High-density Multielectrode Array Research Articles

A Deep Learning Approach for the Classification of Neuronal Cell Types.

Classification of neurons from extracellular recordings is mainly limited to putatively excitatory or inhibitory units based on the spike shape and firing patterns. Narrow waveforms are considered to be fast spiking inhibitory neurons and broad waveforms excitatory neurons. The aim of this work is twofold. First, we intend to use the rich spatial information from high-density Multi-Electrode Arrays (MEAs) to make classification more robust; second, we hope to be able to classify sub-types of excitatory and inhibitory neurons. We first built, in simulation, a large dataset of action potentials from detailed neural models. Then, we extracted spike features from the simulated recordings on a high-density Multi-Electrode Array model. Finally, we used a Convolutional Neural Networks (CNN), to classify the different cell types. Compared with the ground truth data from the simulated dataset, the results show that this forward modelling/machine learning approach is very robust in recognizing excitatory and inhibitory spikes (accuracy $\ge 92.15$%). Additionally, the approach can, to a certain extent, correctly classify different cell sub-types. As the detail and fidelity of neural models increase and high-density recordings become available, this approach could become a viable and robust alternative for classification of neural cell types from in-vivo extracellular recordings.

Independent Component Analysis for Fully Automated Multi-Electrode Array Spike Sorting.

In neural electrophysiology, spike sorting allows to separate different neurons from extracellularly measured recordings. It is an essential processing step in order to understand neural activity and it is an unsupervised problem in nature, since no ground truth information is available. There are several available spike sorting packages, but many of them require a manual intervention to curate the results, which makes the process time consuming and hard to reproduce. Here, we focus on high-density Multi-Electrode Array (MEA) recordings and we present a fully automated pipeline based on Independent Component Analysis (ICA). While ICA has been previously investigated for spike sorting, it has never been compared with fully automated state-of-the-art algorithms. We use realistic simulated datasets to compare the spike sorting performance in terms of complexity, signal-to-noise ratio, and recording duration. We show that an ICA-based fully automated spike sorting approach can be a viable alternative approach due to its precision and robustness, but it needs to be optimized for time constraints and requires sufficient density of electrodes to cover active neurons in the proximity of the MEA.

Multielectrodes join the connectome.

High-density multielectrode arrays in turtle cortex allow parallel assessment of functional and structural connectomics.

(Invited) Invasive Cortical Microelectrode Array Longitudinal Performance: Temporal Dynamics of Electrical Impedance Spectroscopy and Multiunit Activity

