Conductance-based integrate-and-fire models

Neuromorphic computing spiking neural network edge detection model for content based image retrieval

In contemporary times, content-based image retrieval (CBIR) techniques have gained widespread acceptance as a means for end-users to discern and extract specific image content from vast repositories. However, it is noteworthy that a substantial majority of CBIR studies continue to rely on linear methodologies such as gradient-based and derivative-based edge detection techniques. This research explores the integration of bioinspired Spiking Neural Network (SNN) based edge detection within CBIR. We introduce an innovative, computationally efficient SNN-based approach designed explicitly for CBIR applications, outperforming existing SNN models by reducing computational overhead by 2.5 times. The proposed SNN-based edge detection approach is seamlessly incorporated into three distinct CBIR techniques, each employing conventional edge detection methodologies including Sobel, Canny, and image derivatives. Rigorous experimentation and evaluations are carried out utilizing the Corel-10k dataset and crop weed dataset, a widely recognized and frequently adopted benchmark dataset in the realm of image analysis. Importantly, our findings underscore the enhanced performance of CBIR methodologies integrating the proposed SNN-based edge detection approach, with an average increase in mean precision values exceeding 3%. This study conclusively demonstrated the utility of our proposed methodology in optimizing feature extraction, thereby establishing its pivotal role in advancing edge centric CBIR approaches.

Ultrafast simulation of large-scale neocortical microcircuitry with biophysically realistic neurons

Understanding the activity of the mammalian brain requires an integrative knowledge of circuits at distinct scales, ranging from ion channel gating to circuit connectomics. Computational models are regularly employed to understand how multiple parameters contribute synergistically to circuit behavior. However, traditional models of anatomically and biophysically realistic neurons are computationally demanding, especially when scaled to model local circuits. To overcome this limitation, we trained several artificial neural network (ANN) architectures to model the activity of realistic multicompartmental cortical neurons. We identified an ANN architecture that accurately predicted subthreshold activity and action potential firing. The ANN could correctly generalize to previously unobserved synaptic input, including in models containing nonlinear dendritic properties. When scaled, processing times were orders of magnitude faster compared with traditional approaches, allowing for rapid parameter-space mapping in a circuit model of Rett syndrome. Thus, we present a novel ANN approach allowing for rapid, detailed network experiments using inexpensive and commonly available computational resources.

A realistic morpho-anatomical connection strategy for modelling full-scale point-neuron microcircuits

The modeling of extended microcircuits is emerging as an effective tool to simulate the neurophysiological correlates of brain activity and to investigate brain dysfunctions. However, for specific networks, a realistic modeling approach based on the combination of available physiological, morphological and anatomical data is still an open issue. One of the main problems in the generation of realistic networks lies in the strategy adopted to build network connectivity. Here we propose a method to implement a neuronal network at single cell resolution by using the geometrical probability volumes associated with pre- and postsynaptic neurites. This allows us to build a network with plausible connectivity properties without the explicit use of computationally intensive touch detection algorithms using full 3D neuron reconstructions. The method has been benchmarked for the mouse hippocampus CA1 area, and the results show that this approach is able to generate full-scale brain networks at single cell resolution that are in good agreement with experimental findings. This geometric reconstruction of axonal and dendritic occupancy, by effectively reflecting morphological and anatomical constraints, could be integrated into structured simulators generating entire circuits of different brain areas facilitating the simulation of different brain regions with realistic models.

Neural data-driven model of spinal excitability changes induced by transcutaneous electrical stimulation in spinal cord injury subjects

