A Hardware-Efficient Novelty-Aware Spike Sorting Approach for Brain-Implantable Microsystems.

BackgroundPrincipal component analysis (PCA) has been widely employed for automatic neuronal spike sorting. Calculating principal components (PCs) is computationally expensive, and requires complex numerical operations and large memory resources. Substantial hardware resources are therefore needed for hardware implementations of PCA. General Hebbian algorithm (GHA) has been proposed for calculating PCs of neuronal spikes in our previous work, which eliminates the needs of computationally expensive covariance analysis and eigenvalue decomposition in conventional PCA algorithms. However, large memory resources are still inherently required for storing a large volume of aligned spikes for training PCs. The large size memory will consume large hardware resources and contribute significant power dissipation, which make GHA difficult to be implemented in portable or implantable multi-channel recording micro-systems.MethodIn this paper, we present a new algorithm for PCA-based spike sorting based on GHA, namely stream-based Hebbian eigenfilter, which eliminates the inherent memory requirements of GHA while keeping the accuracy of spike sorting by utilizing the pseudo-stationarity of neuronal spikes. Because of the reduction of large hardware storage requirements, the proposed algorithm can lead to ultra-low hardware resources and power consumption of hardware implementations, which is critical for the future multi-channel micro-systems. Both clinical and synthetic neural recording data sets were employed for evaluating the accuracy of the stream-based Hebbian eigenfilter. The performance of spike sorting using stream-based eigenfilter and the computational complexity of the eigenfilter were rigorously evaluated and compared with conventional PCA algorithms. Field programmable logic arrays (FPGAs) were employed to implement the proposed algorithm, evaluate the hardware implementations and demonstrate the reduction in both power consumption and hardware memories achieved by the streaming computingResults and discussionResults demonstrate that the stream-based eigenfilter can achieve the same accuracy and is 10 times more computationally efficient when compared with conventional PCA algorithms. Hardware evaluations show that 90.3% logic resources, 95.1% power consumption and 86.8% computing latency can be reduced by the stream-based eigenfilter when compared with PCA hardware. By utilizing the streaming method, 92% memory resources and 67% power consumption can be saved when compared with the direct implementation of GHA.ConclusionStream-based Hebbian eigenfilter presents a novel approach to enable real-time spike sorting with reduced computational complexity and hardware costs. This new design can be further utilized for multi-channel neuro-physiological experiments or chronic implants.

Spike sorting is a fundamental process for decoding neural activity, involving preprocessing, spike detection, feature extraction, clustering, and validation. However, conventional spike sorting methods are highly fragmented, labor-intensive, and heavily reliant on expert manual curation, limiting their scalability and reproducibility. This challenge has become more pressing with advances in neural recording technology, such as high-density Neuropixels for large-scale neural recording or flexible electrodes for long-term stable recording over months to years. The volume and complexity of these datasets make manual curation infeasible, requiring an automated and scalable solution. Here, we introduce SpikeAgent, a multimodal large language model (LLM)-based AI agent that automates and standardizes the entire spike sorting pipeline. Unlike traditional approaches, SpikeAgent integrates multiple LLM backends, coding functions, and established algorithms, autonomously performing spike sorting with reasoning-based decision-making and real-time interaction with intermediate results. It generates interpretable reports, providing transparent justifications for each sorting decision, enhancing transparency and reliability. We benchmarked SpikeAgent against human experts across various neural recording technology, demonstrating its versatility and ability to achieve curation consistency that are equal to, or even higher than human experts. It also drastically reduces the expertise barrier and accelerates the curation and validation time by orders of magnitude. Moreover, it enables automated interpretability of the neural spiking data, which cannot be achieved by any conventional methods. SpikeAgent presents a paradigm shift in processing signals for neuroscience and brain-computer interfaces, while laying the ground for AI agent-augmented science across various domains.

