A hippocampal cognitive prosthesis: multi-input, multi-output nonlinear modeling and VLSI implementation

Planar amorphous silicon carbide microelectrode arrays for chronic recording in rat motor cortex

Chronic implantation of intracortical microelectrode arrays (MEAs) capable of recording from individual neurons can be used for the development of brain-machine interfaces. However, these devices show reduced recording capabilities under chronic conditions due, at least in part, to the brain's foreign body response (FBR). This creates a need for MEAs that can minimize the FBR to possibly enable long-term recording. A potential approach to reduce the FBR is the use of MEAs with reduced cross-sectional geometries. Here, we fabricated 4-shank amorphous silicon carbide (a-SiC) MEAs and implanted them into the motor cortex of seven female Sprague-Dawley rats. Each a-SiC MEA shank was 8 μm thick by 20 μm wide and had sixteen sputtered iridium oxide film (SIROF) electrodes (4 per shank). A-SiC was chosen as the fabrication base for its high chemical stability, good electrical insulation properties, and amenability to thin film fabrication. Electrochemical analysis and neural recordings were performed weekly for 4 months. MEAs were characterized pre-implantation in buffered saline and in vivo using electrochemical impedance spectroscopy and cyclic voltammetry at 50 mV/s and 50,000 mV/s. Neural recordings were analyzed for single unit activity. At the end of the study, animals were sacrificed for immunohistochemical analysis. We observed statistically significant, but small, increases in 1 and 30 kHz impedance values and 50,000 mV/s charge storage capacity over the 16-week implantation period. Slow sweep 50 mV/s CV and 1 Hz impedance did not significantly change over time. Impedance values increased from 11.6 MΩ to 13.5 MΩ at 1 Hz, 1.2 MΩ-2.9 MΩ at 1 kHz, and 0.11 MΩ-0.13 MΩ at 30 kHz over 16 weeks. The median charge storage capacity of the implanted electrodes at 50 mV/s was 58.1 mC/cm2 on week 1 and 55.9 mC/cm2 on week 16, and at 50,000 mV/s, 4.27 mC/cm2 on week 1 and 5.93 mC/cm2 on week 16. Devices were able to record neural activity from 92% of all active channels at the beginning of the study, At the study endpoint, a-SiC devices were still recording single-unit activity on 51% of electrochemically active electrode channels. In addition, we observed that the signal-to-noise ratio experienced a small decline of -0.19 per week. We also classified observed units as fast and slow repolarizing based on the trough-to-peak time. Although the overall presence of single units declined, fast and slow repolarizing units declined at a similar rate. At recording electrode depth, immunohistochemistry showed minimal tissue response to the a-SiC devices, as indicated by statistically insignificant differences in activated glial cell response between implanted brains slices and contralateral sham slices at 150 μm away from the implant location, as evidenced by GFAP staining. NeuN staining revealed the presence of neuronal cell bodies close to the implantation site, again statistically not different from a contralateral sham slice. These results warrant further investigation of a-SiC MEAs for future long-term implantation neural recording studies.

Neurotechnology and ethics guidelines for human enhancement: The case of the hippocampal cognitive prosthesis

Neurotechnologies offer both therapeutic and enhancement potential. In this article, we demonstrate how ethics guidelines can help with critical reflection on their potential for enhancement. We do this through the case of the hippocampal cognitive prosthesis. This prothesis developed in the US, has primarily therapeutic ends, with scope for enhancement. This technology raises several ethical issues, including as related to identity and memory, autonomy and authenticity. In the first section, we outline what we mean by enhancement, and introduce neurotechnologies generally and the hippocampal cognitive prosthesis specifically, with an introduction to generally relevant ethical issues. In the second section, we outline ethical issues pertinent to the hippocampal cognitive prosthesis and explore how ethics guidelines can help to promote essential critical reflection on a technology like this. Through all this, our emphasis is to balance between technological optimism and caution, especially where technologies have enhancement potential.

