Vassanelli 2011; Vassanelli et al. 2012b,a; Vassanelli 2014; Schröder et al. 2015; Thewes et al. 2016; Jun et al. 2017). However, these novel techniques to acquire neuronal signals generate huge amount of data. And, analyzing this data and mining relevant information is a big challenge. To this goal, individual research groups have contributed towards the development of automated, efficient, and intelligent processing methods and disseminated them to the neuroscientific community (Mahmud and Vassanelli 2016). The interdisciplinary “Neuroengineering” community (Vassanelli and Mahmud 2016) used these tools to mine useful information from these large datasets (Mahmud and Vassanelli 2016). Targeting different applications and needs, these methods deal with processing and analysis of data coming from single or multiple channels. However, with today’s increasing number of recording sites accommodated in a single probe, many of these methods are difficult to rescale and fit to analyze these data. Therefore, the community is still in need of novel analysis tools targeting multichannel neurophysiological data coming from high-resolution neuronal probes. This chapter aims in introducing the reader with available open-source toolboxes capable of performing processing and analysis of multichannel extracellular neuronal signals recorded in vitro.

2 State of the Art of Extracellular Neuronal Signal Analysis

Modern neuroscience research has emerged as a data-driven discipline where both experimental and computational approaches go hand-in-hand (see Fig. 1) (Mahmud et al. 2018). The two approaches of neuroscience research are bridged through a relatively new discipline, called data science, which mainly deals with the analytics of the acquired data, and interpretation of the simulated results and design novel experiments suggested by the obtained results.

Fig. 1
figure 1

Overview of modern neuroscience research which comprises of computational and experimental approaches to neuroscience through the data science domain

Full size image

Focusing on the extracellular neuronal signals acquired in vitro, this chapter is going to deal with the data analytics part, and provide an overview of the available open-source data analytics resources which can be exploited to process and analyze the recorded signals. Using the in vitro experimental paradigm (see Fig. 2), mainly two types of signals can be recorded: (1) neuronal spikes and their trains, and (2) field potentials. Therefore, the following subsections will contain the state-of-the-art open-source resources categorized by signal types, e.g., field potentials, and spikes. Despite that the majority of the available resources are application and signal specific, there exist a few toolboxes which provide methods applicable on multiple types of signals. For the sake of simplicity, the tools are divided into following two categories:

  • Toolboxes for spike trains and field potentials analysis and

  • Toolboxes for spike sorting

Fig. 2
figure 2

Exemplary overview of the in vitro neuronal signal acquisition and analysis paradigm

Full size image

It is also worthy to note that majority of the available toolboxes are developed using MATLAB (Mathworks Inc., Natick, USA; www.mathworks.com) and python (www.python.org) programming languages due to their diffused usage in the neuroscience community. In addition, popular programming languages including C, C++, C#, Delphi7, Java, and R have also been employed in some of the toolboxes.

2.1 Toolboxes for Spike Trains and Field Potentials Analysis

With the growing amount data acquired via simultaneously recorded channels from an increasing number of neurons, the neuroengineering community has developed automated toolboxes addressing the required processing and analyses. The following subsections describe—in alphabetical order—popular publicly available toolboxes. Table 1 summarizes the different packages with their representative features.

Table 1 Popular spike train and field potential processing and analysis toolboxes with their representative features
Full size table

2.1.1 Brain System for Multivariate AutoRegressive Time Series (BSMART)

“BSMART” toolbox is written in Matlab/C for spectral analysis of neurophysiological signals (Cui et al. 2008). It provides multi- or bivariate AutoRegressive modeling, spectral analysis through coherence and Granger causality, and network analysis. The main analyses include: adaptive (bi-/multi-)variate autoregressive model, fast Fourier transform, Granger causality (Granger 1969), coherence, coherent network analysis, and Granger causality network analysis. Available at http://www.brain-smart.org/.

2.1.2 Chronux

“Chronux” toolbox is developed in Matlab for the analysis of both point process and continuous data (Bokil et al. 2010). It provides spike sorting, and local regression and multitaper spectral analysis of neural signals. The main analyses include: hierarchical clustering method (Fee et al. 1996), locally weighted sum of squares (Cleveland 1979), local regression fitting and density estimation (Loader 1999), multitaping method (Thomson 1982), coherence, and spike field coherence. Available at http://chronux.org/.