Spike sorting, i.e., the separation of the firing activity of different neurons from extracellular measurements, is a crucial but often error-prone step in the analysis of neuronal responses. Usually, three different problems have to be solved: the detection of spikes in the extracellular recordings, the estimation of the number of neurons and their prototypical (template) spike waveforms, and the assignment of individual spikes to those putative neurons. If the template spike waveforms are known, template matching can be used to solve the detection and classification problem. Here, we show that for the colored Gaussian noise case the optimal template matching is given by a form of linear filtering, which can be derived via linear discriminant analysis. This provides a Bayesian interpretation for the well-known matched filter output. Moreover, with this approach it is possible to compute a spike detection threshold analytically. The method can be implemented by a linear filter bank derived from the templates, and can be used for online spike sorting of multielectrode recordings. It may also be applicable to detection and classification problems of transient signals in general. Its application significantly decreases the error rate on two publicly available spike-sorting benchmark data sets in comparison to state-of-the-art template matching procedures. Finally, we explore the possibility to resolve overlapping spikes using the template matching outputs and show that they can be resolved with high accuracy.

This thesis was to solve the problem that tremendous amount of data of neural signal seriously confined the number of transmission channels with a limited bandwidth in a digital telemetry system. Although a spike-detection algorithm designed for data reduction had been reported, the signal in between the spikes was lost. Moreover, various non-ideal factors greatly reduced the accuracy of spike detection. This thesis used the commercialized data-compression algorithm, MPEG Layer 3 (MP3), as our data reduction scheme and evaluated its performance on spike sorting. The evaluation results showed that MP3 kept at least 80% accuracy of unit isolation. For reducing the consumption of hardware resource by MP3, MP3 algorithm was further simplified and the corresponding performance was checked. All the algorithm simplification carried out in this thesis compromised barely discernible performance. At last this thesis discussed a method for multi-channel signal analysis, ICA. The effect of MP3 on ICA after the highest level of simplification was simulated. The results showed that the miss probabilities of MP3 were considerably low, regardless of the encoding bit rate. According to all the results mentioned above, we preliminarily validated MP3 a proper neural signal compressor.

Neural spike sorting is used to classify neural spike signals based on neuron type and so is an essential step in decoding brain signals. Because of its computational complexity, spike sorting is generally carried out offline or, at least, using a transmitted signal. In contrast, in-brain spike sorting would reduce the data that needs to be transmitted by orders of magnitude with a corresponding reduction in transmission power. This would enable real-time wireless neural recordings. In this paper, we design and characterize a hardware prototype for in-brain spike sorting. Our design is able to reduce the wireless transmission power by a factor of over 200 over direct transmission. Also, compared with the current state-of-the-art, our design increases the sorting accuracy from 75% to 93% while remaining within hard constraints for power, power density, and real-time processing.

