Nonlinear modeling of causal interrelationships in neuronal ensembles

Recording and Decoding of Vagal Neural Signals Related to Changes in Physiological Parameters and Biomarkers of Disease

Our bodies have built-in neural reflexes that continuously monitor organ function and maintain physiological homeostasis. Whereas the field of bioelectronic medicine has mainly focused on the stimulation of neural circuits to treat various conditions, recent studies have started to investigate the possibility of leveraging the sensory arm of these reflexes to diagnose disease states. To accomplish this, neural signals emanating from the body's built-in biosensors and propagating through peripheral nerves must be recorded and decoded to identify the presence or levels of relevant biomarkers of disease. The process of acquiring these signals poses several technical challenges related to the neural interfaces, surgical techniques, and data-processing framework needed to record and analyze them. However, these challenges can be addressed with a rigorous experimental approach and new advances in implantable electrodes, signal processing, and machine learning methods. Outlined in this review are studies decoding vagus nerve activity as it related to inflammatory, metabolic, and cardiopulmonary biomarkers. Successfully decoding peripheral nerve activity related to disease states will not only enable the development of real-time diagnostic devices, but also help advancing truly closed-loop neuromodulation technologies.

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.

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.

Designing Patient-Specific Optimal Neurostimulation Patterns for Seizure Suppression

Neurostimulation is a promising therapy for abating epileptic seizures. However, it is extremely difficult to identify optimal stimulation patterns experimentally. In this study, human recordings are used to develop a functional 24 neuron network statistical model of hippocampal connectivity and dynamics. Spontaneous seizure-like activity is induced in silico in this reconstructed neuronal network. The network is then used as a testbed to design and validate a wide range of neurostimulation patterns. Commonly used periodic trains were not able to permanently abate seizures at any frequency. A simulated annealing global optimization algorithm was then used to identify an optimal stimulation pattern, which successfully abated 92% of seizures. Finally, in a fully responsive, or closed-loop, neurostimulation paradigm, the optimal stimulation successfully prevented the network from entering the seizure state. We propose that the framework presented here for algorithmically identifying patient-specific neurostimulation patterns can greatly increase the efficacy of neurostimulation devices for seizures.

Mechanism-Based and Input-Output Modeling of the Key Neuronal Connections and Signal Transformations in the CA3-CA1 Regions of the Hippocampus

This letter examines the results of input-output (nonparametric) modeling based on the analysis of data generated by a mechanism-based (parametric) model of CA3-CA1 neuronal connections in the hippocampus. The motivation is to obtain biological insight into the interpretation of such input-output (Volterra-equivalent) models estimated from synthetic data. The insights obtained may be subsequently used to interpretat input-output models extracted from actual experimental data. Specifically, we found that a simplified parametric model may serve as a useful tool to study the signal transformations in the hippocampal CA3-CA1 regions. Input-output modeling of model-based synthetic data show that GABAergic interneurons are responsible for regulating neuronal excitation, controlling the precision of spike timing, and maintaining network oscillations, in a manner consistent with previous studies. The input-output model obtained from real data exhibits intriguing similarities with its synthetic-data counterpart, demonstrating the importance of a dynamic resonance in the system/model response around 2 Hz to 3 Hz. Using the input-output model from real data as a guide, we may be able to amend the parametric model by incorporating more mechanisms in order to yield better-matching input-output model. The approach we present can also be applied to the study of other neural systems and pathways.

A memory efficient implementation scheme of Gauss error function in a Laguerre-Volterra network for neuroprosthetic devices

Cognitive neural prosthesis is a manmade device which can be used to restore or compensate for lost human cognitive modalities. The generalized Laguerre-Volterra (GLV) network serves as a robust mathematical underpinning for the development of such prosthetic instrument. In this paper, a hardware implementation scheme of Gauss error function for the GLV network targeting reconfigurable platforms is reported. Numerical approximations are formulated which transform the computation of nonelementary function into combinational operations of elementary functions, and memory-intensive look-up table (LUT) based approaches can therefore be circumvented. The computational precision can be made adjustable with the utilization of an error compensation scheme, which is proposed based on the experimental observation of the mathematical characteristics of the error trajectory. The precision can be further customizable by exploiting the run-time characteristics of the reconfigurable system. Compared to the polynomial expansion based implementation scheme, the utilization of slice LUTs, occupied slices, and DSP48E1s on a Xilinx XC6VLX240T field-programmable gate array has decreased by 94.2%, 94.1%, and 90.0%, respectively. While compared to the look-up table based scheme, 1.0×10¹⁷ bits of storage can be spared under the maximum allowable error of 1.0×10^-3. The proposed implementation scheme can be employed in the study of large-scale neural ensemble activity and in the design and development of neural prosthetic device.

Computational Modeling of Neurotransmitter Release Evoked by Electrical Stimulation: Nonlinear Approaches to Predicting Stimulation-Evoked Dopamine Release

Neurochemical changes evoked by electrical stimulation of the nervous system have been linked to both therapeutic and undesired effects of neuromodulation therapies used to treat obsessive-compulsive disorder, depression, epilepsy, Parkinson's disease, stroke, hypertension, tinnitus, and many other indications. In fact, interest in better understanding the role of neurochemical signaling in neuromodulation therapies has been a focus of recent government- and industry-sponsored programs whose ultimate goal is to usher in an era of personalized medicine by creating neuromodulation therapies that respond to real-time changes in patient status. A key element to achieving these precision therapeutic interventions is the development of mathematical modeling approaches capable of describing the nonlinear transfer function between neuromodulation parameters and evoked neurochemical changes. Here, we propose two computational modeling frameworks, based on artificial neural networks (ANNs) and Volterra kernels, that can characterize the input/output transfer functions of stimulation-evoked neurochemical release. We evaluate the ability of these modeling frameworks to characterize subject-specific neurochemical kinetics by accurately describing stimulation-evoked dopamine release across rodent (R² = 0.83 Volterra kernel, R² = 0.86 ANN), swine (R² = 0.90 Volterra kernel, R² = 0.93 ANN), and non-human primate (R² = 0.98 Volterra kernel, R² = 0.96 ANN) models of brain stimulation. Ultimately, these models will not only improve understanding of neurochemical signaling in healthy and diseased brains but also facilitate the development of neuromodulation strategies capable of controlling neurochemical release via closed-loop strategies.