The efficacy of trans-spinal direct current stimulation (tsDCS) as neurorehabilitation technology remains sub-optimal, partly due to the variability introduced by subject-specific neurophysiological features and stimulation conditions (e.g. electrode placement, stimulating amplitude, polarity, etc.) Hence, current therapies apply tsDCS in an open-loop fashion, resulting in a lack of standardized protocols for controlling elicited neuronal adaptations in closed-loop. Through the combination of high-density electromyogram (HD-EMG) decomposition, biophysical neuronal modelling and metaheuristic optimization, this work presents a novel neural data-driven framework for estimating subject-specific features and quantifying acute neuronal adaptations elicited by tsDCS on incomplete spinal cord injury subjects. This approach consists of calibrating the anatomical parameters (e.g. soma diameter) of in silico $\alpha-$motoneuron (MN) models for firing similarly to in vivo MNs decoded from HD-EMG. Assuming that cathodal-tsDCS elicits excitability changes in the MN pool, while preserving their anatomical parameters, optimization of an excitability gain common to the entire pool was performed to minimize discrepancies in firing rate and recruitment time between in vivo and in silico MNs after cathodal-tsDCS. This quantification of excitability changes on MN models calibrated in a person specific way enables closing the loop with neuro-modulation devices to tailor neurorehabilitation therapies. Clinical Relevance - This framework addresses a key limitation in non-invasive neuro-modulative technologies via a novel model-assisted framework that enables quantifying acute excitability changes induced on a person-specific in silico MN pool calibrated using in vivo neural data. This will enable the development of advanced controllers for modulating targeted neuronal adaptations in closed-loop.

Learning Long Temporal Sequences in Spiking Networks by Multiplexing Neural Oscillations

Many cognitive and behavioral tasks-such as interval timing, spatial navigation, motor control, and speech-require the execution of precisely-timed sequences of neural activation that cannot be fully explained by a succession of external stimuli. We show how repeatable and reliable patterns of spatiotemporal activity can be generated in chaotic and noisy spiking recurrent neural networks. We propose a general solution for networks to autonomously produce rich patterns of activity by providing a multi-periodic oscillatory signal as input. We show that the model accurately learns a variety of tasks, including speech generation, motor control, and spatial navigation. Further, the model performs temporal rescaling of natural spoken words and exhibits sequential neural activity commonly found in experimental data involving temporal processing. In the context of spatial navigation, the model learns and replays compressed sequences of place cells and captures features of neural activity such as the emergence of ripples and theta phase precession. Together, our findings suggest that combining oscillatory neuronal inputs with different frequencies provides a key mechanism to generate precisely timed sequences of activity in recurrent circuits of the brain.

Neurophysiological correlates of force control improvement induced by sinusoidal vibrotactile stimulation

OBJECTIVE

An optimal level of vibrotactile stimulation has been shown to improve sensorimotor control in healthy and diseased individuals. However, the underlying neurophysiological mechanisms behind the enhanced motor performance caused by vibrotactile stimulation are yet to be fully understood. Therefore, here we aim to evaluate the effect of a cutaneous vibration on the firing behavior of motor units in a condition of improved force steadiness.

APPROACH

Participants performed a visuomotor task, which consisted of low-intensity isometric contractions of the first dorsal interosseous (FDI) muscle, while sinusoidal (175 Hz) vibrotactile stimuli with different intensities were applied to the index finger. High-density surface electromyogram was recorded from the FDI muscle, and a decomposition algorithm was used to extract the motor unit spike trains. Additionally, computer simulations were performed using a multiscale neuromuscular model to provide a potential explanation for the experimental findings.

MAIN RESULTS

Experimental outcomes showed that an optimal level of vibration significantly improved force steadiness (estimated as the coefficient of variation of force). The decreased force variability was accompanied by a reduction in the variability of the smoothed cumulative spike train (as an estimation of the neural drive to the muscle), and the proportion of common inputs to the FDI motor nucleus. However, the interspike interval variability did not change significantly with the vibration. A mathematical approach, together with computer simulation results suggested that vibrotactile stimulation would reduce the variance of the common synaptic input to the motor neuron pool, thereby decreasing the low frequency fluctuations of the neural drive to the muscle and force steadiness.

SIGNIFICANCE

Our results demonstrate that the decreased variability in common input accounts for the enhancement in force control induced by vibrotactile stimulation.