2.1.3 DATA-MEAns

“DATA-MEAns” is a toolbox developed in Borland Delphi 7 (Embarcadero Technologies Inc., Austin, USA) and Matlab (Bonomini et al. 2005). It provides data visualization, basic analysis (i.e., autocorrelations, perievent histograms, rate curves, PSTHs, ISIs, etc.), and nearest neighbor or k-means clustering. The analyses include: poststimulus time and perievent histogram estimation, auto- and cross-correlation, Fano factor and coherence calculation, event synchrony (Quian Quiroga et al. 2002), and nearest neighbor (Cover and Hart 1967) and K-means (MacQueen 1967) clustering. Available at http://cortivis.umh.es/.

2.1.4 Finding Information in Neural Data (FIND)

“FIND” is a platform-independent framework for the analysis of neuronal data based on Matlab (Meier et al. 2008). It provides a unified data import function from various proprietary formats simplifying standardized interfacing with analysis tools and provides means for analysis of discrete series of spike events, continuous time series, and imaging data. Also, it allows simulating multielectrode activity using point-process-based stochastic model. The analyses include: co-variance estimation, point process modeling, pair-wise cross-correlation, asymmetric Savitzky–Golay filter calculation (Savitzky and Golay 1964), response latency differences estimation (Nawrot et al. 2003), and spike detection. Available at http://find.bccn.uni-freiburg.de/.

2.1.5 Information Breakdown ToolBox (ibTB)

“ibTB” is a Matlab-based toolbox which implements information theory methods for spike, LFP, and EEG analysis (Magri et al. 2009). It provides information breakdown technique to decode the encoding of sensory stimuli by different groups of neurons. The major supported analyses include: direct method, quadratic extrapolation (Strong et al. 1998), Panzeri and Treves method (Panzeri and Treves 1996) for bias correction, shuffling procedure (Montemurro et al. 2007), bootstrap bias correction (Optican et al. 1991), and Gaussian method (Misra et al. 2005). The source code can be obtained from the publisher’s website (http://static-content.springer.com/esm/art%3A10.1186%2F1471-2202-10-81/MediaObjects/1471-2202-10-81-S1.zip).

2.1.6 Klusters, NeuroScope, and NDManager

“Klusters,” “NeuroScope,” and “NDManager” are three integrated modules bundled together for processing and analysis of spike and field potential signals (Hazan et al. 2006). Klusters performs spike sorting using KlustaKwik (see Sect. 2.2.3) and displays 2D projection of features, spike traces, correlograms, and error matrix view. NeuroScope allows inspection, selection, and event editing of spike signals as well as local field potentials (LFPs). NDManager facilitates experimental and preprocessing parameter management. The major analyses include: auto- and cross-correlation estimation, spike detection and sorting, and classification expectation–maximization (Celeux and Govaert 1992). Available at http://neurosuite.sourceforge.net/.

2.1.7 MeaBench

“MeaBench” is a toolbox written mainly in C++ with certain parts written in PerlFootnote 1 and Matlab. It is intended for data acquisition and online analysis of commercial multielectrode array recordings from Multichannel Systems GmbH (Reutlingen, Germany) (Wagenaar et al. 2005). It allows real-time data visualization, line and stimulus artifact suppression, and spike and burst detection and validation. Available at www.danielwagenaar.net/res/software/meabench/.

2.1.8 Multivariate Granger Causality Toolbox (MVGC)

“MVGC” is a toolbox written in Matlab that implements Wiener–Granger causality (G-causality) on multiple equivalent representations of a vector autoregressive model in both time and frequency domains (Barnett and Seth 2014). The main analyses supported by the toolbox include: ordinary least squares estimation, Wiener (root mean square) method (Levinson 1946), vector autoregressive maximum likelihood estimator method, cross-power spectral density, multitaper method, fast Fourier transform, and unconditional Granger causality. It can be applied to neuroelectric, neuromagnetic, and fMRI signals and can be obtained from http://www.sussex.ac.uk/sackler/mvgc/.

2.1.9 nSTAT

“nSTAT” toolbox is coded in Matlab and performs spike train analysis in time domain (e.g., Kalman Filtering), frequency domain (e.g., multitaper spectral estimation), and mixed time–frequency domain (e.g., spectrogram) (Cajigas et al. 2012). The supported analyses include: point process generalized linear model (Paninski et al. 2007), generalized linear model-based peristimulus time histogram estimation, Akaike’s and Bayesian information criteria, state-space generalized linear model, Kalman filtering, multitaper method, and spectrogram. Available at www.neurostat.mit.edu/nstat/.