Detection of Impaired Sympathetic Cerebrovascular Control Using Functional Biomarkers Based on Principal Dynamic Mode Analysis

This study sought to determine whether models of cerebrovascular function based on Laguerre-Volterra kernels that account for nonlinear cerebral blood flow (CBF) dynamics can detect the effects of functional cerebral sympathetic blockade. We retrospectively analyzed continuous beat-to-beat blood pressure, middle cerebral blood velocity, and partial-pressure of end-tidal CO₂ (P_ETCO₂) recordings from eighteen healthy individuals who were treated with either an oral dose of the α₁-adrenergic receptor blocker Prazosin or a placebo treatment. The global principal dynamic modes (PDMs) were analyzed using Laguerre-Volterra kernels to examine the nonlinear system dynamics. Our principal findings were: (1) very low frequency (<0.03 Hz) linear components of first-order kernels for BP and P_ETCO₂ are mutually coupled to CBF dynamics with the ability to separate individuals between control and blockade conditions, and (2) the gains of the nonlinear functions associated with low-pass and ≈0.03 Hz global PDMs for the BP are sensitive to sympathetic blockade. Collectively these results suggest that very low frequency global PDMs for BP may have potential utility as functional biomarkers of sympathetic neurovascular dysfunction which can occur in conditions like autonomic failure, stroke and traumatic brain injury.

Mapping Horizontal Spread of Activity in Monkey Motor Cortex Using Single Pulse Microstimulation

Anatomical studies have demonstrated that distant cortical points are interconnected through long range axon collaterals of pyramidal cells. However, the functional properties of these intrinsic synaptic connections, especially their relationship with the cortical representations of body movements, have not been systematically investigated. To address this issue, we used multielectrode arrays chronically implanted in the motor cortex of two rhesus monkeys to analyze the effects of single-pulse intracortical microstimulation (sICMS) applied at one electrode on the neuronal activities recorded at all other electrodes. The temporal and spatial distribution of the evoked responses of single and multiunit activities was quantified to determine the properties of horizontal propagation. The typical responses were characterized by a brief excitatory peak followed by inhibition of longer duration. Significant excitatory responses to sICMS could be evoked up to 4 mm away from the stimulation site, but the strength of the response decreased exponentially and its latency increased linearly with the distance. We then quantified the direction and strength of the propagation in relation to the somatotopic organization of the motor cortex. We observed that following sICMS the propagation of neural activity is mainly directed rostro-caudally near the central sulcus but follows medio-lateral direction at the most anterior electrodes. The fact that these interactions are not entirely symmetrical may characterize a critical functional property of the motor cortex for the control of upper limb movements. Overall, these results support the assumption that the motor cortex is not functionally homogeneous but forms a complex network of interacting subregions.

A cognitive prosthesis for memory facilitation by closed-loop functional ensemble stimulation of hippocampal neurons in primate brain

Very productive collaborative investigations characterized how multineuron hippocampal ensembles recorded in nonhuman primates (NHPs) encode short-term memory necessary for successful performance in a delayed match to sample (DMS) task and utilized that information to devise a unique nonlinear multi-input multi-output (MIMO) memory prosthesis device to enhance short-term memory in real-time during task performance. Investigations have characterized how the hippocampus in primate brain encodes information in a multi-item, rule-controlled, delayed match to sample (DMS) task. The MIMO model was applied via closed loop feedback micro-current stimulation during the task via conformal electrode arrays and enhanced performance of the complex memory requirements. These findings clearly indicate detection of a means by which the hippocampus encodes information and transmits this information to other brain regions involved in memory processing. By employing the nonlinear dynamic multi-input/multi-output (MIMO) model, developed and adapted to hippocampal neural ensemble firing patterns derived from simultaneous recorded multi-neuron CA1 and CA3 activity, it was possible to extract information encoded in the Sample phase of DMS trials that was necessary for successful performance in the subsequent Match phase of the task. The extension of this MIMO model to online delivery of electrical stimulation patterns to the same recording loci that exhibited successful CA1 firing in the DMS Sample Phase provided the means to increase task performance on a trial-by-trial basis. Increased utility of the MIMO model as a memory prosthesis was exhibited by the demonstration of cumulative increases in DMS task performance with repeated MIMO stimulation over many sessions. These results, reported below in this article, provide the necessary demonstrations to further the feasibility of the MIMO model as a memory prosthesis to recover and/or enhance encoding of cognitive information in humans with memory disruptions resulting from brain injury, disease or aging.

Understanding spike-triggered covariance using Wiener theory for receptive field identification

Receptive field identification is a vital problem in sensory neurophysiology and vision. Much research has been done in identifying the receptive fields of nonlinear neurons whose firing rate is determined by the nonlinear interactions of a small number of linear filters. Despite more advanced methods that have been proposed, spike-triggered covariance (STC) continues to be the most widely used method in such situations due to its simplicity and intuitiveness. Although the connection between STC and Wiener/Volterra kernels has often been mentioned in the literature, this relationship has never been explicitly derived. Here we derive this relationship and show that the STC matrix is actually a modified version of the second-order Wiener kernel, which incorporates the input autocorrelation and mixes first- and second-order dynamics. It is then shown how, with little modification of the STC method, the Wiener kernels may be obtained and, from them, the principal dynamic modes, a set of compact and efficient linear filters that essentially combine the spike-triggered average and STC matrix and generalize to systems with both continuous and point-process outputs. Finally, using Wiener theory, we show how these obtained filters may be corrected when they were estimated using correlated inputs. Our correction technique is shown to be superior to those commonly used in the literature for both correlated Gaussian images and natural images.