Curated Model Development Using NEUROiD: A Web-Based NEUROmotor Integration and Design Platform

Decades of research on neuromotor circuits and systems has provided valuable information on neuronal control of movement. Computational models of several elements of the neuromotor system have been developed at various scales, from sub-cellular to system. While several small models abound, their structured integration is the key to building larger and more biologically realistic models which can predict the behavior of the system in different scenarios. This effort calls for integration of elements across neuroscience and musculoskeletal biomechanics. There is also a need for development of methods and tools for structured integration that yield larger in silico models demonstrating a set of desired system responses. We take a small step in this direction with the NEUROmotor integration and Design (NEUROiD) platform. NEUROiD helps integrate results from motor systems anatomy, physiology, and biomechanics into an integrated neuromotor system model. Simulation and visualization of the model across multiple scales is supported. Standard electrophysiological operations such as slicing, current injection, recording of membrane potential, and local field potential are part of NEUROiD. The platform allows traceability of model parameters to primary literature. We illustrate the power and utility of NEUROiD by building a simple ankle model and its controlling neural circuitry by curating a set of published components. NEUROiD allows researchers to utilize remote high-performance computers for simulation, while controlling the model using a web browser.

Dynamic Communications Between GABAA Switch, Local Connectivity, and Synapses During Cortical Development: A Computational Study

Several factors regulate cortical development, such as changes in local connectivity and the influences of dynamical synapses. In this study, we simulated various factors affecting the regulation of neural network activity during cortical development. Previous studies have shown that during early cortical development, the reversal potential of GABA_A shifts from depolarizing to hyperpolarizing. Here we provide the first integrative computational model to simulate the combined effects of these factors in a unified framework (building on our prior work: Khalil et al., 2017a,b). In the current study, we extend our model to monitor firing activity in response to the excitatory action of GABA_A. Precisely, we created a Spiking Neural Network model that included certain biophysical parameters for lateral connectivity (distance between adjacent neurons) and nearby local connectivity (complex connections involving those between neuronal groups). We simulated different network scenarios (for immature and mature conditions) based on these biophysical parameters. Then, we implemented two forms of Short-term synaptic plasticity (depression and facilitation). Each form has two distinct kinds according to its synaptic time constant value. Finally, in both sets of networks, we compared firing rate activity responses before and after simulating dynamical synapses. Based on simulation results, we found that the modulation effect of dynamical synapses for evaluating and shaping the firing activity of the neural network is strongly dependent on the physiological state of GABA_A. Moreover, the STP mechanism acts differently in every network scenario, mirroring the crucial modulating roles of these critical parameters during cortical development. Clinical implications for pathological alterations of GABAergic signaling in neurological and psychiatric disorders are discussed.

Spike-Conducting Integrate-and-Fire Model

Modeling is a useful tool for investigating various biophysical characteristics of neurons. Recent simulation studies of propagating action potentials (spike conduction) along axons include the investigation of neuronal activity evoked by electrical stimulation from implantable prosthetic devices. In contrast to point-neuron simulations, where a large variety of models are readily available, Hodgkin–Huxley-type conductance-based models have been almost the only option for simulating axonal spike conduction, as simpler models cannot faithfully replicate the waveforms of propagating spikes. Since the amount of available physiological data, especially in humans, is usually limited, calibration, and justification of the large number of parameters of a complex model is generally difficult. In addition, not all simulation studies of axons require detailed descriptions of nonlinear ionic dynamics. In this study, we construct a simple model of spike generation and conduction based on the exponential integrate-and-fire model, which can simulate the rapid growth of the membrane potential at spike initiation. In terms of the number of parameters and equations, this model is much more compact than conventional models, but can still reliably simulate spike conduction along myelinated and unmyelinated axons that are stimulated intracellularly or extracellularly. Our simulations of auditory nerve fibers with this new model suggest that, because of the difference in intrinsic membrane properties, the axonal spike conduction of high-frequency nerve fibers is faster than that of low-frequency fibers. The simple model developed in this study can serve as a computationally efficient alternative to more complex models for future studies, including simulations of neuroprosthetic devices.

