A hardware design for in-brain neural spike sorting

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 the process of separating the extracellular recording of the brain signal into one unit activity. There are a number of proposed algorithms for this purpose, but there is still no acceptable solution. In this paper a spike sorting method has been proposed based on the Euclidean distance of the most effective features of spikes represented by principle components (PCs) of the detected and aligned spikes. The assessments of the method, based on signal-to-noise ratio (SNR) representing background noise, showed that the method performed spike sorting to a high level of accuracy.

In this research, we study the effect of feature selection in the spike detection and sorting accuracy. We introduce a new feature representation for neural spikes from multichannel recordings. The features selection plays a significant role in analyzing the response of brain neurons. The more precise selection of features leads to a more accurate spike sorting, which can group spikes more precisely into clusters based on the similarity of spikes. Proper spike sorting will enable the association between spikes and neurons. Different with other threshold-based methods, the cepstrum of spike signals is employed in our method to select the candidates of spike features. To choose the best features among different candidates, the Kolmogorov-Smirnov (KS) test is utilized. Then, we rely on the super paramagnetic method to cluster the neural spikes based on KS features. Simulation results demonstrate that the proposed method not only achieve more accurate clustering results but also reduce computational burden, which implies that it can be applied into real-time spike analysis.

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.

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.

In this paper, we propose an underlay cognitive radio spectrum access (UCRSA) technique to share the licensed 28 GHz millimeter-wave (mmWave) spectrum of one mobile network operator (MNO), termed as primary MNO (p-MNO), with small cells in a building of another MNO, termed as secondary MNO (s-MNO), in a country. The proposed UCRSA considers operating each in-building small cell of an s-MNO at the maximum transmission power on its licensed spectrum, whereas at a reduced transmission power on a shared spectrum of any p-MNO at any time. The reduction in the transmission power varies in accordance with the predefined interference threshold set by the corresponding p-MNO. We derive average capacity, spectral efficiency (SE), and energy efficiency (EE) and carry out extensive numerical and simulation results and analyses for an s-MNO of a country consisting of four MNOs. It is shown that UCRSA can improve SE by 2.67 times, whereas EE by 72.74%, of the s-MNO as compared to that of the traditional static licensed spectrum allocation (SLSA) technique that allocates each MNO an equal amount of 28 GHz mmWave spectrum. Finally, we show that the proposed UCRSA technique can satisfy the expected SE and EE requirements for sixth-generation (6G) mobile systems by reusing the licensed, as well as shared, 28 GHz mmWave spectra of the s-MNO to its small cells of 71.8% less number of buildings than that required by the traditional SLSA technique.

We propose a feature design framework that considers simultaneously performance and computational complexity. In particular, we incorporate these two metrics, which are very important to many low-energy on-chip applications such as implantable neural interfaces, onto an optimization problem. This allows us to strike a balance between the performance of the signal processing task and the computational complexity of the feature extraction process. Simulation results for neural spike sorting demonstrate that by leveraging proposed design framework, we can construct features that outperform other state-of-the-art, low-complexity feature designs, both in terms of classification error and complexity.

A large amount of neurophysiological research is based on the study of biological neural populations. The data is gathered from extra-cellular recordings with multi-electrode arrays (MEAs). The signal is a stream containing an unknown number of neural sources superpositioned with non-stationary noise components. In order to analyze the recorded spike trains, a prior separation into its individual components is required. The increasing number of sensors on a MEA surface demand an automatic spike sorting procedure. In this article the proposed spike sorting method replaces the manual steps with artificial neural networks to enable an unsupervised separation of mixed signal components. Furthermore the artificial neural network based feature extraction allows realtime processing and ensures high flexbibility and adaptivity for unknown signals with non-stationary noise components.

Rate-splitting multiple access (RSMA) and the intelligent reflecting surface (IRS) are novel technologies to improve spectrum and energy efficiencies in multi-antenna networks. In this paper, we investigate a multi-user multiple-input single-output (MISO) simultaneous wireless information and power transfer (SWIPT) system assisted with an IRS and using RSMA. In particular, the IRS assists with data and energy transmission from the base station (BS) to power-splitting (PS) users under a non-linear energy harvesting (EH) model. Our objective is to minimize the transmission power subject to rate and EH requirements at the users. To solve the non-convex problem, we propose a bilevel programming problem in which a genetic algorithm is used to solve the outer optimization problem, and the inner optimization problem is addressed with the semidefinite relaxation (SDR) method. In addition, to study a practical scenario of imperfect channel state information (CSI), we consider a robust beamforming design based on imperfect cascade channels at the BS. Then, we propose a solution based on the S-procedure with the SDR technique. Simulation results show that our proposed RSMA-based scheme provides a significant reduction in transmission power compared to the space-division multiple access (SDMA) and non-orthogonal multiple access (NOMA) methods under perfect and imperfect CSI.