Hippocampal closed-loop modeling and implications for seizure stimulation design

OBJECTIVE

Traditional hippocampal modeling has focused on the series of feedforward synapses known as the trisynaptic pathway. However, feedback connections from CA1 back to the hippocampus through the entorhinal cortex (EC) actually make the hippocampus a closed-loop system. By constructing a functional closed-loop model of the hippocampus, one may learn how both physiological and epileptic oscillations emerge and design efficient neurostimulation patterns to abate such oscillations.

APPROACH

Point process input-output models where estimated from recorded rodent hippocampal data to describe the nonlinear dynamical transformation from CA3 → CA1, via the schaffer-collateral synapse, and CA1 → CA3 via the EC. Each Volterra-like subsystem was composed of linear dynamics (principal dynamic modes) followed by static nonlinearities. The two subsystems were then wired together to produce the full closed-loop model of the hippocampus.

MAIN RESULTS

Closed-loop connectivity was found to be necessary for the emergence of theta resonances as seen in recorded data, thus validating the model. The model was then used to identify frequency parameters for the design of neurostimulation patterns to abate seizures.

SIGNIFICANCE

Deep-brain stimulation (DBS) is a new and promising therapy for intractable seizures. Currently, there is no efficient way to determine optimal frequency parameters for DBS, or even whether periodic or broadband stimuli are optimal. Data-based computational models have the potential to be used as a testbed for designing optimal DBS patterns for individual patients. However, in order for these models to be successful they must incorporate the complex closed-loop structure of the seizure focus. This study serves as a proof-of-concept of using such models to design efficient personalized DBS patterns for epilepsy.

System identification of point-process neural systems using probability based Volterra kernels

BACKGROUND

Neural information processing involves a series of nonlinear dynamical input/output transformations between the spike trains of neurons/neuronal ensembles. Understanding and quantifying these transformations is critical both for understanding neural physiology such as short-term potentiation and for developing cognitive neural prosthetics.

NEW METHOD

A novel method for estimating Volterra kernels for systems with point-process inputs and outputs is developed based on elementary probability theory. These Probability Based Volterra (PBV) kernels essentially describe the probability of an output spike given q input spikes at various lags t1, t2, …, tq.

RESULTS

The PBV kernels are used to estimate both synthetic systems where ground truth is available and data from the CA3 and CA1 regions rodent hippocampus. The PBV kernels give excellent predictive results in both cases. Furthermore, they are shown to be quite robust to noise and to have good convergence and overfitting properties. Through a slight modification, the PBV kernels are shown to also deal well with correlated point-process inputs.

COMPARISON WITH EXISTING METHODS

The PBV kernels were compared with kernels estimated through least squares estimation (LSE) and through the Laguerre expansion technique (LET). The LSE kernels were shown to fair very poorly with real data due to the large amount of input noise. Although the LET kernels gave the best predictive results in all cases, they require prior parameter estimation. It was shown how the PBV and LET methods can be combined synergistically to maximize performance.

CONCLUSIONS

The PBV kernels provide a novel and intuitive method of characterizing point-process input-output nonlinear systems.

Principal dynamic mode analysis of the Hodgkin-Huxley equations

We develop an autoregressive model framework based on the concept of Principal Dynamic Modes (PDMs) for the process of action potential (AP) generation in the excitable neuronal membrane described by the Hodgkin-Huxley (H-H) equations. The model's exogenous input is injected current, and whenever the membrane potential output exceeds a specified threshold, it is fed back as a second input. The PDMs are estimated from the previously developed Nonlinear Autoregressive Volterra (NARV) model, and represent an efficient functional basis for Volterra kernel expansion. The PDM-based model admits a modular representation, consisting of the forward and feedback PDM bases as linear filterbanks for the exogenous and autoregressive inputs, respectively, whose outputs are then fed to a static nonlinearity composed of polynomials operating on the PDM outputs and cross-terms of pair-products of PDM outputs. A two-step procedure for model reduction is performed: first, influential subsets of the forward and feedback PDM bases are identified and selected as the reduced PDM bases. Second, the terms of the static nonlinearity are pruned. The first step reduces model complexity from a total of 65 coefficients to 27, while the second further reduces the model coefficients to only eight. It is demonstrated that the performance cost of model reduction in terms of out-of-sample prediction accuracy is minimal. Unlike the full model, the eight coefficient pruned model can be easily visualized to reveal the essential system components, and thus the data-derived PDM model can yield insight into the underlying system structure and function.

Model-based asessment of an in-vivo predictive relationship from CA1 to CA3 in the rodent hippocampus

Although an anatomical connection from CA1 to CA3 via the Entorhinal Cortex (EC) and through backprojecting interneurons has long been known it exist, it has never been examined quantitatively on the single neuron level, in the in-vivo nonpatholgical, nonperturbed brain. Here, single spike activity was recorded using a multi-electrode array from the CA3 and CA1 areas of the rodent hippocampus (N = 7) during a behavioral task. The predictive power from CA3→CA1 and CA1→CA3 was examined by constructing Multivariate Autoregressive (MVAR) models from recorded neurons in both directions. All nonsignificant inputs and models were identified and removed by means of Monte Carlo simulation methods. It was found that 121/166 (73 %) CA3→CA1 models and 96/145 (66 %) CA1→CA3 models had significant predictive power, thus confirming a predictive 'Granger' causal relationship from CA1 to CA3. This relationship is thought to be caused by a combination of truly causal connections such as the CA1→EC→CA3 pathway and common inputs such as those from the Septum. All MVAR models were then examined in the frequency domain and it was found that CA3 kernels had significantly more power in the theta and beta range than those of CA1, confirming CA3's role as an endogenous hippocampal pacemaker.