Physiological models of the lateral superior olive

In computational biology, modeling is a fundamental tool for formulating, analyzing and predicting complex phenomena. Most neuron models, however, are designed to reproduce certain small sets of empirical data. Hence their outcome is usually not compatible or comparable with other models or datasets, making it unclear how widely applicable such models are. In this study, we investigate these aspects of modeling, namely credibility and generalizability, with a specific focus on auditory neurons involved in the localization of sound sources. The primary cues for binaural sound localization are comprised of interaural time and level differences (ITD/ILD), which are the timing and intensity differences of the sound waves arriving at the two ears. The lateral superior olive (LSO) in the auditory brainstem is one of the locations where such acoustic information is first computed. An LSO neuron receives temporally structured excitatory and inhibitory synaptic inputs that are driven by ipsi- and contralateral sound stimuli, respectively, and changes its spike rate according to binaural acoustic differences. Here we examine seven contemporary models of LSO neurons with different levels of biophysical complexity, from predominantly functional ones (‘shot-noise’ models) to those with more detailed physiological components (variations of integrate-and-fire and Hodgkin-Huxley-type). These models, calibrated to reproduce known monaural and binaural characteristics of LSO, generate largely similar results to each other in simulating ITD and ILD coding. Our comparisons of physiological detail, computational efficiency, predictive performances, and further expandability of the models demonstrate (1) that the simplistic, functional LSO models are suitable for applications where low computational costs and mathematical transparency are needed, (2) that more complex models with detailed membrane potential dynamics are necessary for simulation studies where sub-neuronal nonlinear processes play important roles, and (3) that, for general purposes, intermediate models might be a reasonable compromise between simplicity and biological plausibility.

Computational models help our understanding of complex biological systems, by identifying their key elements and revealing their operational principles. Close comparisons between model predictions and empirical observations ensure our confidence in a model as a building block for further applications. Most current neuronal models, however, are constructed to replicate only a small specific set of experimental data. Thus, it is usually unclear how these models can be generalized to different datasets and how they compare with each other. In this paper, seven neuronal models are examined that are designed to reproduce known physiological characteristics of auditory neurons involved in the detection of sound source location. Despite their different levels of complexity, the models generate largely similar results when their parameters are tuned with common criteria. Comparisons show that simple models are computationally more efficient and theoretically transparent, and therefore suitable for rigorous mathematical analyses and engineering applications including real-time simulations. In contrast, complex models are necessary for investigating the relationship between underlying biophysical processes and sub- and suprathreshold spiking properties, although they have a large number of unconstrained, unverified parameters. Having identified their advantages and drawbacks, these auditory neuron models may readily be used for future studies and applications.

Spike patterning in oxytocin neurons: Capturing physiological behaviour with Hodgkin-Huxley and integrate-and-fire models

Integrate-and-fire (IF) models can provide close matches to the discharge activity of neurons, but do they oversimplify the biophysical properties of the neurons? A single compartment Hodgkin-Huxley (HH) model of the oxytocin neuron has previously been developed, incorporating biophysical measurements of channel properties obtained in vitro. A simpler modified integrate-and-fire model has also been developed, which can match well the characteristic spike patterning of oxytocin neurons as observed in vivo. Here, we extended the HH model to incorporate synaptic input, to enable us to compare spike activity in the model with experimental data obtained in vivo. We refined the HH model parameters to closely match the data, and then matched the same experimental data with a modified IF model, using an evolutionary algorithm to optimise parameter matching. Finally we compared the properties of the modified HH model with those of the IF model to seek an explanation for differences between spike patterning in vitro and in vivo. We show that, with slight modifications, the original HH model, like the IF model, is able to closely match both the interspike interval (ISI) distributions of oxytocin neurons and the observed variability of spike firing rates in vivo and in vitro. This close match of both models to data depends on the presence of a slow activity-dependent hyperpolarisation (AHP); this is represented in both models and the parameters used in the HH model representation match well with optimal parameters of the IF model found by an evolutionary algorithm. The ability of both models to fit data closely also depends on a shorter hyperpolarising after potential (HAP); this is explicitly represented in the IF model, but in the HH model, it emerges from a combination of several components. The critical elements of this combination are identified.