Distinctive properties of biological neural networks and recent advances in bottom-up approaches toward a better biologically plausible neural network

Although it may appear infeasible and impractical, building artificial intelligence (AI) using a bottom-up approach based on the understanding of neuroscience is straightforward. The lack of a generalized governing principle for biological neural networks (BNNs) forces us to address this problem by converting piecemeal information on the diverse features of neurons, synapses, and neural circuits into AI. In this review, we described recent attempts to build a biologically plausible neural network by following neuroscientifically similar strategies of neural network optimization or by implanting the outcome of the optimization, such as the properties of single computational units and the characteristics of the network architecture. In addition, we proposed a formalism of the relationship between the set of objectives that neural networks attempt to achieve, and neural network classes categorized by how closely their architectural features resemble those of BNN. This formalism is expected to define the potential roles of top-down and bottom-up approaches for building a biologically plausible neural network and offer a map helping the navigation of the gap between neuroscience and AI engineering.

Nonlinear dynamical modeling of neural activity using volterra series with GA-enhanced particle swarm optimization algorithm

In order to improve the modeling performance of Volterra sequence for nonlinear neural activity, in this paper, a new optimization algorithm is proposed to identify Volterra sequence parameters. Algorithm combines the advantages of particle swarm optimization (PSO) and genetic algorithm (GA) improve the performance of the identification of nonlinear model parameters from rapidity and accuracy. In the modeling experiments of neural signal data generated by the neural computing model and clinical neural data set in this paper, the proposed algorithm shows its excellent potential in nonlinear neural activity modeling. Compared with PSO and GA, the algorithm can achieve less identification error, and better balance the convergence speed and identification error. Further, we explore the influence of algorithm parameters on identification efficiency, which provides possible guiding significance for parameter setting in practical application of the algorithm.

A Double-Layer Multi-Resolution Classification Model for Decoding Spatiotemporal Patterns of Spikes With Small Sample Size

We build a double-layer, multiple temporal-resolution classification model for decoding single-trial spatiotemporal patterns of spikes. The model takes spiking activities as input signals and binary behavioral or cognitive variables as output signals and represents the input-output mapping with a double-layer ensemble classifier. In the first layer, to solve the underdetermined problem caused by the small sample size and the very high dimensionality of input signals, B-spline functional expansion and L1-regularized logistic classifiers are used to reduce dimensionality and yield sparse model estimations. A wide range of temporal resolutions of neural features is included by using a large number of classifiers with different numbers of B-spline knots. Each classifier serves as a base learner to classify spatiotemporal patterns into the probability of the output label with a single temporal resolution. A bootstrap aggregating strategy is used to reduce the estimation variances of these classifiers. In the second layer, another L1-regularized logistic classifier takes outputs of first-layer classifiers as inputs to generate the final output predictions. This classifier serves as a meta-learner that fuses multiple temporal resolutions to classify spatiotemporal patterns of spikes into binary output labels. We test this decoding model with both synthetic and experimental data recorded from rats and human subjects performing memory-dependent behavioral tasks. Results show that this method can effectively avoid overfitting and yield accurate prediction of output labels with small sample size. The double-layer, multi-resolution classifier consistently outperforms the best single-layer, single-resolution classifier by extracting and utilizing multi-resolution spatiotemporal features of spike patterns in the classification.

Neural interface systems with on-device computing: machine learning and neuromorphic architectures

Development of neural interface and brain-machine interface (BMI) systems enables the treatment of neurological disorders including cognitive, sensory, and motor dysfunctions. While neural interfaces have steadily decreased in form factor, recent developments target pervasive implantables. Along with advances in electrodes, neural recording, and neurostimulation circuits, integration of disease biomarkers and machine learning algorithms enables real-time and on-site processing of neural activity with no need for power-demanding telemetry. This recent trend on combining artificial intelligence and machine learning with modern neural interfaces will lead to a new generation of low-power, smart, and miniaturized therapeutic devices for a wide range of neurological and psychiatric disorders. This paper reviews the recent development of the 'on-chip' machine learning and neuromorphic architectures, which is one of the key puzzles in devising next-generation clinically viable neural interface systems.