Donor/recipient enhancement of memory in rat hippocampus

The critical role of the mammalian hippocampus in the formation, translation and retrieval of memory has been documented over many decades. There are many theories of how the hippocampus operates to encode events and a precise mechanism was recently identified in rats performing a short-term memory task which demonstrated that successful information encoding was promoted via specific patterns of activity generated within ensembles of hippocampal neurons. In the study presented here, these “representations” were extracted via a customized non-linear multi-input multi-output (MIMO) mathematical model which allowed prediction of successful performance on specific trials within the testing session. A unique feature of this characterization was demonstrated when successful information encoding patterns were derived online from well-trained “donor” animals during difficult long-delay trials and delivered via online electrical stimulation to synchronously tested naïve “recipient” animals never before exposed to the delay feature of the task. By transferring such model-derived trained (donor) animal hippocampal firing patterns via stimulation to coupled naïve recipient animals, their task performance was facilitated in a direct “donor-recipient” manner. This provides the basis for utilizing extracted appropriate neural information from one brain to induce, recover, or enhance memory related processing in the brain of another subject.

On parsing the neural code in the prefrontal cortex of primates using principal dynamic modes

Nonlinear modeling of multi-input multi-output (MIMO) neuronal systems using Principal Dynamic Modes (PDMs) provides a novel method for analyzing the functional connectivity between neuronal groups. This paper presents the PDM-based modeling methodology and initial results from actual multi-unit recordings in the prefrontal cortex of non-human primates. We used the PDMs to analyze the dynamic transformations of spike train activity from Layer 2 (input) to Layer 5 (output) of the prefrontal cortex in primates performing a Delayed-Match-to-Sample task. The PDM-based models reduce the complexity of representing large-scale neural MIMO systems that involve large numbers of neurons, and also offer the prospect of improved biological/physiological interpretation of the obtained models. PDM analysis of neuronal connectivity in this system revealed "input-output channels of communication" corresponding to specific bands of neural rhythms that quantify the relative importance of these frequency-specific PDMs across a variety of different tasks. We found that behavioral performance during the Delayed-Match-to-Sample task (correct vs. incorrect outcome) was associated with differential activation of frequency-specific PDMs in the prefrontal cortex.

Identification of sparse neural functional connectivity using penalized likelihood estimation and basis functions

One key problem in computational neuroscience and neural engineering is the identification and modeling of functional connectivity in the brain using spike train data. To reduce model complexity, alleviate overfitting, and thus facilitate model interpretation, sparse representation and estimation of functional connectivity is needed. Sparsities include global sparsity, which captures the sparse connectivities between neurons, and local sparsity, which reflects the active temporal ranges of the input-output dynamical interactions. In this paper, we formulate a generalized functional additive model (GFAM) and develop the associated penalized likelihood estimation methods for such a modeling problem. A GFAM consists of a set of basis functions convolving the input signals, and a link function generating the firing probability of the output neuron from the summation of the convolutions weighted by the sought model coefficients. Model sparsities are achieved by using various penalized likelihood estimations and basis functions. Specifically, we introduce two variations of the GFAM using a global basis (e.g., Laguerre basis) and group LASSO estimation, and a local basis (e.g., B-spline basis) and group bridge estimation, respectively. We further develop an optimization method based on quadratic approximation of the likelihood function for the estimation of these models. Simulation and experimental results show that both group-LASSO-Laguerre and group-bridge-B-spline can capture faithfully the global sparsities, while the latter can replicate accurately and simultaneously both global and local sparsities. The sparse models outperform the full models estimated with the standard maximum likelihood method in out-of-sample predictions.

Design of optimal stimulation patterns for neuronal ensembles based on Volterra-type hierarchical modeling

This paper presents a general methodology for the optimal design of stimulation patterns applied to neuronal ensembles in order to elicit a desired effect. The methodology follows a variant of the hierarchical Volterra modeling approach that utilizes input-output data to construct predictive models that describe the effects of interactions among multiple input events in an ascending order of interaction complexity. The illustrative example presented in this paper concerns the multi-unit activity of CA1 neurons in the hippocampus of a rodent performing a learned delayed-nonmatch-to-sample (DNMS) task. The multi-unit activity of the hippocampal CA1 neurons is recorded via chronically implanted multi-electrode arrays during this task. The obtained model quantifies the likelihood of having correct performance of the specific task for a given multi-unit (spatiotemporal) activity pattern of a CA1 neuronal ensemble during the 'sample presentation' phase of the DNMS task. The model can be used to determine computationally (off-line) the 'optimal' multi-unit stimulation pattern that maximizes the likelihood of inducing the correct performance of the DNMS task. Our working hypothesis is that application of this optimal stimulation pattern will enhance performance of the DNMS task due to enhancement of memory formation and storage during the 'sample presentation' phase of the task.

A nonlinear autoregressive Volterra model of the Hodgkin-Huxley equations

We propose a new variant of Volterra-type model with a nonlinear auto-regressive (NAR) component that is a suitable framework for describing the process of AP generation by the neuron membrane potential, and we apply it to input-output data generated by the Hodgkin-Huxley (H-H) equations. Volterra models use a functional series expansion to describe the input-output relation for most nonlinear dynamic systems, and are applicable to a wide range of physiologic systems. It is difficult, however, to apply the Volterra methodology to the H-H model because is characterized by distinct subthreshold and suprathreshold dynamics. When threshold is crossed, an autonomous action potential (AP) is generated, the output becomes temporarily decoupled from the input, and the standard Volterra model fails. Therefore, in our framework, whenever membrane potential exceeds some threshold, it is taken as a second input to a dual-input Volterra model. This model correctly predicts membrane voltage deflection both within the subthreshold region and during APs. Moreover, the model naturally generates a post-AP afterpotential and refractory period. It is known that the H-H model converges to a limit cycle in response to a constant current injection. This behavior is correctly predicted by the proposed model, while the standard Volterra model is incapable of generating such limit cycle behavior. The inclusion of cross-kernels, which describe the nonlinear interactions between the exogenous and autoregressive inputs, is found to be absolutely necessary. The proposed model is general, non-parametric, and data-derived.