Addition of neural sensing elements to neurostimulation devices can ‘close the loop’ to provide dynamic, patient-specific neuromodulation. Recently approved medical devices, such as the Neuropace RNS System for drug-resistant epilepsy and the Inspire Upper Airway Stimulation device monitor the patient’s electrophysiological activity and titrate stimulation accordingly, improving therapeutic outcomes. Current devices rely on a few, large electrode contacts; however, high-density electrode interfaces will provide greater information content to maximize the therapeutic benefit. High-density neural interfaces have been explored in the context of brain computer interface devices, which utilize microelectrode arrays to detect neural activity used to control prosthetic devices. Although the high-density multielectrode arrays implanted in human cortex can record from hundreds of neurons simultaneously, the device performance gradually declines within six to twelve months post-implantation as the number of recorded neurons dwindles. Preclinical investigations have identified several key factors implicated in this failure, including poor tissue integration resulting in inflammation, blood-brain barrier compromise and electrode material breakdown. To differentiate between tissue-related, biological failure modalities and material-related physical modalities, we performed electrical impedance spectroscopy (EIS) before, during and after the long-term implantation of high-density electrode arrays in the mouse motor cortex. Three different types of commercial research-grade electrodes were sterilized with ethylene oxide and implanted into the neocortex (Blackrock, Neuronexus and Microprobes for Life Science). Implant duration lasted for between 6-12 months, and EIS was collected across the implant duration. EIS was measured over frequencies ranging from 1Hz to 1MHz with 5 frequency points per decade using a Gamry Reference 600 potentiostat (Gamry Instruments). Multi-unit activity was also collected throughout the course of implantation during periods of spontaneous movement in awake, behaving animals using a 16-channel data acquisition system (Neuralynx). As expected, electrode implantation lead to an overall increase in impedance across three device types. However, the largest increase in impedance was not centered around 1kHz (a common frequency for impedance spectroscopy in vivo), but instead centered between 10-100kHz for all array types. Much of this increase was reduced when EIS was measured after the arrays were explanted from the brain, suggesting that changes in the resistive component of the tissue medium is the largest contribution to in vivo impedance dynamics. Interestingly, all arrays demonstrated a considerable change in impedance phase, centered between 100kHz and 1MHz during the course of implantation. The electrophysiological signals recorded from each animal were analyzed using custom routines (MATLAB, Igor). Mechanical and EMG artifacts were subtracted using coherence-based common average referencing, and then two metrics of neural firing were calculated; the event rate of signals 3.5SD above the mean, and power in the spectral band between 1000 and 1500 Hz. Temporal dynamics of electrophysiological and EIS signals were compared for each electrode across the implant duration. In addition, scanning electron microscopy (SEM) images of the explanted arrays were scored for measures of physical damage, including tip breakage, metallization loss, insulation delamination, cracking or loss and broken connectivity. Characteristics from each of these separate modalities were combined into custom machine learning algorithms to identify relationships between factors. Through this work, we were able to characterize the longitudinal impedance dynamics up to a year post implantation, to dissociate the contribution of biological and physical factors to the impedance change, and perform electrode-by-electrode correlation between impedance, physical damage and electrophysiological signal quality.

Localizing neuronal somata from Multi-Electrode Array in-vivo recordings using deep learning.

With the latest development in the design and fabrication of high-density Multi-Electrode Arrays (MEA) for in-vivo neural recordings, the spatiotemporal information in the recorded signals allows for refined estimation of a neuron's location around the probe. In parallel, advances in computational models for neural activity enables simulation of recordings from neurons with detailed morphology. Our approach uses deep learning algorithms on a large set of such simulation data to extract the 3D position of the neuronal somata. Multi-compartment models from 13 different neural morphologies in layer 5 (L5) of the rat's neocortex are placed at random locations and with different alignments with respect to the MEA. The sodium trough and repolarisation peak images on the MEA serve as input features for a Convolutional Neural Network (CNN), which predicts the neural location robustly and with low error rates. The forward modeling/machine learning approach yields very accurate results for the different morphologies and is able to cope with different neuron alignments.

Unsupervised Spike Sorting for Large-Scale, High-Density Multielectrode Arrays

We present a method for automated spike sorting for recordings with high-density, large-scale multielectrode arrays. Exploiting the dense sampling of single neurons by multiple electrodes, an efficient, low-dimensional representation of detected spikes consisting of estimated spatial spike locations and dominant spike shape features is exploited for fast and reliable clustering into single units. Millions of events can be sorted in minutes, and the method is parallelized and scales better than quadratically with the number of detected spikes. Performance is demonstrated using recordings with a 4,096-channel array and validated using anatomical imaging, optogeneticstimulation, and model-based quality control. A comparison with semi-automated, shape-based spike sorting exposes significant limitations of conventional methods. Our approach demonstrates that it is feasible to reliably isolate the activity of up to thousands of neurons and that dense, multi-channel probes substantially aid reliable spike sorting.