A neuromimetic realization of hippocampal CA1 for theta wave generation

Recent advances in neural engineering allowed the development of neuroprostheses which facilitate functionality in people with neurological problems. In this research, a real-time neuromorphic system is proposed to artificially reproduce the theta wave and firing patterns of different neuronal populations in the CA1, a sub-region of the hippocampus. The hippocampal theta oscillations (4-12 Hz) are an important electrophysiological rhythm that contributes in various cognitive functions, including navigation, memory, and novelty detection. The proposed CA1 neuromimetic circuit includes 100 linearized Pinsky-Rinzel neurons and 668 excitatory and inhibitory synapses on a field programmable gate array (FPGA). The implemented spiking neural network of the CA1 includes the main neuronal populations for the theta rhythm generation: excitatory pyramidal cells, PV+ basket cells, and Oriens Lacunosum-Moleculare (OLM) cells which are inhibitory interneurons. Moreover, the main inputs to the CA1 region from the entorhinal cortex via the perforant pathway, the CA3 via Schaffer collaterals, and the medial septum via fimbria-fornix are also implemented on the FPGA using a bursting leaky-integrate and fire (LIF) neuron model. The results of hardware realization show that the proposed CA1 neuromimetic circuit successfully reconstructs the theta oscillations and functionally illustrates the phase relations between firing responses of the different neuronal populations. It is also evaluated the impact of medial septum elimination on the firing patterns of the CA1 neuronal population and the theta wave's characteristics. This neuromorphic system can be considered as a potential platform that opens opportunities for neuroprosthetic applications in future works.

The future of neuromodulation: smart neuromodulation

Introduction: The International Neuromodulation Society defines neuromodulation as the alteration of nerve activity through targeted delivery of a stimulus, such as electrical stimulation or chemical agents, to specific neurological sites in the body.Areas covered: In the near future (<5 years) increasingly complex implantable neuromodulation systems will enter the market. These devices are capable of closed-loop stimulation and the delivery of novel stimulation designs, pushing the need for upgradability. But what about the near-to-far future, meaning 5-10 years from now?Expert opinion: We propose that neuromodulation in the near to far future (5-10 years) will involve integration of adaptive network neuromodulation with predictive artificial intelligence, automatically adjusted by brain and external sensors, and controlled via cloud-based applications. The components will be introduced in a phased approach, culminating in a fully autonomous brain-stimulator-cloud interface. This may, in the long future (>10 years), lead to the brain of the future, a brain with integrated artificial intelligence.

The complement cascade at the Utah microelectrode-tissue interface

Devices implanted within the central nervous system (CNS) are subjected to tissue reactivity due to the lack of biocompatibility between implanted material and the cells' microenvironment. Studies have attributed blood-brain barrier disruption, inflammation, and oxidative stress as main contributing factors that lead to electrode recording failure. The complement cascade is a part of the innate immunity that focuses on recognizing and targeting foreign objects; however, its role in the context of neural implants is substantially unknown. In this study, we implanted a non-functional 4x4 Utah microelectrode array (UEA) into the somatosensory cortex and studied the complement cascade via combined gene and immunohistochemistry quantification at acute (48-h), sub-acute (1-week), and early chronic (4-weeks) time points. The results of this study demonstrate the activation and continuation of the complement cascade at the electrode-tissue interface, illustrating the therapeutic potential of modulating the foreign body response via the complement cascade.

Plasticity and Adaptation in Neuromorphic Biohybrid Systems

Neuromorphic systems take inspiration from the principles of biological information processing to form hardware platforms that enable the large-scale implementation of neural networks. The recent years have seen both advances in the theoretical aspects of spiking neural networks for their use in classification and control tasks and a progress in electrophysiological methods that is pushing the frontiers of intelligent neural interfacing and signal processing technologies. At the forefront of these new technologies, artificial and biological neural networks are tightly coupled, offering a novel "biohybrid" experimental framework for engineers and neurophysiologists. Indeed, biohybrid systems can constitute a new class of neuroprostheses opening important perspectives in the treatment of neurological disorders. Moreover, the use of biologically plausible learning rules allows forming an overall fault-tolerant system of co-developing subsystems. To identify opportunities and challenges in neuromorphic biohybrid systems, we discuss the field from the perspectives of neurobiology, computational neuroscience, and neuromorphic engineering.