Nonlinear modeling of dynamic interactions within neuronal ensembles using Principal Dynamic Modes

A methodology for nonlinear modeling of multi-input multi-output (MIMO) neuronal systems is presented that utilizes the concept of Principal Dynamic Modes (PDM). The efficacy of this new methodology is demonstrated in the study of the dynamic interactions between neuronal ensembles in the Pre-Frontal Cortex (PFC) of a behaving non-human primate (NHP) performing a Delayed Match-to-Sample task. Recorded spike trains from Layer-2 and Layer-5 neurons were viewed as the "inputs" and "outputs", respectively, of a putative MIMO system/model that quantifies the dynamic transformation of multi-unit neuronal activity between Layer-2 and Layer-5 of the PFC. Model prediction performance was evaluated by means of computed Receiver Operating Characteristic (ROC) curves. The PDM-based approach seeks to reduce the complexity of MIMO models of neuronal ensembles in order to enable the practicable modeling of large-scale neural systems incorporating hundreds or thousands of neurons, which is emerging as a preeminent issue in the study of neural function. The "scaling-up" issue has attained critical importance as multi-electrode recordings are increasingly used to probe neural systems and advance our understanding of integrated neural function. The initial results indicate that the PDM-based modeling methodology may greatly reduce the complexity of the MIMO model without significant degradation of performance. Furthermore, the PDM-based approach offers the prospect of improved biological/physiological interpretation of the obtained MIMO models.

Functional connectivity during surround suppression in macaque area V4

Surround suppression is a common feature of sensory neurons. For neurons of the visual cortex, it occurs when a visual stimulus extends beyond a neuron's classical receptive field, reducing the neuron's firing rate. While several studies have been attributing the suppression effect on horizontal, long-range lateral or feedback connections, the underlying circuitry for surround modulation remain unidentified. Since most of these models have been relying on single neuron recordings, the contribution of lateral connections can only be suggested from the surround field properties. A more straightforward approach would be to detect these connections and their dynamics using simultaneous recordings from multiple neurons in one or more visual areas. We have developed a method for estimating these connections and we analyzed data obtained from 100-electrode Utah arrays chronically implanted into area V4 of the macaque monkey. Using a method based on the nonlinear Volterra modeling approach, we computed estimates of the strength and statistical reliability of connections among neurons, including nonlinear interactions and excitatory and inhibitory connections. Our results thus far reveal a pattern of connectivity within V4 that conforms to the results of previous anatomical work: Excitatory connections are far more common than inhibitory connections (∼65%), stronger connections are found among neurons that are physically near one another, and connections are stronger among neurons with similar receptive field properties. However, this connectivity is capable of reorganizing on short time scales according to the stimulus: Stimuli that evoke strong suppression at the single-unit level introduce stronger inhibition among V4 neurons, identifying recurrent connectivity as the source of the suppression. Overall, these results provide insight into the dynamic nature of neuronal organization within V4 and its contribution to surround suppression.

Closing the loop for memory prosthesis: detecting the role of hippocampal neural ensembles using nonlinear models

A major factor involved in providing closed loop feedback for control of neural function is to understand how neural ensembles encode online information critical to the final behavioral endpoint. This issue was directly assessed in rats performing a short-term delay memory task in which successful encoding of task information is dependent upon specific spatio-temporal firing patterns recorded from ensembles of CA3 and CA1 hippocampal neurons. Such patterns, extracted by a specially designed nonlinear multi-input multi-output (MIMO) nonlinear mathematical model, were used to predict successful performance online via a closed loop paradigm which regulated trial difficulty (time of retention) as a function of the "strength" of stimulus encoding. The significance of the MIMO model as a neural prosthesis has been demonstrated by substituting trains of electrical stimulation pulses to mimic these same ensemble firing patterns. This feature was used repeatedly to vary "normal" encoding as a means of understanding how neural ensembles can be "tuned" to mimic the inherent process of selecting codes of different strength and functional specificity. The capacity to enhance and tune hippocampal encoding via MIMO model detection and insertion of critical ensemble firing patterns shown here provides the basis for possible extension to other disrupted brain circuitry.

A nonlinear model for hippocampal cognitive prosthesis: memory facilitation by hippocampal ensemble stimulation

Collaborative investigations have characterized how multineuron hippocampal ensembles encode memory necessary for subsequent successful performance by rodents in a delayed nonmatch to sample (DNMS) task and utilized that information to provide the basis for a memory prosthesis to enhance performance. By employing a unique nonlinear dynamic multi-input/multi-output (MIMO) model, developed and adapted to hippocampal neural ensemble firing patterns derived from simultaneous recorded CA1 and CA3 activity, it was possible to extract information encoded in the sample phase necessary for successful performance in the nonmatch phase of the task. The extension of this MIMO model to online delivery of electrical stimulation delivered to the same recording loci that mimicked successful CA1 firing patterns, provided the means to increase levels of performance on a trial-by-trial basis. Inclusion of several control procedures provides evidence for the specificity of effective MIMO model generated patterns of electrical stimulation. Increased utility of the MIMO model as a prosthesis device was exhibited by the demonstration of cumulative increases in DNMS task performance with repeated MIMO stimulation over many sessions on both stimulation and nonstimulation trials, suggesting overall system modification with continued exposure. Results reported here are compatible with and extend prior demonstrations and further support the candidacy of the MIMO model as an effective cortical prosthesis.

