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.

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.

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.

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.

ASIC Implementation of a Nonlinear Dynamical Model for Hippocampal Prosthesis.

A hippocampal prosthesis is a very large scale integration (VLSI) biochip that needs to be implanted in the biological brain to solve a cognitive dysfunction. In this letter, we propose a novel low-complexity, small-area, and low-power programmable hippocampal neural network application-specific integrated circuit (ASIC) for a hippocampal prosthesis. It is based on the nonlinear dynamical model of the hippocampus: namely multi-input, multi-output (MIMO)-generalized Laguerre-Volterra model (GLVM). It can realize the real-time prediction of hippocampal neural activity. New hardware architecture, a storage space configuration scheme, low-power convolution, and gaussian random number generator modules are proposed. The ASIC is fabricated in 40 nm technology with a core area of 0.122 mm[Formula: see text] and test power of 84.4 [Formula: see text]W. Compared with the design based on the traditional architecture, experimental results show that the core area of the chip is reduced by 84.94% and the core power is reduced by 24.30%.

Identification of Stable Spike-Timing-Dependent Plasticity from Spiking Activity with Generalized Multilinear Modeling.

Characterization of long-term activity-dependent plasticity from behaviorally driven spiking activity is important for understanding the underlying mechanisms of learning and memory. In this letter, we present a computational framework for quantifying spike-timing-dependent plasticity (STDP) during behavior by identifying a functional plasticity rule solely from spiking activity. First, we formulate a flexible point-process spiking neuron model structure with STDP, which includes functions that characterize the stationary and plastic properties of the neuron. The STDP model includes a novel function for prolonged plasticity induction, as well as a more typical function for synaptic weight change based on the relative timing of input-output spike pairs. Consideration for system stability is incorporated with weight-dependent synaptic modification. Next, we formalize an estimation technique using a generalized multilinear model (GMLM) structure with basis function expansion. The weight-dependent synaptic modification adds a nonlinearity to the model, which is addressed with an iterative unconstrained optimization approach. Finally, we demonstrate successful model estimation on simulated spiking data and show that all model functions can be estimated accurately with this method across a variety of simulation parameters, such as number of inputs, output firing rate, input firing type, and simulation time. Since this approach requires only naturally generated spikes, it can be readily applied to behaving animal studies to characterize the underlying mechanisms of learning and memory.

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.

Developing a hippocampal neural prosthetic to facilitate human memory encoding and recall of stimulus features and categories.

Objective: Here, we demonstrate the first successful use of static neural stimulation patterns for specific information content. These static patterns were derived by a model that was applied to a subject's own hippocampal spatiotemporal neural codes for memory. Approach: We constructed a new model of processes by which the hippocampus encodes specific memory items via spatiotemporal firing of neural ensembles that underlie the successful encoding of targeted content into short-term memory. A memory decoding model (MDM) of hippocampal CA3 and CA1 neural firing was computed which derives a stimulation pattern for CA1 and CA3 neurons to be applied during the encoding (sample) phase of a delayed match-to-sample (DMS) human short-term memory task. Main results: MDM electrical stimulation delivered to the CA1 and CA3 locations in the hippocampus during the sample phase of DMS trials facilitated memory of images from the DMS task during a delayed recognition (DR) task that also included control images that were not from the DMS task. Across all subjects, the stimulated trials exhibited significant changes in performance in 22.4% of patient and category combinations. Changes in performance were a combination of both increased memory performance and decreased memory performance, with increases in performance occurring at almost 2 to 1 relative to decreases in performance. Across patients with impaired memory that received bilateral stimulation, significant changes in over 37.9% of patient and category combinations was seen with the changes in memory performance show a ratio of increased to decreased performance of over 4 to 1. Modification of memory performance was dependent on whether memory function was intact or impaired, and if stimulation was applied bilaterally or unilaterally, with nearly all increase in performance seen in subjects with impaired memory receiving bilateral stimulation. Significance: These results demonstrate that memory encoding in patients with impaired memory function can be facilitated for specific memory content, which offers a stimulation method for a future implantable neural prosthetic to improve human memory.

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.

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.

Columnar processing in primate pFC: evidence for executive control microcircuits.

A common denominator for many cognitive disorders of human brain is the disruption of neural activity within pFC, whose structural basis is primarily interlaminar (columnar) microcircuits or "minicolumns." The importance of this brain region for executive decision-making has been well documented; however, because of technological constraints, the minicolumnar basis is not well understood. Here, via implementation of a unique conformal multielectrode recording array, the role of interlaminar pFC minicolumns in the executive control of task-related target selection is demonstrated in nonhuman primates performing a visuomotor DMS task. The results reveal target-specific, interlaminar correlated firing during the decision phase of the trial between multielectrode recording array-isolated minicolumnar pairs of neurons located in parallel in layers 2/3 and layer 5 of pFC. The functional significance of individual pFC minicolumns (separated by 40 µm) was shown by reduced correlated firing between cell pairs within single minicolumns on error trials with inappropriate target selection. To further demonstrate dependence on performance, a task-disrupting drug (cocaine) was administered in the middle of the session, which also reduced interlaminar firing in minicolumns that fired appropriately in the early (nondrug) portion of the session. The results provide a direct demonstration of task-specific, real-time columnar processing in pFC indicating the role of this type of microcircuit in executive control of decision-making in primate brain.