Pan-retinal characterisation of Light Responses from Ganglion Cells in the Developing Mouse Retina

We have investigated the ontogeny of light-driven responses in mouse retinal ganglion cells (RGCs). Using a large-scale, high-density multielectrode array, we recorded from hundreds to thousands of RGCs simultaneously at pan-retinal level, including dorsal and ventral locations. Responses to different contrasts not only revealed a complex developmental profile for ON, OFF and ON-OFF responses, but also unveiled differences between dorsal and ventral RGC responses. At eye-opening, dorsal RGCs of all types were more responsive to light, perhaps indicating an environmental priority to nest viewing for pre-weaning pups. The developmental profile of ON and OFF responses exhibited antagonistic behaviour, with the strongest ON responses shortly after eye-opening, followed by an increase in the strength of OFF responses later on. Further, we found that with maturation receptive field (RF) center sizes decrease, spike-triggered averaged responses to white noise become stronger, and centers become more circular while maintaining differences between RGC types. We conclude that the maturation of retinal functionality is not spatially homogeneous, likely reflecting ecological requirements that favour earlier maturation of the dorsal retina.

Propagation and synchronization of reverberatory bursts in developing cultured networks.

Developing networks of neural systems can exhibit spontaneous, synchronous activities called neural bursts, which can be important in the organization of functional neural circuits. Before the network matures, the activity level of a burst can reverberate in repeated rise-and-falls in periods of hundreds of milliseconds following an initial wave-like propagation of spiking activity, while the burst itself lasts for seconds. To investigate the spatiotemporal structure of the reverberatory bursts, we culture dissociated, rat cortical neurons on a high-density multi-electrode array to record the dynamics of neural activity over the growth and maturation of the network. We find the synchrony of the spiking significantly reduced following the initial wave and the activities become broadly distributed spatially. The synchrony recovers as the system reverberates until the end of the burst. Using a propagation model we infer the spreading speed of the spiking activity, which increases as the culture ages. We perform computer simulations of the system using a physiological model of spiking networks in two spatial dimensions and find the parameters that reproduce the observed resynchronization of spiking in the bursts. An analysis of the simulated dynamics suggests that the depletion of synaptic resources causes the resynchronization. The spatial propagation dynamics of the simulations match well with observations over the course of a burst and point to an interplay of the synaptic efficacy and the noisy neural self-activation in producing the morphology of the bursts.

Rank Order Coding: a Retinal Information Decoding Strategy Revealed by Large-Scale Multielectrode Array Retinal Recordings.

How a population of retinal ganglion cells (RGCs) encodes the visual scene remains an open question. Going beyond individual RGC coding strategies, results in salamander suggest that the relative latencies of a RGC pair encode spatial information. Thus, a population code based on this concerted spiking could be a powerful mechanism to transmit visual information rapidly and efficiently. Here, we tested this hypothesis in mouse by recording simultaneous light-evoked responses from hundreds of RGCs, at pan-retinal level, using a new generation of large-scale, high-density multielectrode array consisting of 4096 electrodes. Interestingly, we did not find any RGCs exhibiting a clear latency tuning to the stimuli, suggesting that in mouse, individual RGC pairs may not provide sufficient information. We show that a significant amount of information is encoded synergistically in the concerted spiking of large RGC populations. Thus, the RGC population response described with relative activities, or ranks, provides more relevant information than classical independent spike count- or latency- based codes. In particular, we report for the first time that when considering the relative activities across the whole population, the wave of first stimulus-evoked spikes is an accurate indicator of stimulus content. We show that this coding strategy coexists with classical neural codes, and that it is more efficient and faster. Overall, these novel observations suggest that already at the level of the retina, concerted spiking provides a reliable and fast strategy to rapidly transmit new visual scenes.

Long-term High-density Microelectrode Array Recordings Evidence Stability of Neuron Locations in Organotypic Hippocampus Slices

Long-term High-density Microelectrode Array Recordings Evidence Stability of Neuron Locations in Organotypic Hippocampus Slices

UnitBrowser - A Tool to Evaluate and Post-Process Units Sorted by Automatic Spike Sorting Algorithms

UnitBrowser - A Tool to Evaluate and Post-Process Units Sorted by Automatic Spike Sorting Algorithms

Spike Detection for Large Neural Populations Using High Density Multielectrode Arrays