Event Abstract Back to Event Optimizing Spike-Sorting Yield through Selection of Optimal Recording Sites in High-density Microelectrode Arrays Jan Müller1*, Felix Franke2, Michele Fiscella1, Urs Frey3, Douglas Bakkum1 and Andreas Hierlemann1 1 ETH Zurich, Department of Biosystems Science and Engineering, Switzerland 2 ETH Zürich, Department of Biosystems Science and Engineering, Switzerland 3 Riken, Quantitative Biology Center, Japan Motivation Recently, various high-density microelectrode arrays (HDMEA) were developed for in-vitro and in-vivo applications. Some of these arrays feature hundreds to thousands of electrodes, from which a subset can be selected for recording at one time [1-4]. Our goal here is to simultaneously extract the activity of a maximum number of identifiable neurons from such recordings. Careful choice of recording electrodes is required, when the number of recorded neurons exceeds the number of available recording channels. When analyzing data from networks of neurons, one is typically interested in the exact relative timing of the occurrence of action potentials in different neurons. Here, we consider the assignment of recorded action potentials to individual neurons a classification problem. Therefore, we propose a metric relying on linear discriminant analysis (LDA) to compute the expected classification error for a given electrode arrangement. Moreover, we propose an algorithm to automatically find optimal electrode configurations. Material and Methods HDMEA: To demonstrate the feasibility of our method, we used a recently developed HDMEA featuring 26,400 Pt microelectrodes, arranged in a grid-like configuration with a center-to-center pitch of 17.5 μm [3]. 1024 of these electrodes can be selected for simultaneous recording by means of low-noise amplifiers (2.4 μV rms in the action potential signal band 300 Hz–10 kHz). Algorithm: The extent to which action potentials from two neurons can be clearly distinguished depends on their multi-electrode waveforms. The best separation can be achieved if information from all electrodes is available. However, only a subset of all available electrodes can be used for recording at the same time. Our method starts out by successively recording through blocks of all 26,400 electrodes, spike sorting each of the recordings, and building templates for each neuron (Step 1). Using the templates from all neurons and linear discriminant analysis [5], we compute the expected spike sorting error (ESSE) using all electrodes. We then compute for each electrode the ESSE under the assumption that the corresponding electrode would not be used for recording (Step 2). Electrodes, the removal of which increases the ESSE by the least amount will then be discarded (Step 3). Step 2 and 3 are iterated until the desired number of electrodes, i.e., the number of available readout channels is reached. Results We recorded from a patch of hamster retina for ten minutes per electrode. Based on this data, action potential templates were computed, and the algorithm has been applied to identify the optimal recording electrodes. The classification error decays exponentially upon increasing the number of used electrodes. Discussion/Conclusion We have presented a framework to assess the quality of a particular recording scenario. The framework quantifies the error probability for discriminating between all a-priori known templates for all present neurons. The algorithm was applied to recordings from a retina performed with a recently developed HDMEA. Our approach is related to earlier work [6]. However, by relying on linear discriminant analysis, we make - in a Bayesian sense - optimal use of the available recording resources. [1] Lopez, Carolina M., et al. "A 966-electrode neural probe with 384 configurable channels in 0.13μm SOI CMOS", ISSCC (2016), 392-393. [2] Viswam, Vijay, et al. "Multi-Functional Microelectrode Array System featuring 59’760 Electrodes, 2048 Electrophysiology Channels, Impedance and Neurotransmitter Measurement Units", ISSCC (2016), 394-396. [3] Ballini, Marco, et al. "A 1024-Channel CMOS microelectrode array with 26,400 electrodes for recording and stimulation of electrogenic cells in vitro." Solid-State Circuits, IEEE Journal of 49.11 (2014): 2705-2719. [4] Seidl, Karsten, et al. "CMOS-based high-density silicon microprobe arrays for electronic depth control in intracortical neural recording" Microelectromechanical Systems, Journal of, 20(6):1439– 1448, 2011. [5] Franke, Felix, et al. "Bayes Optimal Template Matching for Spike Sorting – Combining Fisher Discriminant Analysis with Optimal Filtering, " J. Comput. Neurosci., vol. 38, no. 3, pp. 439–459, 2015. [6] Vysotska, Olga, "Automatic channel selection and neural signal estimation across channels of neural probes", International Conference on Intelligent Robots and Systems (2014), 1453–1459. Keywords: Retina, CMOS, Spike-sorting, high-density microelectrode array (HDMEA) Conference: MEA Meeting 2016 | 10th International Meeting on Substrate-Integrated Electrode Arrays, Reutlingen, Germany, 28 Jun - 1 Jul, 2016. Presentation Type: Poster Presentation Topic: MEA Meeting 2016 Citation: Müller J, Franke F, Fiscella M, Frey U, Bakkum D and Hierlemann A (2016). Optimizing Spike-Sorting Yield through Selection of Optimal Recording Sites in High-density Microelectrode Arrays. Front. Neurosci. Conference Abstract: MEA Meeting 2016 | 10th International Meeting on Substrate-Integrated Electrode Arrays. doi: 10.3389/conf.fnins.2016.93.00122 Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters. The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated. Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed. For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions. Received: 22 Jun 2016; Published Online: 24 Jun 2016. * Correspondence: Dr. Jan Müller, ETH Zurich, Department of Biosystems Science and Engineering, Basel, Switzerland, jan.mueller@bsse.ethz.ch Login Required This action requires you to be registered with Frontiers and logged in. To register or login click here. Abstract Info Abstract The Authors in Frontiers Jan Müller Felix Franke Michele Fiscella Urs Frey Douglas Bakkum Andreas Hierlemann Google Jan Müller Felix Franke Michele Fiscella Urs Frey Douglas Bakkum Andreas Hierlemann Google Scholar Jan Müller Felix Franke Michele Fiscella Urs Frey Douglas Bakkum Andreas Hierlemann PubMed Jan Müller Felix Franke Michele Fiscella Urs Frey Douglas Bakkum Andreas Hierlemann Related Article in Frontiers Google Scholar PubMed Abstract Close Back to top Javascript is disabled. Please enable Javascript in your browser settings in order to see all the content on this page.