Accelerating input-output model estimation with parallel computing for testing hippocampal memory prostheses in human.

BACKGROUND: Hippocampal memory prosthesis is defined as a closed-loop biomimetic system that can be used for restoration and enhancement of memory functions impaired in diseases or injuries. To build such a prosthesis, we have developed two types of input-output models, i.e., a multi-input multi-output (MIMO) model for predicting output spike trains based on input spikes, and a double-layer multi-resolution memory decoding (MD) model for classifying spatio-temporal patterns of spikes into memory categories. Both models can achieve high prediction accuracy using human hippocampal spikes data and can be used to derive electrical stimulation patterns to test the hippocampal memory prosthesis. METHODS: However, testing hippocampal memory prostheses in human epilepsy patients with such models has to be performed within a much shorter time window (48-72 h) due to clinical limitations. To solve this problem, we have developed parallelization strategies to decompose the overall model estimation task into multiple independent sub-tasks involving different outputs and cross-validation folds. These sub-tasks are then accomplished in parallel on different computer nodes to reduce model estimation time. RESULTS: Implementing both parallel schemes with a high-performance computer cluster, we successfully reduced the computing time of model estimations from hundreds of hours to tens of hours. COMPARISON WITH EXISTING METHOD: We have tested the two parallel computing schemes for both MIMO and MD models with data collected from 11 human subjects. The performances of the parallel schemes are compared with the performance of the non-parallel scheme. CONCLUSION: Such strategies allow us to complete the modeling procedure within the required time frame to further test input-output model-driven electrical stimulations for the hippocampal memory prosthesis. It has important implications to test the model-based DBS intraoperatively and developing clinically viable hippocampal memory prostheses.

Altered adult neurogenesis and gliogenesis in patients with mesial temporal lobe epilepsy.

The hippocampus is the most common seizure focus in people. In the hippocampus, aberrant neurogenesis plays a critical role in the initiation and progression of epilepsy in rodent models, but it is unknown whether this also holds true in humans. To address this question, we used immunofluorescence on control healthy hippocampus and surgical resections from mesial temporal lobe epilepsy (MTLE), plus neural stem-cell cultures and multi-electrode recordings of ex vivo hippocampal slices. We found that a longer duration of epilepsy is associated with a sharp decline in neuronal production and persistent numbers in astrogenesis. Further, immature neurons in MTLE are mostly inactive, and are not observed in cases with local epileptiform-like activity. However, immature astroglia are present in every MTLE case and their location and activity are dependent on epileptiform-like activity. Immature astroglia, rather than newborn neurons, therefore represent a potential target to continually modulate adult human neuronal hyperactivity.

Corrigendum: Patterned hippocampal stimulation facilitates memory in patients with a history of head impact and/or brain injury.

[This corrects the article DOI: 10.3389/fnhum.2022.933401.].

Patterned Hippocampal Stimulation Facilitates Memory in Patients With a History of Head Impact and/or Brain Injury.

RATIONALE: Deep brain stimulation (DBS) of the hippocampus is proposed for enhancement of memory impaired by injury or disease. Many pre-clinical DBS paradigms can be addressed in epilepsy patients undergoing intracranial monitoring for seizure localization, since they already have electrodes implanted in brain areas of interest. Even though epilepsy is usually not a memory disorder targeted by DBS, the studies can nevertheless model other memory-impacting disorders, such as Traumatic Brain Injury (TBI). METHODS: Human patients undergoing Phase II invasive monitoring for intractable epilepsy were implanted with depth electrodes capable of recording neurophysiological signals. Subjects performed a delayed-match-to-sample (DMS) memory task while hippocampal ensembles from CA1 and CA3 cell layers were recorded to estimate a multi-input, multi-output (MIMO) model of CA3-to-CA1 neural encoding and a memory decoding model (MDM) to decode memory information from CA3 and CA1 neuronal signals. After model estimation, subjects again performed the DMS task while either MIMO-based or MDM-based patterned stimulation was delivered to CA1 electrode sites during the encoding phase of the DMS trials. Each subject was sorted (post hoc) by prior experience of repeated and/or mild-to-moderate brain injury (RMBI), TBI, or no history (control) and scored for percentage successful delayed recognition (DR) recall on stimulated vs. non-stimulated DMS trials. The subject's medical history was unknown to the experimenters until after individual subject memory retention results were scored. RESULTS: When examined compared to control subjects, both TBI and RMBI subjects showed increased memory retention in response to both MIMO and MDM-based hippocampal stimulation. Furthermore, effects of stimulation were also greater in subjects who were evaluated as having pre-existing mild-to-moderate memory impairment. CONCLUSION: These results show that hippocampal stimulation for memory facilitation was more beneficial for subjects who had previously suffered a brain injury (other than epilepsy), compared to control (epilepsy) subjects who had not suffered a brain injury. This study demonstrates that the epilepsy/intracranial recording model can be extended to test the ability of DBS to restore memory function in subjects who previously suffered a brain injury other than epilepsy, and support further investigation into the beneficial effect of DBS in TBI patients.