Improving voltage-sensitive dye imaging: with a little help from computational approaches

Voltage-sensitive dye imaging (VSDI) is a key neurophysiological recording tool because it reaches brain scales that remain inaccessible to other techniques. The development of this technique from in vitro to the behaving nonhuman primate has only been made possible thanks to the long-lasting, visionary work of Amiram Grinvald. This work has opened new scientific perspectives to the great benefit to the neuroscience community. However, this unprecedented technique remains largely under-utilized, and many future possibilities await for VSDI to reveal new functional operations. One reason why this tool has not been used extensively is the inherent complexity of the signal. For instance, the signal reflects mainly the subthreshold neuronal population response and is not linked to spiking activity in a straightforward manner. Second, VSDI gives access to intracortical recurrent dynamics that are intrinsically complex and therefore nontrivial to process. Computational approaches are thus necessary to promote our understanding and optimal use of this powerful technique. Here, we review such approaches, from computational models to dissect the mechanisms and origin of the recorded signal, to advanced signal processing methods to unravel new neuronal interactions at mesoscopic scale. Only a stronger development of interdisciplinary approaches can bridge micro- to macroscales.

Reassessment of Non-Monosynaptic Excitation from the Motor Cortex to Motoneurons in Single Motor Units of the Human Biceps Brachii

Corticospinal excitation is mediated by polysynaptic pathways in several vertebrates, including dexterous monkeys. However, indirect non-monosynaptic excitation has not been clearly observed following transcranial electrical stimulation (TES) or cervicomedullary stimulation (CMS) in humans. The present study evaluated indirect motor pathways in normal human subjects by recording the activities of single motor units (MUs) in the biceps brachii (BB) muscle. The pyramidal tract was stimulated with weak TES, CMS, and transcranial magnetic stimulation (TMS) contralateral to the recording side. During tasks involving weak co-contraction of the BB and hand muscles, all stimulation methods activated MUs with short latencies. Peristimulus time histograms (PSTHs) showed that responses with similar durations were induced by TES (1.9 ± 1.4 ms) and CMS (2.0 ± 1.4 ms), and these responses often showed multiple peaks with the PSTH peak having a long duration (65.3% and 44.9%, respectively). Such long-duration excitatory responses with multiple peaks were rarely observed in the finger muscles following TES or in the BB following stimulation of the Ia fibers. The responses obtained with TES were compared in the same 14 BB MUs during the co-contraction and isolated BB contraction tasks. Eleven and three units, respectively, exhibited activation with multiple peaks during the two tasks. In order to determine the dispersion effects on the axon conduction velocities (CVs) and synaptic noise, a simulation study that was comparable to the TES experiments was performed with a biologically plausible neuromuscular model. When the model included the monosynaptic-pyramidal tract, multiple peaks were obtained in about 34.5% of the motoneurons (MNs). The experimental and simulation results indicated the existence of task-dependent disparate inputs from the pyramidal tract to the MNs of the upper limb. These results suggested that intercalated interneurons are present in the spinal cord and that these interneurons might be equivalent to those identified in animal experiments.

Impact of slow K(+) currents on spike generation can be described by an adaptive threshold model

A neuron that is stimulated by rectangular current injections initially responds with a high firing rate, followed by a decrease in the firing rate. This phenomenon is called spike-frequency adaptation and is usually mediated by slow K(+) currents, such as the M-type K(+) current (I M ) or the Ca(2+)-activated K(+) current (I AHP ). It is not clear how the detailed biophysical mechanisms regulate spike generation in a cortical neuron. In this study, we investigated the impact of slow K(+) currents on spike generation mechanism by reducing a detailed conductance-based neuron model. We showed that the detailed model can be reduced to a multi-timescale adaptive threshold model, and derived the formulae that describe the relationship between slow K(+) current parameters and reduced model parameters. Our analysis of the reduced model suggests that slow K(+) currents have a differential effect on the noise tolerance in neural coding.