Protocol for machine learning-based classification of neuronal cell types from spike waveforms

Brain-machine interfaces (BMIs) based on extracellular recordings with microelectrodes provide means of observing the activities of neurons that orchestrate fundamental brain function, and are therefore powerful tools for exploring the function of the brain. Due to physical restrictions and risks for post-surgical complications, wired BMIs are not suitable for long-term studies in freely behaving animals. Wireless BMIs ideally solve these problems, but they call for low-complexity techniques for data compression that ensure maximum utilization of the wireless link and energy resources, as well as minimum heat dissipation in the surrounding tissues. In this paper, we analyze the performances of various system architectures that involve spike detection, spike alignment and spike compression. Performance is analyzed in terms of spike reconstruction and spike sorting performance after wireless transmission of the compressed spike waveforms. Compression is performed with transform coding, using five different compression bases, one of which we pay special attention to. That basis is a fixed basis derived, by singular value decomposition, from a large assembly of experimentally obtained spike waveforms, and therefore represents a generic basis specially suitable for compressing spike waveforms. Our results show that a compression factor of 99.8%, compared to transmitting the raw acquired data, can be achieved using the fixed generic compression basis without compromising performance in spike reconstruction and spike sorting. Besides illustrating the relative performances of various system architectures and compression bases, our findings show that compression of spikes with a fixed generic compression basis derived from spike data provides better performance than compression with downsampling or the Haar basis, given that no optimization procedures are implemented for compression coefficients, and the performance is similar to that obtained when the optimal SVD based basis is used.

Brain Machine Interface (BMI) systems demand real-time spike sorting to instantaneously decode the spike trains of simultaneously recorded cortical neurons. Real-time spike sorting, however, requires extensive computational power that is not feasible to implement in implantable BMI architectures, thereby requiring transmission of high-bandwidth raw neural data to an external computer. In this work, we describe a miniaturized, low power, programmable hardware module capable of performing this task within the resource constraints of an implantable chip. The module computes a sparse representation of the spike waveforms followed by "smart" thresholding. This cascade restricts the sparse representation to a subset of projections that preserve the discriminative features of neuron-specific spike waveforms. In addition, it further reduces telemetry bandwidth making it feasible to wirelessly transmit only the important biological information to the outside world, thereby improving the efficiency, practicality and viability of BMI systems in clinical applications.

Spike sorting is difficult when there is high waveforms similarity between different spike or when there is a large number of superimposed spikes in the sample. A new sample optimization method is proposed in the paper, called window-gradient feature. Every spike waveform is segmented into successive fragments in terms of the width σ, and each segment is called a window. Then calculate the gradient change of each window and use the ratio as the new alternative feature of the window. Finally, all window-gradient features of the spike waveforms are used to replace the original waveform features for spike sporting. The method is verified on the simulation data at different SNR. The experimental results show that when using SVM for spike sorting, the optimization effect of the proposed method is better than PCA, especially for data sets data sets with high noise or large sampling waveform similarity.

During the analysis of neurobiological signals acquired during neurosurgical procedures such as Deep Brain Stimulation, one focuses mainly on the signal’s background and spiking activity. Spikes, the electrical impulses generated by neuron cells, are events lasting about 1.5 ms with voltage typically well below 200 \(\upmu \)V. As the shape of spike derives from the physical construction of the neuron’s cell membrane and its distance from the recording point, one can infer that spikes of different shapes were generated by different neuron cells. This information can be beneficial during the analysis of the acquired signals, and the procedure of grouping spikes according to their shapes, is called spike sorting. Typically in this procedure, there are defined various attributes characterizing spike shape, and subsequently, the standard clustering method is used to produce groups of spikes of different shapes. This may cause some difficulties that arise both from the definition of good discriminating attributes as well as from the fact that one does not know beforehand how many clusters of shapes to expect. The approach presented in this paper is different, using the fully unsupervised and explainable AI, it generates groups of spike shapes in a fully automatic way. No attributes have to be defined nor the expected number of shape clusters provided. Spike waveforms are fed into the autoencoder that has defined additional constraints on its latent layers. After training, data obtained from those layers can be used for the purpose of spike shape clustering. Beside shape class, one also receives information on how much given spike shape differs from the mean shape of its class.