Memory encoding in hippocampal ensembles is negatively influenced by cannabinoid CB1 receptors.

It has previously been demonstrated that the detrimental effect on the performance of a delayed nonmatch to sample (DNMS) memory task by exogenously administered cannabinoid (CB1) receptor agonist, WIN 55212-2 (WIN), is reversed by the receptor antagonist rimonabant. In addition, rimonabant administered alone elevates DNMS performance, presumably through the suppression of negative modulation by released endocannabinoids during normal task performance. Other investigations have shown that rimonabant enhances encoding of DNMS task-relevant information on a trial-by-trial, delay-dependent basis. In this study, these reciprocal pharmacological actions were completely characterized by long-term, chronic intrahippocampal infusion of both agents (WIN and rimonabant) in successive 2-week intervals. Such long-term exposure allowed extraction and confirmation of task-related firing patterns, in which rimonabant reversed the effects of CB1 agonists. This information was then utilized to artificially impose the facilitatory effects of rimonabant and to reverse the effects of WIN on DNMS performance, by delivering multichannel electrical stimulation in the same firing patterns to the same hippocampal regions. Direct comparison of normal and WIN-injected subjects, in which rimonabant injections and ensemble firing facilitated performance, verified reversal of the modulation of hippocampal memory processes by CB1 receptor agonists, including released endocannabinoids.

A sparse multiscale nonlinear autoregressive model for seizure prediction.

Objectives.Accurate seizure prediction is highly desirable for medical interventions such as responsive electrical stimulation. We aim to develop a classification model that can predict seizures by identifying preictal states, i.e. the precursor of a seizure, based on multi-channel intracranial electroencephalography (iEEG) signals.Approach.A two-level sparse multiscale classification model was developed to classify interictal and preictal states from iEEG data. In the first level, short time-scale linear dynamical features were extracted as autoregressive (AR) model coefficients; arbitrary (usually long) time-scale linear and nonlinear dynamical features were extracted as Laguerre-Volterra AR model coefficients; root-mean-square error of model prediction was used as a feature representing model unpredictability. In the second level, all features were fed into a sparse classifier to discriminate the iEEG data between interictal and preictal states.Main results. The two-level model can accurately classify seizure states using iEEG data recorded from ten canine and human subjects. Adding arbitrary (usually long) time-scale and nonlinear features significantly improves model performance compared with the conventional AR modeling approach. There is a high degree of variability in the types of features contributing to seizure prediction across different subjects.Significance. This study suggests that seizure generation may involve distinct linear/nonlinear dynamical processes caused by different underlying neurobiological mechanisms. It is necessary to build patient-specific classification models with a wide range of dynamical features.

Bridging Hierarchies in Multi-Scale Models of Neural Systems: Look-Up Tables Enable Computationally Efficient Simulations of Non-linear Synaptic Dynamics.

Synapses are critical actors of neuronal transmission as they form the basis of chemical communication between neurons. Accurate computational models of synaptic dynamics may prove important in elucidating emergent properties across hierarchical scales. Yet, in large-scale neuronal network simulations, synapses are often modeled as highly simplified linear exponential functions due to their small computational footprint. However, these models cannot capture the complex non-linear dynamics that biological synapses exhibit and thus, are insufficient in representing synaptic behavior accurately. Existing detailed mechanistic synapse models can replicate these non-linear dynamics by modeling the underlying kinetics of biological synapses, but their high complexity prevents them from being a suitable option in large-scale models due to long simulation times. This motivates the development of more parsimonious models that can capture the complex non-linear dynamics of synapses accurately while maintaining a minimal computational cost. We propose a look-up table approach that stores precomputed values thereby circumventing most computations at runtime and enabling extremely fast simulations for glutamatergic receptors AMPAr and NMDAr. Our results demonstrate that this methodology is capable of replicating the dynamics of biological synapses as accurately as the mechanistic synapse models while offering up to a 56-fold increase in speed. This powerful approach allows for multi-scale neuronal networks to be simulated at large scales, enabling the investigation of how low-level synaptic activity may lead to changes in high-level phenomena, such as memory and learning.

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.

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.