Restorative encoding memory integrative neural device: "REMIND"

Construction and application of a neural prosthesis device that enhances existing and replaces lost memory capacity in humans is the focus of research described here in rodents. A unique approach for the analysis and application of neural population firing has been developed to decipher the pattern in which information is successfully encoded by the hippocampus where mnemonic accuracy is critical. A nonlinear dynamic multi-input multi-output (MIMO) model is utilized to extract memory relevant firing patterns in CA3 and CA1 and to predict online what the consequences of the encoded firing patterns reflect for subsequent information retrieval for successful performance of delayed-nonmatch-to-sample (DNMS) memory task in rodents. The MIMO model has been tested successfully in a number of different contexts, each of which produced improved performance by a) utilizing online predicted codes to regulate task difficulty, b) employing electrical stimulation of CA1 output areas in the same pattern as successful cell firing, c) employing electrical stimulation to recover cell firing compromised by pharmacological agents and d) transferring and improving performance in naïve animals using the same stimulation patterns that are effective in fully trained animals. The results in rodents formed the basis for extension of the MIMO model to nonhuman primates in the same type of memory task that is now being tested in the last step prior to its application in humans.

Relationships between spike-free local field potentials and spike timing in human temporal cortex

Intracortical recordings comprise both fast events, action potentials (APs), and slower events, known as local field potentials (LFPs). Although it is believed that LFPs mostly reflect local synaptic activity, it is unclear which of their signal components are most closely related to synaptic potentials and would therefore be causally related to the occurrence of individual APs. This issue is complicated by the significant contribution from AP waveforms, especially at higher LFP frequencies. In recordings of single-cell activity and LFPs from the human temporal cortex, we computed quantitative, nonlinear, causal dynamic models for the prediction of AP timing from LFPs, at millisecond resolution, before and after removing AP contributions to the LFP. In many cases, the timing of a significant number of single APs could be predicted from spike-free LFPs at different frequencies. Not surprisingly, model performance was superior when spikes were not removed. Cells whose activity was predicted by the spike-free LFP models generally fell into one of two groups: in the first group, neuronal spike activity was associated with specific phases of low LFP frequencies, lower spike activity at high LFP frequencies, and a stronger linear component in the spike-LFP model; in the second group, neuronal spike activity was associated with larger amplitude of high LFP frequencies, less frequent phase locking, and a stronger nonlinear model component. Spike timing in the first group was better predicted by the sign and level of the LFP preceding the spike, whereas spike timing in the second group was better predicted by LFP power during a certain time window before the spike.

A cortical neural prosthesis for restoring and enhancing memory

A primary objective in developing a neural prosthesis is to replace neural circuitry in the brain that no longer functions appropriately. Such a goal requires artificial reconstruction of neuron-to-neuron connections in a way that can be recognized by the remaining normal circuitry, and that promotes appropriate interaction. In this study, the application of a specially designed neural prosthesis using a multi-input/multi-output (MIMO) nonlinear model is demonstrated by using trains of electrical stimulation pulses to substitute for MIMO model derived ensemble firing patterns. Ensembles of CA3 and CA1 hippocampal neurons, recorded from rats performing a delayed-nonmatch-to-sample (DNMS) memory task, exhibited successful encoding of trial-specific sample lever information in the form of different spatiotemporal firing patterns. MIMO patterns, identified online and in real-time, were employed within a closed-loop behavioral paradigm. Results showed that the model was able to predict successful performance on the same trial. Also, MIMO model-derived patterns, delivered as electrical stimulation to the same electrodes, improved performance under normal testing conditions and, more importantly, were capable of recovering performance when delivered to animals with ensemble hippocampal activity compromised by pharmacologic blockade of synaptic transmission. These integrated experimental-modeling studies show for the first time that, with sufficient information about the neural coding of memories, a neural prosthesis capable of real-time diagnosis and manipulation of the encoding process can restore and even enhance cognitive, mnemonic processes.

Nonlinear dynamic modeling of synaptically driven single hippocampal neuron intracellular activity

A high-order nonlinear dynamic model of the input-output properties of single hippocampal CA1 pyramidal neurons was developed based on synaptically driven intracellular activity. The purpose of this study is to construct a model that: 1) can capture the nonlinear dynamics of both subthreshold activities [postsynaptic potentials (PSPs)] and suprathreshold activities (action potentials) in a single formalism; 2) is sufficiently general to be applied to any spike-input and spike-output neurons (point process input and point process output neural systems); and 3) is computationally efficient. The model consisted of three major components: 1) feedforward kernels (up to third order) that transform presynaptic action potentials into PSPs; 2) a constant threshold, above which action potentials are generated; and 3) a feedback kernel (first order) that describes spike-triggered after-potentials. The model was applied to CA1 pyramidal cells, as they were electrically stimulated with broadband Poisson random impulse trains through the Schaffer collaterals. The random impulse trains used here have physiological properties similar to spiking patterns observed in CA3 hippocampal neurons. PSPs and action potentials were recorded from the soma of CA1 pyramidal neurons using whole-cell patch-clamp recording. We evaluated the model performance separately with respect to PSP waveforms and the occurrence of spikes. The average normalized mean square error of PSP prediction is 14.4%. The average spike prediction error rate is 18.8%. In summary, although prediction errors still could be reduced, the model successfully captures the majority of high-order nonlinear dynamics of the single-neuron intracellular activity. The model captures the general biophysical processes with a small set of open parameters that are directly constrained by the intracellular recording, and thus, can be easily applied to any spike-input and spike-output neuron.