2.1.10 PANDORA

“PANDORA” is a Matlab-based toolbox that extracts user-defined characteristics from spike train signals and creates numerical database tables from them (Gunay et al. 2009). Further analyses (e.g., drug and parameter effects, spike shape characterization, histogramming and comparison of distributions, cross-correlation, etc.) can then be performed on these tables. However, spike detection and feature extraction can also be performed. The supported analyses include: rational database creation from datasets, extraction of spike shape characteristics, Kullback–Leibler divergence measure (Kullback and Leibler 1951) estimation, and resistor-average distance (Johnson et al. 2001) estimation. It is available at https://github.com/cengique/pandora-matlab.

2.1.11 QSpike Tools

“QSpike Tools” is a Linux/Unix-based cloud-computing framework, modeled using client–server architecture and developed in Matlab/Bash scripts,Footnote 2 for processing and analysis of extracellular spike trains (Mahmud et al. 2014). It performs batch preprocessing of CPU-intensive operations for each channel (e.g., filtering, multiunit activity detection, spike sorting, etc.), in parallel, by delegating them to a multicore computer or to a computers cluster. The main analyses include: spike detection and validation, poststimulus time and perievent histogram estimation, burst detection and validation, and spike sorting through Wave_Clus package (see Sect. 2.2.11). It can be obtained from https://sites.google.com/site/qspiketool/.

2.1.12 SigMate

“SigMate” is a Matlab-based comprehensive framework that allows preprocessing and analysis of EEG, LFPs, and spike signals (Mahmud et al. 2012a). Its main contribution is in the analysis of LFPs which includes data display, file operations, baseline correction, artifact removal, noise characterization, current source density (CSD) analysis, latency estimation from LFPs and CSDs, determination of cortical layer activation order using LFPs and CSDs, and single LFP clustering. The main processing and analyses include: various file operations (e.g., file splitting, concatenation, and column rearranging), latency calculation (Mahmud et al. 2016), detection of cortical layer activation order (Mahmud et al. 2010), current source density analysis (Mahmud et al. 2011), classification of single-trial LFPs (Mahmud et al. 2012c), and spike analysis. The spike analyses are provided through Wave_Clus (see Sect. 2.2.11) toolbox. It can be obtained from https://sites.google.com/site/muftimahmud/codes.

2.1.13 sigTOOL

“sigTOOL” toolbox is written in Matlab and allows direct loading of a wide range of proprietary file formats (Lidierth 2009). The usable data file formats originate from various hardware vendors such as Alpha Omega, Axon Instruments, Blackrock Microsystems, Cambridge Electronic Design, Heka, MultiChannel Systems, NeuroExplorer, NeuroShare native, and Plexon. The major analyses supported by the toolbox include: auto- or cross-correlation, power spectral analysis, and coherence estimation in addition to usual spike train analysis (i.e., ISI, event auto- and cross-correlations, spike-triggered averaging, perievent time histograms, frequencygrams, etc.). Available at http://sigtool.sourceforge.net/.

2.1.14 SpiCoDyn

“SpiCoDyn” is an open-source windows-only graphical user interface-based toolbox focusing on functional-effective connectivity analysis and spiking and bursting dynamics analysis (Pastore et al. 2018). Developed in C#, under the Microsoft .NET platform, it supports HDF5,Footnote 3 level 5 MAT filesFootnote 4 and text files. The toolbox provides optimized implementations of two main transfer entropy algorithms (i.e., delayed transfer entropy and high-order transfer entropy) and provides analysis platform for multiple spike trains originating from large number of electrodes. Available at https://www.nitrc.org/projects/spicodyn/.

2.1.15 Spike Train Analysis Toolkit (STAToolkit)

“STAToolkit” is a Matlab/C-hybrid toolbox implementing information theoretic methods to quantify how well the stimuli can be distinguished based on the timing of neuronal firing patterns in a spike train (Goldberg et al. 2009). The main analyses include: direct method (Strong et al. 1998), metric space method (Victor and Purpura 1997), binless method (Victor 2002), asymptotically debiased method (Treves and Panzeri 1995), Jackknife debiased method (Thomson and Chave 1991), debiased Ma bound method (Ma 1981), best upper bound method (Paninski 2003), coverage-adjusted method (Chao and Shen 2003), and Bayesian with Dirichlet prior (Wolpert and Wolf 1995). Available at http://neuroanalysis.org.