A compiler for biological networks on silicon chips

The explosive growth in semiconductor integrated circuits was made possible in large part by design automation software. The design and/or analysis of synthetic and natural circuits in living cells could be made more scalable using the same approach. We present a compiler which converts standard representations of chemical reaction networks and circuits into hardware configurations that can be used to simulate the network on specialized cytomorphic hardware. The compiler also creates circuit–level models of the target configuration, which enhances the versatility of the compiler and enables the validation of its functionality without physical experimentation with the hardware. We show that this compiler can translate networks comprised of mass–action kinetics, classic enzyme kinetics (Michaelis–Menten, Briggs–Haldane, and Botts–Morales formalisms), and genetic repressor kinetics, thereby allowing a large class of models to be transformed into a hardware representation. Rule–based models are particularly well–suited to this approach, as we demonstrate by compiling a MAP kinase model. Development of specialized hardware and software for simulating biological networks has the potential to enable the simulation of larger kinetic models than are currently feasible or allow the parallel simulation of many smaller networks with better performance than current simulation software.

We present a “silicon compiler” that is capable of translating biochemical models encoded in the SBML standard into specialized analog cytomorphic hardware and transfer function–level simulations of such hardware. We show how the compiler and hardware address challenges in analog computing: 1) We ensure that the integration of errors due to the mismatch between analog circuit parameters does not become infinite over time but always remains finite via the use of total variables (the solution of the “divergence problem”); 2) We describe the compilation process through a series of examples using building blocks of biological networks, and show the results of compiling two SBML models from the literature: the Elowitz repressilator model and a rule–based model of a MAP kinase cascade. Source code for the compiler is available at https://doi.org/10.5281/zenodo.3948393.

A comprehensive model-based framework for optimal design of biomimetic patterns of electrical stimulation for prosthetic sensation

OBJECTIVE

Touch and proprioception are essential to motor function as shown by the movement deficits that result from the loss of these senses, e.g. due to neuropathy of sensory nerves. To achieve a high-performance brain-controlled prosthetic arm/hand thus requires the restoration of somatosensation, perhaps through intracortical microstimulation (ICMS) of somatosensory cortex (S1). The challenge is to generate patterns of neuronal activation that evoke interpretable percepts. We present a framework to design optimal spatiotemporal patterns of ICMS (STIM) that evoke naturalistic patterns of neuronal activity and demonstrate performance superior to four previous approaches.

APPROACH

We recorded multiunit activity from S1 during a center-out reach task (from proprioceptive neurons in Brodmann's area 2) and during application of skin indentations (from cutaneous neurons in Brodmann's area 1). We implemented a computational model of a cortical hypercolumn and used a genetic algorithm to design STIM that evoked patterns of model neuron activity that mimicked their experimentally-measured counterparts. Finally, from the ICMS patterns, the evoked neuronal activity, and the stimulus parameters that gave rise to it, we trained a recurrent neural network (RNN) to learn the mapping function between the physical stimulus and the biomimetic stimulation pattern, i.e. the sensory encoder to be integrated into a neuroprosthetic device.

MAIN RESULTS

We identified ICMS patterns that evoked simulated responses that closely approximated the measured responses for neurons within 50 µm of the electrode tip. The RNN-based sensory encoder generalized well to untrained limb movements or skin indentations. STIM designed using the model-based optimization approach outperformed STIM designed using existing linear and nonlinear mappings.

SIGNIFICANCE

The proposed framework produces an encoder that converts limb state or patterns of pressure exerted onto the prosthetic hand into STIM that evoke naturalistic patterns of neuronal activation.

Model Predictive Control for Seizure Suppression Based on Nonlinear Auto-Regressive Moving-Average Volterra Model

This article investigates a closed-loop brain stimulation method based on model predictive control strategy to suppress epileptic seizures. A neural mass model (NMM), exhibiting the normal and various epileptic seizures by changing physiologically meaningful parameters, is used as a black-box model of the brain. Based on system identification, an auto-regressive moving-average Volterra model is established to reveal the relationship between stimulation and neuronal responses. Then, the model predictive control strategy is implemented based the Volterra model, which can generate an optimal stimulation waveform to eliminate epileptiform waves. The computational simulation results indicate the proposed closed-loop control strategy can optimize the stimulation waveform without particular knowledge of the physiological properties of the system. The robustness of the proposed control strategy to system disturbances makes it more appropriate for future clinical application.

Hippocampal CA1 and CA3 neural recording in the human brain: validation of depth electrode placement through high-resolution imaging and electrophysiology

OBJECTIVE

Intracranial human brain recordings typically utilize recording systems that do not distinguish individual neuron action potentials. In such cases, individual neurons are not identified by location within functional circuits. In this paper, verified localization of singly recorded hippocampal neurons within the CA3 and CA1 cell fields is demonstrated.