Dynamic nonlinear modeling of interactions between neuronal ensembles using principal dynamic modes

We present a novel methodology for modeling the interactions between neuronal ensembles that utilizes the concept of Principal Dynamic Modes (PDM) and their associated nonlinear functions (ANF). This new approach seeks to reduce the complexity of the multi-input/multi-output (MIMO) model of the interactions between neuronal ensembles--an issue of critical practical importance in scaling up the MIMO models to incorporate hundreds (or even thousands) of input-output neurons. Global PDMs were extracted from the data using estimated first-order and second-order kernels and singular value decomposition (SVD). These global PDMs represent an efficient "coordinate system" for the representation of the MIMO model. The ANFs of the PDMs are estimated from the histograms of the combinations of PDM output values that lead to output spikes. For initial testing and validation of this approach, we applied it to a set of data collected at the pre-frontal cortex of a non-human primate during a behavioral task (Delayed Match-to-Sample). Recorded spike trains from Layer-2 neurons were viewed as the "inputs" and from Layer-5 neurons as the outputs. Model prediction performance was evaluated by means of computed Receiver Operating Characteristic (ROC) curves. The results indicate that this methodology may greatly reduce the complexity of the MIMO model without significant degradation of performance.

Removal of spurious correlations between spikes and local field potentials

Single neurons carry out important sensory and motor functions related to the larger networks in which they are embedded. Understanding the relationships between single-neuron spiking and network activity is therefore of great importance and the latter can be readily estimated from low-frequency brain signals known as local field potentials (LFPs). In this work we examine a number of issues related to the estimation of spike and LFP signals. We show that spike trains and individual spikes contain power at the frequencies that are typically thought to be exclusively related to LFPs, such that simple frequency-domain filtering cannot be effectively used to separate the two signals. Ground-truth simulations indicate that the commonly used method of estimating the LFP signal by low-pass filtering the raw voltage signal leads to artifactual correlations between spikes and LFPs and that these correlations exert a powerful influence on popular metrics of spike-LFP synchronization. Similar artifactual results were seen in data obtained from electrophysiological recordings in macaque visual cortex, when low-pass filtering was used to estimate LFP signals. In contrast LFP tuning curves in response to sensory stimuli do not appear to be affected by spike contamination, either in simulations or in real data. To address the issue of spike contamination, we devised a novel Bayesian spike removal algorithm and confirmed its effectiveness in simulations and by applying it to the electrophysiological data. The algorithm, based on a rigorous mathematical framework, outperforms other methods of spike removal on most metrics of spike-LFP correlations. Following application of this spike removal algorithm, many of our electrophysiological recordings continued to exhibit spike-LFP correlations, confirming previous reports that such relationships are a genuine aspect of neuronal activity. Overall, these results show that careful preprocessing is necessary to remove spikes from LFP signals, but that when effective spike removal is used, spike-LFP correlations can potentially yield novel insights about brain function.

The Neurobiological Basis of Cognition: Identification by Multi-Input, Multioutput Nonlinear Dynamic Modeling: A method is proposed for measuring and modeling human long-term memory formation by mathematical analysis and computer simulation of nerve-cell dynamics

The successful development of neural prostheses requires an understanding of the neurobiological bases of cognitive processes, i.e., how the collective activity of populations of neurons results in a higher level process not predictable based on knowledge of the individual neurons and/or synapses alone. We have been studying and applying novel methods for representing nonlinear transformations of multiple spike train inputs (multiple time series of pulse train inputs) produced by synaptic and field interactions among multiple subclasses of neurons arrayed in multiple layers of incompletely connected units. We have been applying our methods to study of the hippocampus, a cortical brain structure that has been demonstrated, in humans and in animals, to perform the cognitive function of encoding new long-term (declarative) memories. Without their hippocampi, animals and humans retain a short-term memory (memory lasting approximately 1 min), and long-term memory for information learned prior to loss of hippocampal function. Results of more than 20 years of studies have demonstrated that both individual hippocampal neurons, and populations of hippocampal cells, e.g., the neurons comprising one of the three principal subsystems of the hippocampus, induce strong, higher order, nonlinear transformations of hippocampal inputs into hippocampal outputs. For one synaptic input or for a population of synchronously active synaptic inputs, such a transformation is represented by a sequence of action potential inputs being changed into a different sequence of action potential outputs. In other words, an incoming temporal pattern is transformed into a different, outgoing temporal pattern. For multiple, asynchronous synaptic inputs, such a transformation is represented by a spatiotemporal pattern of action potential inputs being changed into a different spatiotemporal pattern of action potential outputs. Our primary thesis is that the encoding of short-term memories into new, long-term memories represents the collective set of nonlinearities induced by the three or four principal subsystems of the hippocampus, i.e., entorhinal cortex-to-dentate gyrus, dentate gyrus-to-CA3 pyramidal cell region, CA3-to-CA1 pyramidal cell region, and CA1-to-subicular cortex. This hypothesis will be supported by studies using in vivo hippocampal multineuron recordings from animals performing memory tasks that require hippocampal function. The implications for this hypothesis will be discussed in the context of "cognitive prostheses"-neural prostheses for cortical brain regions believed to support cognitive functions, and that often are subject to damage due to stroke, epilepsy, dementia, and closed head trauma.

Consistent recovery of sensory stimuli encoded with MIMO neural circuits

We consider the problem of reconstructing finite energy stimuli encoded with a population of spiking leaky integrate-and-fire neurons. The reconstructed signal satisfies a consistency condition: when passed through the same neuron, it triggers the same spike train as the original stimulus. The recovered stimulus has to also minimize a quadratic smoothness optimality criterion. We formulate the reconstruction as a spline interpolation problem for scalar as well as vector valued stimuli and show that the recovery has a unique solution. We provide explicit reconstruction algorithms for stimuli encoded with single as well as a population of integrate-and-fire neurons. We demonstrate how our reconstruction algorithms can be applied to stimuli encoded with ON-OFF neural circuits with feedback. Finally, we extend the formalism to multi-input multi-output neural circuits and demonstrate that vector-valued finite energy signals can be efficiently encoded by a neural population provided that its size is beyond a threshold value. Examples are given that demonstrate the potential applications of our methodology to systems neuroscience and neuromorphic engineering.