2.1.16 SPKTool

“SPKTool” is coded in Matlab for the detection and analysis of neural spiking activity (Liu et al. 2011). It performs spike detection, feature extraction, and manual and semiautomatic clustering of spike trains. Spike detection by thresholding and raw and nonlinear energy of signal, extraction of various spike features (e.g., principal components, peaks, valleys, energy, timestamps, slice of waveforms, etc.), implementation of popular spike-sorting techniques (e.g., K-means, template matching, EM of Gaussian mixed model, valley seeking, manual contour based method, etc.), ISI, Poincare maps, correlation, firing rate histograms, and perievent rasters and histograms. Available at http://spktool.sourceforge.net/.

2.1.17 ToolConnect

“ToolConnect” is a standalone windows-only tool, developed in C#, targeting to infer functional connectivity from spike trains data (Pastore et al. 2016). This toolbox has been optimized for in vitro networks grown on high-density multielectrode arrays. The implemented analyses include: cross- and partial correlation, and joint and transfer entropy. Additionally, it contains several add-ons for the visualization of the functional connectivity graphs as well as extraction of topological features of the network. Available at https://www.nitrc.org/projects/toolconnect/.

2.2 Toolboxes for Spike Sorting

A great amount of efforts have been put towards the development of sophisticated tools capable of accurate spike sorting and analysis. Rey et al. (2015), in their review, outline the primary concepts of spike sorting, the requirements for applying specific techniques, and most importantly, the shortcomings of currently available algorithms. Due to the vast amount of methods and tools available for the purpose, we restrict our discussion only to the popular open-source toolboxes.

2.2.1 EToS

“EToS” or Efficient Technology of Spike sorting is a spike-sorting toolbox aimed mainly at heterogeneous neural population recordings (Takekawa et al. 2012). It is written in C++ implementing multimodality-weighted PCA for feature extraction and variational Bayes for student’s t mixture model for clustering. The spike-sorting code is parallelized through OpenMP (www.openmp.org) and available at http://etos.sourceforge.net/.

2.2.2 KiloSort

“KiloSort” is a spike-sorting package developed in Matlab for handling data coming from MEAs with large number of recording electrodes (Pachitariu et al. 2016). The package uses template matching for detection and clustering of spikes and can handle batch processing using both GPUs and CPUs. The source code can be downloaded from https://github.com/cortex-lab/KiloSort.

2.2.3 KlustaKwik

“KlustaKwik” is a standalone program written in C++ for automatic clustering analysis (Harris et al. 2000) of high-dimensional spiking data by fitting a mixture of Gaussians and hard expectation–maximization algorithm with unconstrained covariance matrices (Kadir et al. 2014; Rossant et al. 2016). The package can be downloaded from https://github.com/klusta-team/klustakwik.

2.2.4 MClust

“MClust” is a spike-sorting toolbox developed in Matlab. It supports both manual and automated clustering with possibility to manual feature selection (Redish 2014) on data recorded from single electrode, stereotrode, and tetrode. It allows manual corrections to automated clustering results and can be obtained from http://redishlab.neuroscience.umn.edu/MClust/MClust.html.

2.2.5 NEV2lkit

“NEV2lKit” is a package written in C++ with routines for analysis, visualization, and classification of spikes (Bongard et al. 2014). It is a preprocessor for the analysis of intra- and extracellular spiking neuronal signals. The main purposes served by the toolbox includes: loading ASCII files as well as neural event files (NEV) and extract spike events from them, PCA-based sorting of spikes based on the spike waveform, and saving of spike timestamps, unit information, and the spike signals into ASCII or NEV files. It provides accurate, efficient, and consistency across experiments. Available at http://nev2lkit.sourceforge.net/.

2.2.6 OSort

“OSort” is a template-based, unsupervised, online spike-sorting algorithm written in Matlab (Rutishauser et al. 2006). It uses residual-sum-of-squares-based distance method and custom thresholds to on-the-fly sort of the recorded spikes. As the algorithm is online, the tool uses a technique where the clusters are built and adapted iteratively over the course of the recording. Available at http://www.urut.ch/new/serendipity/index.php?/pages/osort.html.

2.2.7 SpikeOMatic

“SpikeOMatic” is a spike-sorting package developed in R (Pouzat and Chaffiol 2009). It implements two sophisticated data generation models, namely Gaussian mixture model (GMM) and dynamic hidden Markov model (DHMM). For statistical inference for the abovementioned models, the tool makes use of expectation–maximization for GMM and Markov chain Monte Carlo method for DHMM. The package can be downloaded from http://www.biomedicale.univ-paris5.fr/SpikeOMatic/.