METHODS

Macro-micro depth electrodes were implanted in 23 human patients undergoing invasive monitoring for identification of epileptic seizure foci. Individual neurons were isolated and identified via extracellular action potential waveforms recorded via macro-micro depth electrodes localized within the hippocampus. A morphometric survey was performed using 3T MRI scans of hippocampi from the 23 implanted patients, as well as 46 normal (i.e., nonepileptic) patients and 26 patients with a history of epilepsy but no history of depth electrode placement, which provided average dimensions of the hippocampus along typical implantation tracks. Localization within CA3 and CA1 cell fields was tentatively assigned on the basis of recording electrode site, stereotactic positioning of the depth electrode in comparison with the morphometric survey, and postsurgical MRI. Cells were selected as candidate CA3 and CA1 principal neurons on the basis of waveform and firing rate characteristics and confirmed within the CA3-to-CA1 neural projection pathways via measures of functional connectivity.

RESULTS

Cross-correlation analysis confirmed that nearly 80% of putative CA3-to-CA1 cell pairs exhibited positive correlations compatible with feed-forward connection between the cells, while only 2.6% exhibited feedback (inverse) connectivity. Even though synchronous and long-latency correlations were excluded, feed-forward correlation between CA3-CA1 pairs was identified in 1071 (26%) of 4070 total pairs, which favorably compares to reports of 20%-25% feed-forward CA3-CA1 correlation noted in published animal studies.

CONCLUSIONS

This study demonstrates the ability to record neurons in vivo from specified regions and subfields of the human brain. As brain-machine interface and neural prosthetic research continues to expand, it is necessary to be able to identify recording and stimulation sites within neural circuits of interest.

Accelerating Estimation of a Multi-Input Multi-Output Model of the Hippocampus with a Parallel Computing Strategy

To build hippocampal memory prosthesis for restoring memory functions, we previously developed and implemented a multi-input multi-output (MIMO) nonlinear dynamic model of the hippocampus. This model can successfully predict hippocampal output spike activities based on input spike activities, and thus be used to drive microstimulation to bypass the damaged hippocampal region. Building such a MIMO model involves estimations of a large number of model coefficients, which typically takes hundreds of hours using a single personal computer. In practice, however, due to the requirement of medical care and clinical trials, the modeling processes must be completed within 72 hours after the recording, so that models can be used to drive stimulations. To solve this problem, we utilized a parallelization strategy to divide the whole MIMO model computation involving iterative estimation and optimization into independent computing tasks that can be performed simultaneously in multiple computer nodes. Such a strategy was implemented on the high-performance computing cluster at the University of Southern California. It reduced the model estimation time to tens of hours and thus allowed us to complete the modeling process within the required time frame to further test model-driven electrical stimulation for the hippocampal memory prosthesis.

Nonlinear dynamics based machine learning: Utilizing dynamics-based flexibility of nonlinear circuits to implement different functions

The core element of machine learning is a flexible, universal function approximator that can be trained and fit into the data. One of the main challenges in modern machine learning is to understand the role of nonlinearity and complexity in these universal function approximators. In this research, we focus on nonlinear complex systems, and show their capability in representation and learning of different functions. Complex nonlinear dynamics and chaos naturally yield an almost infinite diversity of dynamical behaviors and functions. Physical, biological and engineered systems can utilize this diversity to implement adaptive, robust behaviors and operations. A nonlinear dynamical system can be considered as an embodiment of a collection of different possible behaviors or functions, from which different behaviors or functions can be chosen as a response to different conditions or problems. This process of selection can be manual in the sense that one can manually pick and choose the right function through directly setting parameters. Alternatively, we can automate the process and allow the system itself learn how to do it. This creates an approach to machine learning, wherein the nonlinear dynamics represents and embodies different possible functions, and it learns through training how to pick the right function from this function space. We report on how we utilized nonlinear dynamics and chaos to design and fabricate nonlinear dynamics based, morphable hardware in silicon as a physical embodiment for different possible functions. We demonstrate how this flexible, morphable hardware learns through learning and searching algorithms such as genetic algorithm to implement different desired functions. In this approach, we combine two powerful natural and biological phenomenon, Darwinian evolution and nonlinear dynamics and chaos, as a dynamics-oriented approach to designing intelligent, adaptive systems with applications. Nonlinear dynamics embodies different functions at the hardware level, while an evolutionary method is utilized in order to find the parameters to implement the right function.