Boolean modeling of neural systems with point-process inputs and outputs. Part II: Application to the rat hippocampus

This paper presents a pilot application of the Boolean-Volterra modeling methodology presented in the companion paper (Part I) that is suitable for the analysis of systems with point-process inputs and outputs (e.g., recordings of the activity of neuronal ensembles). This application seeks to discover the causal links between two neuronal ensembles in the hippocampus of a behaving rat. The experimental data come from multi-unit recordings in the CA3 and CA1 regions of the hippocampus in the form of sequences of action potentials-treated mathematically as point-processes and computationally as spike-trains-that are collected in vivo during two behavioral tasks. The modeling objective is to identify and quantify the causal links among the neurons generating the recorded activity, using Boolean-Volterra models estimated directly from the data according to the methodological framework presented in the companion paper. The obtained models demonstrate the feasibility of the proposed approach using short data-records and provide some insights into the functional properties of the system (e.g., regarding the presence of rhythmic characteristics in the neuronal dynamics of these ensembles), making the proposed methodology an attractive tool for the analysis and modeling of multi-unit recordings from neuronal systems in a practical context.

A wireless recording system that utilizes Bluetooth technology to transmit neural activity in freely moving animals

A new wireless transceiver is described for recording individual neuron firing from behaving rats utilizing Bluetooth transmission technology and a processor onboard for discrimination of neuronal waveforms and associated time stamps. This universal brain activity transmitter (UBAT) is attached to rodents via a backpack and amplifier headstage and can transmit 16 channels of captured neuronal firing data via a Bluetooth transceiver chip over very large and unconstrained distances. The onboard microprocessor of the UBAT allows flexible online control over waveform isolation criteria via transceiver instruction and the two-way communication capacity allows for closed-loop applications between neural events and behavioral or physiological processes which can be modified by transceiver instructions. A detailed description of the multiplexer processing of channel data as well as examples of neuronal recordings in different behavioral testing contexts is provided to demonstrate the capacity for robust transmission within almost any laboratory environment. A major advantage of the UBAT is the long transmission range and lack of object-based line of sight interference afforded by Bluetooth technology, allowing flexible recording capabilities within multiple experimental paradigms without interruption. Continuous recordings over very large distance separations from the monitor station are demonstrated providing experimenters with recording advantages not previously available with other telemetry devices.

Boolean modeling of neural systems with point-process inputs and outputs. Part I: theory and simulations

This paper presents a new modeling approach for neural systems with point-process (spike) inputs and outputs that utilizes Boolean operators (i.e. modulo 2 multiplication and addition that correspond to the logical AND and OR operations respectively, as well as the AND_NOT logical operation representing inhibitory effects). The form of the employed mathematical models is akin to a "Boolean-Volterra" model that contains the product terms of all relevant input lags in a hierarchical order, where terms of order higher than first represent nonlinear interactions among the various lagged values of each input point-process or among lagged values of various inputs (if multiple inputs exist) as they reflect on the output. The coefficients of this Boolean-Volterra model are also binary variables that indicate the presence or absence of the respective term in each specific model/system. Simulations are used to explore the properties of such models and the feasibility of their accurate estimation from short data-records in the presence of noise (i.e. spurious spikes). The results demonstrate the feasibility of obtaining reliable estimates of such models, with excitatory and inhibitory terms, in the presence of considerable noise (spurious spikes) in the outputs and/or the inputs in a computationally efficient manner. A pilot application of this approach to an actual neural system is presented in the companion paper (Part II).

Nonlinear modeling of neural population dynamics for hippocampal prostheses

Developing a neural prosthesis for the damaged hippocampus requires restoring the transformation of population neural activities performed by the hippocampal circuitry. To bypass a damaged region, output spike trains need to be predicted from the input spike trains and then reinstated through stimulation. We formulate a multiple-input, multiple-output (MIMO) nonlinear dynamic model for the input-output transformation of spike trains. In this approach, a MIMO model comprises a series of physiologically-plausible multiple-input, single-output (MISO) neuron models that consist of five components each: (1) feedforward Volterra kernels transforming the input spike trains into the synaptic potential, (2) a feedback kernel transforming the output spikes into the spike-triggered after-potential, (3) a noise term capturing the system uncertainty, (4) an adder generating the pre-threshold potential, and (5) a threshold function generating output spikes. It is shown that this model is equivalent to a generalized linear model with a probit link function. To reduce model complexity and avoid overfitting, statistical model selection and cross-validation methods are employed to choose the significant inputs and interactions between inputs. The model is applied successfully to the hippocampal CA3-CA1 population dynamics. Such a model can serve as a computational basis for the development of hippocampal prostheses.

Functional connectivity through nonlinear modeling: an application to the rat hippocampus

Implementation of neuroprosthetic devices requires a reliable and accurate quantitative representation of the input-output transformations performed by the involved neuronal populations. Nonparametric, data driven models with predictive capabilities are excellent candidates for these purposes. When modeling input-output relations in multi-input neuronal systems, it is important to select the subset of inputs that are functionally and causally related to the output. Inputs that do not convey information about the actual transformation not only increase the computational burden but also affect the generalization of the model. Moreover, a reliable functional connectivity measure can provide patterns of information flow that can be linked to physiological and anatomical properties of the system. We propose a method based on the Volterra modeling approach that selects distinct subsets of inputs for each output based on the prediction of the respective models and its statistical evaluation. The algorithm builds successive models with increasing number of inputs and examines whether the inclusion of additional inputs benefits the predictive accuracy of the overall model. It also explores possible second-order (inter-modulatory) interactions among the inputs. The method was applied to multi-unit recordings from the CA3 (input) and CA1 (output) regions of the hippocampus in behaving rats, in order to reveal spatiotemporal connectivity maps of the input-output transformation taking place in the CA3-CA1 synapse.