The optimal neural strategy for a stable motor task requires a compromise between level of muscle cocontraction and synaptic gain of afferent feedback

Increasing joint stiffness by cocontraction of antagonist muscles and compensatory reflexes are neural strategies to minimize the impact of unexpected perturbations on movement. Combining these strategies, however, may compromise steadiness, as elements of the afferent input to motor pools innervating antagonist muscles are inherently negatively correlated. Consequently, a high afferent gain and active contractions of both muscles may imply negatively correlated neural drives to the muscles and thus an unstable limb position. This hypothesis was systematically explored with a novel computational model of the peripheral nervous system and the mechanics of one limb. Two populations of motor neurons received synaptic input from descending drive, spinal interneurons, and afferent feedback. Muscle force, simulated based on motor unit activity, determined limb movement that gave rise to afferent feedback from muscle spindles and Golgi tendon organs. The results indicated that optimal steadiness was achieved with low synaptic gain of the afferent feedback. High afferent gains during cocontraction implied increased levels of common drive in the motor neuron outputs, which were negatively correlated across the two populations, constraining instability of the limb. Increasing the force acting on the joint and the afferent gain both effectively minimized the impact of an external perturbation, and suboptimal adjustment of the afferent gain could be compensated by muscle cocontraction. These observations show that selection of the strategy for a given contraction implies a compromise between steadiness and effectiveness of compensations to perturbations. This indicates that a task-dependent selection of neural strategy for steadiness is necessary when acting in different environments.

Estimation of the synaptic input firing rates and characterization of the stimulation effects in an auditory neuron

To understand information processing in neuronal circuits, it is important to infer how a sensory stimulus impacts on the synaptic input to a neuron. An increase in neuronal firing during the stimulation results from pure excitation or from a combination of excitation and inhibition. Here, we develop a method for estimating the rates of the excitatory and inhibitory synaptic inputs from a membrane voltage trace of a neuron. The method is based on a modified Ornstein-Uhlenbeck neuronal model, which aims to describe the stimulation effects on the synaptic input. The method is tested using a single-compartment neuron model with a realistic description of synaptic inputs, and it is applied to an intracellular voltage trace recorded from an auditory neuron in vivo. We find that the excitatory and inhibitory inputs increase during stimulation, suggesting that the acoustic stimuli are encoded by a combination of excitation and inhibition.

Spinal mechanisms may provide a combination of intermittent and continuous control of human posture: predictions from a biologically based neuromusculoskeletal model

Several models have been employed to study human postural control during upright quiet stance. Most have adopted an inverted pendulum approximation to the standing human and theoretical models to account for the neural feedback necessary to keep balance. The present study adds to the previous efforts in focusing more closely on modelling the physiological mechanisms of important elements associated with the control of human posture. This paper studies neuromuscular mechanisms behind upright stance control by means of a biologically based large-scale neuromusculoskeletal (NMS) model. It encompasses: i) conductance-based spinal neuron models (motor neurons and interneurons); ii) muscle proprioceptor models (spindle and Golgi tendon organ) providing sensory afferent feedback; iii) Hill-type muscle models of the leg plantar and dorsiflexors; and iv) an inverted pendulum model for the body biomechanics during upright stance. The motor neuron pools are driven by stochastic spike trains. Simulation results showed that the neuromechanical outputs generated by the NMS model resemble experimental data from subjects standing on a stable surface. Interesting findings were that: i) an intermittent pattern of muscle activation emerged from this posture control model for two of the leg muscles (Medial and Lateral Gastrocnemius); and ii) the Soleus muscle was mostly activated in a continuous manner. These results suggest that the spinal cord anatomy and neurophysiology (e.g., motor unit types, synaptic connectivities, ordered recruitment), along with the modulation of afferent activity, may account for the mixture of intermittent and continuous control that has been a subject of debate in recent studies on postural control. Another finding was the occurrence of the so-called “paradoxical” behaviour of muscle fibre lengths as a function of postural sway. The simulations confirmed previous conjectures that reciprocal inhibition is possibly contributing to this effect, but on the other hand showed that this effect may arise without any anticipatory neural control mechanism.

The control of upright stance is a challenging task since the objective is to maintain the equilibrium of an intrinsically unstable biomechanical system. Somatosensory information is used by the central nervous system to modulate muscle contraction, which prevents the body from falling. While the visual and vestibular systems also provide important additional sensory information, a human being with only somatosensory inputs is able to maintain an upright stance. In this study, we used a biologically-based large-scale neuromusculoskeletal model driven only by somatosensory feedback to investigate human postural control from a neurophysiological point of view. No neural structures above the spinal cord were included in the model. The results showed that the model based on a spinal control of posture can reproduce several neuromechanical outcomes previously reported in the literature, including an intermittent muscle activation. Since this intermittent muscular recruitment is an emergent property of this spinal-like controller, we argue that the so-called intermittent control of upright stance might be produced by an interplay between spinal cord properties and modulated sensory inflow.