Delaying memory decline: different options and emerging solutions

Memory decline can be a devastating disease and increases in aging Western populations. Memory enhancement technologies hold promise for this and other conditions. Approaches include stem cell transplantation, which improved memory in several animal studies as well as vaccination against Alzheimer´s disease (AD) by β-amyloid antibodies. For a positive clinical effect, the vaccine should probably be administered over a long period of time and before amyloid pathologies manifest in the brain. Different drugs, such as erythropoietin or antiplatelet therapy, improve memory in neuropsychiatric diseases or AD or at least in animal studies. Omega-3 polyunsaturated fatty acid-rich diets improve memory through the gut-brain axis by altering the gut flora through probiotics. Sports, dancing, and memory techniques (e.g., Method of Loci) utilize behavioral approaches for memory enhancement, and were effective in several studies. Augmented reality (AR) is an auspicious way for enhancing memory in real time. Future approaches may include memory prosthesis for head-injured patients and light therapy for restoring memory in AD. Memory enhancement in humans in health and disease holds big promises for the future. Memory training helps only in mild or no impairment. Clinical application requires further investigation.

A K-Medoids based Point-Process Modeling on Neural Spike Transformation using Binless Kernel

A neural prosthesis is designed to compensate for cognitive functional losses by modeling the information transmission among cortical areas. Existing methods generally build a generalized linear model to approximate the nonlinear transformation among two areas, and use the temporal information of the neural spike with low efficiency. It is essential to efficiently model the nonlinearity embedded in spike generation and transmission for the real-time. This paper proposes a nonlinear point-process model to describe spike-in and spike-out transformation using the theory of reproducing kernel Hilbert space (RKHS) and the binless kernel on spike trains. The binless kernel efficiently maps exact spike timing information to the RKHS to describe nonlinear transformations with global minimum regardless of the weight initialization. A streaming K-medoids algorithm is introduced to select typical and important features in this nonlinear binless kernel for further modeling. We test our model on the nonlinearly generated synthetic neural spike trains, and compare with the existing spike transformation methods, such as Volterra model and staged point-process model. The results show that our model has higher goodness-of-fit evaluated by Kolmogorov-Smirnov test and less variance on the prediction results, which indicates the potential better modeling approach for neural prosthesis application.

Real-time cerebellar neuroprosthetic system based on a spiking neural network model of motor learning

OBJECTIVE

Damage to the brain, as a result of various medical conditions, impacts the everyday life of patients and there is still no complete cure to neurological disorders. Neuroprostheses that can functionally replace the damaged neural circuit have recently emerged as a possible solution to these problems. Here we describe the development of a real-time cerebellar neuroprosthetic system to substitute neural function in cerebellar circuitry for learning delay eyeblink conditioning (DEC).

APPROACH

The system was empowered by a biologically realistic spiking neural network (SNN) model of the cerebellar neural circuit, which considers the neuronal population and anatomical connectivity of the network. The model simulated synaptic plasticity critical for learning DEC. This SNN model was carefully implemented on a field programmable gate array (FPGA) platform for real-time simulation. This hardware system was interfaced in in vivo experiments with anesthetized rats and it used neural spikes recorded online from the animal to learn and trigger conditioned eyeblink in the animal during training.

MAIN RESULTS

This rat-FPGA hybrid system was able to process neuronal spikes in real-time with an embedded cerebellum model of ~10 000 neurons and reproduce learning of DEC with different inter-stimulus intervals. Our results validated that the system performance is physiologically relevant at both the neural (firing pattern) and behavioral (eyeblink pattern) levels.

SIGNIFICANCE

This integrated system provides the sufficient computation power for mimicking the cerebellar circuit in real-time. The system interacts with the biological system naturally at the spike level and can be generalized for including other neural components (neuron types and plasticity) and neural functions for potential neuroprosthetic applications.