An emerging generation of high-density microelectrode arrays (MEAs) is now capable of recording spiking activity simultaneously from thousands of neurons with closely spaced electrodes. Reliable spike detection and analysis in such recordings is challenging due to the large amount of raw data and the dense sampling of spikes with closely spaced electrodes. Here, we present a highly efficient, online capable spike detection algorithm, and an offline method with improved detection rates, which enables estimation of spatial event locations at a resolution higher than that provided by the array by combining information from multiple electrodes. Data acquired with a 4096 channel MEA from neuronal cultures and the neonatal retina, as well as synthetic data, was used to test and validate these methods. We demonstrate that these algorithms outperform conventional methods due to a better noise estimate and an improved signal-to-noise ratio (SNR) through combining information from multiple electrodes. Finally, we present a new approach for analyzing population activity based on the characterization of the spatio-temporal event profile, which does not require the isolation of single units. Overall, we show how the improved spatial resolution provided by high density, large scale MEAs can be reliably exploited to characterize activity from large neural populations and brain circuits.

A super-resolution approach for receptive fields estimation of neuronal ensembles

The Spike Triggered Average (STA) is a classical technique to find a discrete approximation of the Receptive Fields (RFs) of sensory neurons [1], a required analysis in most experimental studies. One important parameter of the STA is the spatial resolution of the estimation, corresponding to the size of the blocks of the checkerboard stimulus images. In general, it is experimentally fixed to reach a compromise: If too small, neuronal responses might be too weak thus leading to RF with low Signal-to-Noise-Ratio; on the contrary, if too large, small RF will be lost, or not described with enough details, because of the coarse approximation. Other solutions were proposed consisting in starting from a small block size and updating it following the neuron response in a closed-loop to increase its response [2; 3; 4]. However, these solutions were designed for single cells and cannot be applied to simultaneous recordings of ensembles of neurons (since each RF has its own size and preferred stimulus). To solve this problem, we introduced a modified checkerboard stimulus where blocks are shifted randomly in space at fixed time steps. This idea is inspired from super-resolution techniques developed in image processing [4]. The main interest is that the block size can be large, enabling strong responses, while the resolution can be finer since it depends on the shift minimum size. In [5] was shown that the STA remains an unbiased RF estimator and, using simulated spike trains from an ensemble of Linear Nonlinear Poisson cascade neurons, it was predicted that this approach improves RF estimation over the neuron ensemble. Here, we test these predictions experimentally on the RFs estimation of 8460 ganglion cells from two mouse retinas, using recordings performed with a large scale high-density multielectrode array. To illustrate the main interest of the approach, in Figure 1 we show a representative example of STA for one neuron where RFs have been obtained using the three following stimuli (all presented during 15min, for one retina displayed at 10 Hz, for the other at 30 Hz): (A) standard checkerboard stimulus with block size of 160μm, (B) standard checkerboard stimulus with block size of 40μm, (C) checkerboard stimulus with block size of 160μm and arbitrary shifts of 40μm in x and y-directions. Results show spatial resolution can be improved in case (C), while nothing could be obtained in (B) by changing only the block size of the standard stimulus. At the population level, plot (D) shows the number of the RFs that could be recovered for each stimuli, using a decision criteria based of the RFs value distribution. Most of the RFs were mapped with both methods (A) and (C) (49.9%). However, the proposed case (C) allows to recover 51% of the mapped RFs at a resolution of 40μm, while in the classical case (A), 41% of the RFs could be found at a resolution of only 160μm. Thus, the method does improve the quality of the RF estimation and the amount of successfully mapped RFs in neural ensembles.

High-density MEA recordings unveil the dynamics of bursting events in Cell Cultures.