We propose an unsupervised compressed sensing (CS)-based framework to compress, recover, and cluster neural action potentials. This framework can be easily integrated into high-density multi-electrode neural recording VLSI systems. Embedding spectral clustering and group structures in dictionary learning, we extend the proposed framework to unsupervised spike sorting without prior label information. Additionally, we incorporate group sparsity concepts in the dictionary learning to enable the framework for multi-channel neural recordings, as in tetrodes. To further improve spike sorting success rates in the CS framework, we embed template matching in sparse coding to jointly predict clusters of spikes. Our experimental results demonstrate that the proposed CS-based framework can achieve a high compression ratio (8:1 to 20:1), with a high quality reconstruction performance (>8 dB) and a high spike sorting accuracy (>90%).

Spike sorting is one of the cornerstones of extracellular electrophysiology. By leveraging advanced signal processing and data analysis techniques, spike sorting makes it possible to detect, isolate, and map single neuron spiking activity from both in vivo and in vitro extracellular electrophysiological recordings. A crucial step of any spike sorting pipeline is to reduce the dimensionality of the recorded spike waveform data. Reducing the dimensionality of the processed data is a near-universal practice, fundamentally motivated by the use of clustering algorithms responsible to detect, isolate, and sort the recorded putative neurons. In this paper, we propose and illustrate on both synthetic and experimental data that employing the nonlinear dimensionality reduction technique Uniform Manifold Approximation and Projection (UMAP) can drastically improve the performance, efficiency, robustness, and scalability of spike sorting pipelines without increasing their computational cost. We show how replacing the linear or ad hoc, expert-defined, supervised nonlinear dimensionality reduction methods commonly used in spike sorting pipelines by the unsupervised, mathematically grounded, nonlinear dimensionality reduction method provided by UMAP drastically increases the number of correctly sorted neurons, makes the identification of quieter, seldom spiking neurons more reliable, enables deeper and more precise explorations and analysis of the neural code, and paves new ways toward more efficient and end-to-end automatable spike sorting pipelines of large-scale extracellular neural recording as those produced by high-density multielectrode arrays.

This paper proposed a design of a closed-loop, bi-directional brain-machine-interface (BMI). The proposed chip consists of 16-channel neural acquisition, 8-channel neural stimulation, action potential detection, feature extraction and support vector machine (SVM) for spike sorting. A closed-loop control strategy is utilized to trigger different stimulation pattern according to different detected spike categories. The 16-channel neural signal acquisition is comprised of 16-channel low noise amplifier (LNA) which shares a single Programmable Gain Amplifier (PGA) draws a total current of 218uA under a power supply of 3.3V. Arbitrary combination is realized for the stimulator channel control. The frequency, the output current amplitude, the pulse width and the burst number are all tunable. The first and second derivative extrema method is applied to the detected action potential before spike sorting, realizing a great reduction on the data to be processed in the SVM. A single two-class sorting module is implemented as a results from the tradeoff between area and latency. More than two-class spike sorting is realized by a repeating of the two-class sorting procedure. The design has been fabricated in TSMC 180nm HV COMS process, occupying a silicon area of 5mm × 2mm.

Objective. The OSort algorithm, a pivotal unsupervised spike sorting method, has been implemented in dedicated hardware devices for real-time spike sorting. However, due to the inherent complexity of neural recording environments, OSort still grapples with numerous transient cluster occurrences during the practical sorting process. This leads to substantial memory usage, heavy computational load, and complex hardware architectures, especially in noisy recordings and multi-channel systems. Approach. This study introduces an optimized OSort algorithm (opt-OSort) which utilizes correlation coefficient (CC), instead of Euclidean distance as classification criterion. The CC method not only bolsters the robustness of spike classification amidst the diverse and ever-changing conditions of physiological and recording noise environments, but also can finish the entire sorting procedure within a fixed number of cluster slots, thus preventing a large number of transient clusters. Moreover, the opt-OSort incorporates two configurable validation loops to efficiently reject cluster outliers and track recording variations caused by electrode drifting in real-time. Main results. The opt-OSort significantly reduces transient cluster occurrences by two orders of magnitude and decreases memory usage by 2.5–80 times in the number of pre-allocated transient clusters compared with other hardware implementations of OSort. The opt-OSort maintains an accuracy comparable to offline OSort and other commonly-used algorithms, with a sorting time of 0.68 µs as measured by the hardware-implemented system in both simulated datasets and experimental data. The opt-OSort’s ability to handle variations in neural activity caused by electrode drifting is also demonstrated. Significance. These results present a rapid, precise, and robust spike sorting solution suitable for integration into low-power, portable, closed-loop neural control systems and brain–computer interfaces.