A Neuromorphic Prosthesis to Restore Communication in Neuronal Networks

Recent advances in bioelectronics and neural engineering allowed the development of brain machine interfaces and neuroprostheses, capable of facilitating or recovering functionality in people with neurological disability. To realize energy-efficient and real-time capable devices, neuromorphic computing systems are envisaged as the core of next-generation systems for brain repair. We demonstrate here a real-time hardware neuromorphic prosthesis to restore bidirectional interactions between two neuronal populations, even when one is damaged or missing. We used in vitro modular cell cultures to mimic the mutual interaction between neuronal assemblies and created a focal lesion to functionally disconnect the two populations. Then, we employed our neuromorphic prosthesis for bidirectional bridging to artificially reconnect two disconnected neuronal modules and for hybrid bidirectional bridging to replace the activity of one module with a real-time hardware neuromorphic Spiking Neural Network. Our neuroprosthetic system opens avenues for the exploitation of neuromorphic-based devices in bioelectrical therapeutics for health care.

Detection of reaching intention using EEG signals and nonlinear dynamic system identification

BACKGROUND AND OBJECTIVES

Low frequency electroencephalography (EEG) signals are associated with preparation of movement and thus provide valuable information for brain-machine interface applications. The purpose of this study was to detect movement intention from EEG signals before execution of self-paced arm reaching movements.

METHODS

Ten healthy individuals were recruited. Movement onset was determined from surface electromyography recordings time-locked with EEG signals. Unlike previous studies, which employed feature extraction and classification for decoding, a nonlinear dynamic multiple-input/single output (MISO) model was developed. The MISO model consisted of a cascade of Volterra structures and a threshold block, generating the binary output corresponding to intention/no-intention. The modeling process included input selection from a pool of different EEG channels. The predictive performance of the model was evaluated using the receiver operating characteristics curve, from which the optimum threshold was also selected. The Mann-Whitney statistics was employed to select the significant EEG channels for the output by examining the statistical significance of improvement in the predictive capability of the model when the respective channels were included.

RESULTS

With the proposed approach, movement intention was detected approximately 500 ms before the movement onset and on average, with an accuracy of 96.37 ± 0.94%, a sensitivity of 77.93 ± 4.40% and a specificity of 98.52 ± 1.19%.

CONCLUSIONS

The model output can be converted to motion commands for neuroprosthetic devices and exoskeletons in future applications.

Multi-Input, Multi-Output Neuronal Mode Network Approach to Modeling the Encoding Dynamics and Functional Connectivity of Neural Systems

This letter proposes a novel method, multi-input, multi-output neuronal mode network (MIMO-NMN), for modeling encoding dynamics and functional connectivity in neural ensembles such as the hippocampus. Compared with conventional approaches such as the Volterra-Wiener model, linear-nonlinear-cascade (LNC) model, and generalized linear model (GLM), the NMN has several advantages in terms of estimation accuracy, model interpretation, and functional connectivity analysis. We point out the limitations of current neural spike modeling methods, especially the estimation biases caused by the imbalanced class problem when the number of zeros is significantly larger than ones in the spike data. We use synthetic data to test the performance of NMN with a comparison of the traditional methods, and the results indicate the NMN approach could reduce the imbalanced class problem and achieve better predictions. Subsequently, we apply the MIMO-NMN method to analyze data from the human hippocampus. The results indicate that the MIMO-NMN method is a promising approach to modeling neural dynamics and analyzing functional connectivity of multi-neuronal data.

A Biohybrid Setup for Coupling Biological and Neuromorphic Neural Networks

Developing technologies for coupling neural activity and artificial neural components, is key for advancing neural interfaces and neuroprosthetics. We present a biohybrid experimental setting, where the activity of a biological neural network is coupled to a biomimetic hardware network. The implementation of the hardware network (denoted NeuroSoC) exhibits complex dynamics with a multiplicity of time-scales, emulating 2880 neurons and 12.7 M synapses, designed on a VLSI chip. This network is coupled to a neural network in vitro, where the activities of both the biological and the hardware networks can be recorded, processed, and integrated bidirectionally in real-time. This experimental setup enables an adjustable and well-monitored coupling, while providing access to key functional features of neural networks. We demonstrate the feasibility to functionally couple the two networks and to implement control circuits to modify the biohybrid activity. Overall, we provide an experimental model for neuromorphic-neural interfaces, hopefully to advance the capability to interface with neural activity, and with its irregularities in pathology.