High density multielectrode arrays (MEAs) based on CMOS technology (CMOS-MEAs) can simultaneously record extracellular spiking activity in neuronal cultures from 4096 closely spaced microelectrodes. This allows for a finer investigation of neuronal network activity compared to conventional MEAs with a few tens of electrodes. However, the sensing properties of these devices differ. To highlight this aspect, here we investigate and discuss the differences observed when quantifying spontaneous synchronized bursting events (SBEs) in datasets acquired with conventional MEAs and high-density MEAs from comparable hippocampal cultures. We found that datasets acquired with high-density MEAs exhibit collective dynamics similar to conventional arrays, but are characterized by a higher percentage of random spikes, i.e. spikes that are not part of a burst, most probably resulting from the larger recording capability. Additionally, the percentage of electrodes that record a burst is remarkably small on high-density MEAs compared to what can be observed on conventional MEAs and SBEs appear to be propagating in time across the electrode array, by involving shorter sequences of spikes per electrode. Overall, these results highlight a lower level of network synchronization involved in SBEs compared to what has been debated for several decades based on conventional MEA recordings from cell cultures.

Investigating cell culture dynamics combining high density recordings with dimensional reduction techniques.

High density multielectrode array recordings with CMOS-MEAs allow to monitor cell culture activity with unprecedent details respect to previous recording techniques. This is clarifying how network activity develops and is motivating the development of novel data analysis tools. Here, in order to advance in the exploitation of the richness of these large-scale array recordings, we introduce a principal component analysis approach that aims at improving on existing methodologies to describe neural activity events within large networks.

Sloppiness in spontaneously active neuronal networks.

Various plasticity mechanisms, including experience-dependent, spontaneous, as well as homeostatic ones, continuously remodel neural circuits. Yet, despite fluctuations in the properties of single neurons and synapses, the behavior and function of neuronal assemblies are generally found to be very stable over time. This raises the important question of how plasticity is coordinated across the network. To address this, we investigated the stability of network activity in cultured rat hippocampal neurons recorded with high-density multielectrode arrays over several days. We used parametric models to characterize multineuron activity patterns and analyzed their sensitivity to changes. We found that the models exhibited sloppiness, a property where the model behavior is insensitive to changes in many parameter combinations, but very sensitive to a few. The activity of neurons with sloppy parameters showed faster and larger fluctuations than the activity of a small subset of neurons associated with sensitive parameters. Furthermore, parameter sensitivity was highly correlated with firing rates. Finally, we tested our observations from cell cultures on an in vivo recording from monkey visual cortex and we confirm that spontaneous cortical activity also shows hallmarks of sloppy behavior and firing rate dependence. Our findings suggest that a small subnetwork of highly active and stable neurons supports group stability, and that this endows neuronal networks with the flexibility to continuously remodel without compromising stability and function.

Anatomical identification of extracellularly recorded cells in large-scale multielectrode recordings.

This study combines for the first time two major approaches to understanding the function and structure of neural circuits: large-scale multielectrode recordings, and confocal imaging of labeled neurons. To achieve this end, we develop a novel approach to the central problem of anatomically identifying recorded cells, based on the electrical image: the spatiotemporal pattern of voltage deflections induced by spikes on a large-scale, high-density multielectrode array. Recordings were performed from identified ganglion cell types in the macaque retina. Anatomical images of cells in the same preparation were obtained using virally transfected fluorescent labeling or by immunolabeling after fixation. The electrical image was then used to locate recorded cell somas, axon initial segments, and axon trajectories, and these signatures were used to identify recorded cells. Comparison of anatomical and physiological measurements permitted visualization and physiological characterization of numerically dominant ganglion cell types with high efficiency in a single preparation.

Tracking rotors with minimal electrodes: modulation index-based strategy.

High-frequency periodic sources during cardiac fibrillation can be detected by phase mapping techniques. To enable practical therapeutic options for modulating periodic sources (existing techniques require high density multielectrode arrays and real time simultaneous mapping capability), a method to identify electrogram morphologies colocalizing to rotors that can be implemented on few electrograms needs to be devised. Multichannel ventricular fibrillation electrogram data from 7 isolated human hearts using Langendorff setup and intraoperative clinical data from 2 human hearts were included in the analysis. The spatial locations of rotors were identified using phase maps constructed from 112 electrograms. Electrograms were analyzed for repeating patterns and discriminating signal morphologies around the locations of rotors and nonrotors were identified and quantified. Features were extracted from the unipolar electrogram patterns, which corroborated well with the spatial location of rotors. The results suggest that using the proposed modulation index feature, and as low as 1 sample point in the vicinity of the rotors, an accuracy as high as 86% (P<0.001) was obtained in separating rotor locations versus nonrotor locations. The analysis of bipolar electrogram signatures in the vicinity of the rotor locations suggest that 62.5% of the rotors occur at locations where the bipolar electrogram demonstrates continuous activities during ventricular fibrillation. Unipolar electrogram extracted modulation index-based detection of rotors is feasible with few electrodes and has greater detection rate than bipolar approach. This strategy may be suitable for nonarray-based single mapping catheter enabled detection of rotors.