2.2.8 Spyke

“Spyke” is a python toolbox for visualizing, navigating, and spike sorting of high-density multichannel extracellular spikes (Spacek et al. 2009). It uses PCA for dimensionality reduction and modified gradient ascent clustering algorithm (Fukunaga and Hostetler 1975; Swindale and Spacek 2014) to classify the features. The sorting method initially assigns each event to a channel and then these channel-based clusters are subdivided into possible distinguished clusters. Available at http://spyke.github.io/.

2.2.9 SpyKING CIRCUS

“SpyKING CIRCUS” is a python toolbox aiming to provide spike sorting for high-density multichannel extracellular spikes (Yger et al. 2018). This semiautomatic spike-sorting package performs highly parallel code execution to handle large number of recording electrodes. Based on a greedy template matching approach and with the help of a smart clustering technique, the package can efficiently sort spikes from up to 4225 channels. The code is available at https://github.com/spyking-circus/spyking-circus with a ground truth dataset at https://zenodo.org/record/1205233/export/hx#.WrORP3XwaV4.

2.2.10 UltraMegaSort2000

“UltraMegaSort2000” is a Matlab-based toolbox for spike detection and clustering which implements a hierarchical clustering scheme using similarities of spike shape and spike timing statistics, and provides false-positive and false-negative errors as quality evaluation metrics (Fee et al. 1996; Hill et al. 2011). The toolbox also provides the users with tools to manually correct the automatically generated clusters. Available at http://physics.ucsd.edu/neurophysics/software.php.

2.2.11 Wave_Clus

“Wave_Clus” is probably the most popular spike-sorting package to date. Developed in Matlab, it uses sophisticated wavelet transformation-based time–frequency analysis for feature selection and a “temperature”-based superparamagnetic clustering (Blatt et al. 1996) method to sort the features into different clusters (Quian Quiroga et al. 2004). It is available for downloading at https://github.com/csn-le/wave_clus.

3 Future Perspectives

In terms of offline processing and analysis of extracellular neuronal signals, the neuroscientific community has seen active participation and contribution from diverse disciplines where individual laboratories developed novel tools and techniques and shared them with the community. As a result of these great efforts, now it has been possible to establish some standardized analyses from these available tools. However, it is noteworthy that, to date, only a few tools are available to the community which are capable of doing the extracellular neuronal signal processing and analysis online which is huge bottleneck for future generation closed-loop real-time systems (Mahmud and Vassanelli 2016).

Also, keeping pace with the technological advancements and speed of the Internet, next-generation online distributed processing and analysis tools are becoming increasingly essential. Exploiting the existing institutional infrastructures, it might be possible to transform the interinstitutional competition into multi-institutional collaboration. In such scenario, a vital requirement is to have secure infrastructures where multi-institutional neuronal signal analysis facilities shall be capable of integrating their data and tools seamlessly (Mahmud et al. 2012b) with the option to protect individual institute’s proprietary data. Additionally, these infrastructures should support user-friendly interfaces allowing even experimentalists with minimal information technology skills to explore, navigate, and use provided scientific data and services. Towards these goals, cloud computing and service-oriented architectures might be utilized through the distributed infrastructure. These approaches allow better representation of responsibilities taken by the different users in accordance to their granted privileges. As an early example of such systems, the Spike-Sorting Evaluation Project initiative (http://spike.g-node.org/) aims to lay a platform for evaluating the performance of spike-sorting algorithms through sharing benchmark data and receiving spike-sorting results for comparison.

Furthermore, considering the growing usage of in vitro MEA technology in applications contributing to improved quality of life for patients, such as pharmacological screening and stem-cell-derived neuronal cultures (Tanskanen et al. 2018), it is becoming increasingly important to standardize experimental protocols and analysis procedures, and data and code sharing to foster experimental reproducibility and validation of obtained results for a combined and accelerated maturation of ground-breaking discoveries.

In our opinion, the development is expected towards:

  • Development of novel tools capable of performing online analyses,

  • Design and implementation of secure and protected systems,

  • Advance on cloud-based web applications,

  • Facilitate easy deployment of data,

  • Reusability and sharing of tools with adaptability to changing requirements, and

  • Empower researchers to share data and functionalities that they want to publish.

4 Conclusion

With the tremendous growth of neurotechnologies, scientists can acquire data from multiple levels and multiple sources. This poses a great challenge to the neuroscientific community to automatically process and analyze those data in order to find meaningful conclusions towards understanding brain’s functioning and to devise translatable technologies towards autonomous diagnosis and treatment strategies for treating brain diseases. This chapter introduced the reader to the popular open-source automated methods for processing and analysis of extracellularly recorded neuronal signals from in vitro devices. Towards the end, some perspective research lines—where future developments are expected—have also been outlined.