Analysis of Light Propagation on Physiological Properties of Neurons for Nanoscale Optogenetics

Miniaturization of implantable devices is an important challenge for future brain-computer interface applications, and in particular for achieving precise neuron stimulation. For stimulation that utilizes light, i.e., optogenetics, the light propagation behavior and interaction at the nanoscale with elements within the neuron is an important factor that needs to be considered when designing the device. This paper analyzes the effect of light behavior for a single neuron stimulation and focuses on the impact from different cell shapes. Based on the Mie scattering theory, the paper analyzes how the shape of the soma and the nucleus contributes to the focusing effect resulting in an intensity increase, which ensures that neurons can assist in transferring light through the tissue toward the target cells. At the same time, this intensity increase can in turn also stimulate neighboring cells leading to interference within the neural circuits. This paper also analyzes the ideal placements of the device with respect to the angle and position within the cortex that can enable axonal biophoton communications, which can contain light within the cell to avoid the interference.

Unsupervised and real-time spike sorting chip for neural signal processing in hippocampal prosthesis

BACKGROUND

Damage to the hippocampus will result in the loss of ability to form new long-term memories and cognitive disorders. At present, there is no effective medical treatment for this issue. Hippocampal cognitive prosthesis is proposed to replace damaged regions of the hippocampus to mimic the function of original biological tissue. This prosthesis requires a spike sorter to detect and classify spikes in the recorded neural signal.

NEW METHOD

A 16-channel spike sorting processor is presented in this paper, where all channels are considered as independent. An automatic threshold estimation method suitable for hardware implementation is proposed for the Osort clustering algorithm. A new distance metric is also introduced to facilitate clustering. Bayes optimal template matching classification algorithm is optimized to reduce computational complexity by introducing a preselection mechanism.

RESULTS

The chip was fabricated in 40-nm CMOS process with a core area of 0.0175 mm²/ch and power consumption of 19.0 μW/ch. Synthetic and realistic test data are used to evaluate the chip. The test result shows that it has high performance on both data.

COMPARISON WITH EXISTING METHOD(S)

Compared with the other three spike sorting processors, the proposed chip achieves the highest detection and classification accuracy. It also has the ability to deal with partially overlapping spikes, which is not reported in the other work.

CONCLUSIONS

We have developed a 16-channel spike sorting chip used in hippocampal prosthesis, which provides unsupervised clustering and real-time detection and classification. It also has the ability to deal with partially overlapping spikes.

Nonlinear Modeling of Neural Interaction for Spike Prediction Using the Staged Point-Process Model

Neurons communicate nonlinearly through spike activities. Generalized linear models (GLMs) describe spike activities with a cascade of a linear combination across inputs, a static nonlinear function, and an inhomogeneous Bernoulli or Poisson process, or Cox process if a self-history term is considered. This structure considers the output nonlinearity in spike generation but excludes the nonlinear interaction among input neurons. Recent studies extend GLMs by modeling the interaction among input neurons with a quadratic function, which considers the interaction between every pair of input spikes. However, quadratic effects may not fully capture the nonlinear nature of input interaction. We therefore propose a staged point-process model to describe the nonlinear interaction among inputs using a few hidden units, which follows the idea of artificial neural networks. The output firing probability conditioned on inputs is formed as a cascade of two linear-nonlinear (a linear combination plus a static nonlinear function) stages and an inhomogeneous Bernoulli process. Parameters of this model are estimated by maximizing the log likelihood on output spike trains. Unlike the iterative reweighted least squares algorithm used in GLMs, where the performance is guaranteed by the concave condition, we propose a modified Levenberg-Marquardt (L-M) algorithm, which directly calculates the Hessian matrix of the log likelihood, for the nonlinear optimization in our model. The proposed model is tested on both synthetic data and real spike train data recorded from the dorsal premotor cortex and primary motor cortex of a monkey performing a center-out task. Performances are evaluated by discrete-time rescaled Kolmogorov-Smirnov tests, where our model statistically outperforms a GLM and its quadratic extension, with a higher goodness-of-fit in the prediction results. In addition, the staged point-process model describes nonlinear interaction among input neurons with fewer parameters than quadratic models, and the modified L-M algorithm also demonstrates fast convergence.