Realization of a diamond based high density multi electrode array by means of Deep Ion Beam Lithography

In the present work we report about a parallel-processing ion beam fabrication technique whereby high-density sub-superficial graphitic microstructures can be created in diamond. Ion beam implantation is an effective tool for the structural modification of diamond: in particular ion-damaged diamond can be converted into graphite, therefore obtaining an electrically conductive phase embedded in an optically transparent and highly insulating matrix.The proposed fabrication process consists in the combination of Deep Ion Beam Lithography (DIBL) and Focused Ion Beam (FIB) milling. FIB micromachining is employed to define micro-apertures in the contact masks consisting of thin (<10μm) deposited metal layers through which ions are implanted in the sample. A prototypical single-cell biosensor was realized with the above described technique. The biosensor has 16 independent electrodes converging inside a circular area of 20μm diameter (typical neuroendocrine cells size) for the simultaneous recording of amperometric signals.

Principles of high–fidelity, high–density 3–d neural recording

New probe technologies, neural amplifier systems, and data acquisition systems enable the extracellular electrical recording of ever greater numbers of neurons in the live mammalian brain. These recordings have the potential to increase our understanding of neuronal network dynamics, but much remains uncharacterized about the possibilities and limitations of extracellular techniques. We explore these possibilities and limitations in the context of spike sorting and probe design. Spike sorting is a critical analysis step for extracellular data, which attempts to separate raw electrode traces into the activity patterns of individual neurons. Given the labor associated with manual spike sorting of large datasets, the necessity for automated spike sorting method will only increase. An automated method would ideally commit no errors in spike assignment – that is, it would associate each extracted individual neuron with all the spikes fired by a single neuron, and with no spikes not fired by that same neuron. The elimination of errors would reduce the reliance on manual validation, saving large amounts of analysis time, and also reduce downstream biases in data analyses introduced by errors in spike sorting. We explore designs for multi-electrode probes and spike sorting methods that in combination allow high accuracy in spike assignment with many neurons extracted. To this end, first, we ask if an automated spike sorting method with zero spike assignment errors is possible. Second, we explore what multi-electrode probe designs produce optimal yield using this method. We have constructed a spike sorting method for the case of spatially dense high channel count extracellular recordings, first applying a well-established source separation technique called Independent Components Analysis (ICA) to continuous recordings. Second, we apply a classifier to the ICA components, keeping only putative single neuron units that are well separated from noise and other units. To test this algorithm, we simulated multielectrode probe data that encompasses many of the realistic variations and noisinesses of natural neural data, including spatial non-linearities in spike shape from individual cells. Running this algorithm against this simulated data, for a wide range of classifier parameters, we find that no spike assignment errors are committed. This result is robust to changes in neural firing rate, neural density, Gaussian noise, and increasing electrode density on the probes. However, for probes with electrode counts similar to those in commercial probes (10-50), only a handful of neurons are extracted. Exploring the space of probe designs, we find that designing probes with higher electrode density (for a fixed area) can compensate for this low yield. As the electrode density on the probe increases, the number of neurons extracted increases to some saturation. We also find that as the electrode density increases, we are able to extract neurons with spike peak magnitudes below the thresholding noise floor. Thus, the construction of very high density multielectrode arrays, coupled to the algorithm here proposed, may yield experimental approaches for recording very large numbers of neurons in the live brain, and automatically analyzing the resulting spike trains.