Spike sorting is an important processing step in various neuroscientific and clinical studies. Energy-efficient spike-sorting ASICs are necessary to allow real-time processing of multi-channel, wireless neural recordings. Spike-sorting ASICs have to meet stringent power-density constraints and must provide significant data-rate reduction for wireless transmission. Most existing designs either provide only spike detection for multi-channel processing, or they provide detection and feature extraction only for a single channel. In this paper, we demonstrate the design of a spike-sorting DSP chip that can perform detection, alignment, and feature extraction simultaneously for 64 channels. Spike-sorting algorithms chosen based on a complexity-performance analysis were implemented on ASIC using a MATLAB/Simulink-based architecture design framework. Energy-delay tradeoffs of the design were analyzed to identify the optimal degree of interleaving. The chip was implemented with a modular architecture, and can be configured to process 16, 32, 48, or 64 channels. Inactive cores are power-gated when the chip is operated to process a reduced number of channels. The chip, implemented in a 90-nm CMOS process, has a power dissipation of 130 μW (power density of 30 μW/mm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> ) when processing all 64 channels and provides a data-rate reduction of 91.25% (11.71 Mb/s to 1.02 Mb/s).

Transmitting large amounts of data sensed by multi-electrode array devices is widely considered to be a challenging problem in designing implantable neural recording systems. Spike sorting is an important step to reducing the data bandwidth before wireless data transmission. The feasibility of spike sorting algorithms in scaled CMOS technologies, which typically operate on low frequency neural spikes, is determined largely by its power consumption, a dominant portion of which is leakage power. Our goal is to explore energy saving architectures for online spike sorting without sacrificing performance. In the face of non-stationary neural data, training is not always a feasible option. We present a low-power architecture for real-time online spike sorting that does not require any training period and has the capability to quickly respond to the changing spike shapes.

This paper investigates energy and interference aware Cooperative Routing algorithms in Cognitive Radio Ad-Hoc Networks. Two routing algorithms have been proposed. In Algorithm I, we propose to minimize transmission power based on the rate required, and investigate the total transmission power consumption of the shortest path from source to destination whereby each link on the path is given the option of selecting a relay to further reduce transmission power. In Algorithm II, we consider Primary User (PU) receiver protection by reducing the fractional overlap area of each secondary node with the primary coverage area, and use the average interference probability to optimize route choice again allowing a cooperative relay to be chosen for each link on each possible path. Simulation results show that the proposed Algorithm I can achieve a transmission power saving of up to 30%, while Algorithm II can reduce interference probability for PU receivers by up to 6%.

In recent years, spike sorting has become an emerging technique in multi-channel recording for neuroprosthetic applications. To achieve on-chip real-time processing, it is necessary to design reliable yet low complexity feature extraction and dimensionality reduction to suit low power hardware resources. To satisfy this criterion, this paper proposes asynchronous sampling of first derivative spike waveform features as a dimensionality reduction algorithm. The resulting accuracy of this approach enables identification of the differences temporally localized between the clusters. Directions with maximized or minimized mutual differences are chosen. The classification accuracy of this method is compared with other approaches using different datasets. Using the k-Means clustering algorithm the proposed method achieves an average classification accuracy of > 94% and has very significantly less complexity compared with techniques such as principal component analysis and discrete wavelet transform.

In dense wireless local area network (WLAN) environments, the haphazard deployment of WLAN devices causes them to overlap. In such situations, a coverage adaptation through transmission power control (TPC) is required. At the same time, coverage holes should be taken into consideration for connectivity maintenance among devices. In this paper, a TPC scheme is proposed that reduces overlaps in an autonomous manner, without generating target coverage holes. The problem is formulated as a target coverage problem, in which wireless stations (STAs) are treated as target points, and the transmission power is appropriately controlled by the proposed scheme. In addition, we prove that the target coverage game is a potential game, and the proposed method is guaranteed to converge to a Nash equilibrium after a finite number of iterations. In order to prevent the occurrence of coverage holes that cannot be detected by the available STA information, we propose the exclusion of coverage boundary access points (APs) from reducing transmission power, with the idea that each boundary AP always contains the region covered by it alone. The coverage boundary APs are determined through Delaunay triangulation of all APs. The proposed scheme is evaluated by performing some simulations. The simulation results indicate that the proposed TPC scheme is effective in reducing coverage overlaps without generating coverage holes.

This paper shows that the maximal end-to-end bandwidth in a wireless multihop transmission is 1/2 of bandwidth of communication links and is 1/3 if there are no transmission power control. Then, we propose a method for extension of end-to-end bandwidth by successive reduction of wireless signal transmission power along a multihop message transmission route. Our proposed protocol designed based on AODV achieves 60% wider bandwidth with small reduction of connectivity in simulation.