Influences of premotoneuronal command statistics on the scaling of motor output variability during isometric plantar flexion

This study focuses on neuromuscular mechanisms behind ankle torque and EMG variability during a maintained isometric plantar flexion contraction. Experimentally obtained torque standard deviation (SD) and soleus, medial gastrocnemius, and lateral gastrocnemius EMG envelope mean and SD increased with mean torque for a wide range of torque levels. Computer simulations were performed on a biophysically-based neuromuscular model of the triceps surae consisting of premotoneuronal spike trains (the global input, GI) driving the motoneuron pools of the soleus, medial gastrocnemius, and lateral gastrocnemius muscles, which activate their respective muscle units. Two types of point processes were adopted to represent the statistics of the GI: Poisson and Gamma. Simulations showed a better agreement with experimental results when the GI was modeled by Gamma point processes having lower orders (higher variability) for higher target torques. At the same time, the simulations reproduced well the experimental data of EMG envelope mean and SD as a function of mean plantar flexion torque, for the three muscles. These results suggest that the experimentally found relations between torque-EMG variability as a function of mean plantar flexion torque level depend not only on the intrinsic properties of the motoneuron pools and the muscle units innervated, but also on the increasing variability of the premotoneuronal GI spike trains when their mean rates increase to command a higher plantar flexion torque level. The simulations also provided information on spike train statistics of several hundred motoneurons that compose the triceps surae, providing a wide picture of the associated mechanisms behind torque and EMG variability.

Model of an excitatory synapse based on stochastic processes

We present a mathematical model of a biological synapse based on stochastic processes to establish the temporal behavior of the postsynaptic potential following a quantal synaptic transmission. This potential form is the basis of the neural code. We suppose that the release of neurotransmitters in the synaptic cleft follows a Poisson process, and that they diffuse according to integrated Ornstein-Uhlenbeck processes in 3-D with random initial positions and velocities. The diffusion occurs in an isotropic environment between two infinite parallel planes representing the pre- and postsynaptic membrane. We state that the presynaptic membrane is perfectly reflecting and that the other is perfectly absorbing. The activation of the receptors polarizes the postsynaptic membrane according to a parallel RC circuit scheme. We present the results obtained by simulations according to a Gillespie algorithm and we show that our model exhibits realistic postsynaptic behaviors from a simple quantal occurrence.

Models of passive and active dendrite motoneuron pools and their differences in muscle force control

Motoneuron (MN) dendrites may be changed from a passive to an active state by increasing the levels of spinal cord neuromodulators, which activate persistent inward currents (PICs). These exert a powerful influence on MN behavior and modify the motor control both in normal and pathological conditions. Motoneuronal PICs are believed to induce nonlinear phenomena such as the genesis of extra torque and torque hysteresis in response to percutaneous electrical stimulation or tendon vibration in humans. An existing large-scale neuromuscular simulator was expanded to include MN models that have a capability to change their dynamic behaviors depending on the neuromodulation level. The simulation results indicated that the variability (standard deviation) of a maintained force depended on the level of neuromodulatory activity. A force with lower variability was obtained when the motoneuronal network was under a strong influence of PICs, suggesting a functional role in postural and precision tasks. In an additional set of simulations when PICs were active in the dendrites of the MN models, the results successfully reproduced experimental results reported from humans. Extra torque was evoked by the self-sustained discharge of spinal MNs, whereas differences in recruitment and de-recruitment levels of the MNs were the main reason behind torque and electromyogram (EMG) hysteresis. Finally, simulations were also used to study the influence of inhibitory inputs on a MN pool that was under the effect of PICs. The results showed that inhibition was of great importance in the production of a phasic force, requiring a reduced co-contraction of agonist and antagonist muscles. These results show the richness of functionally relevant behaviors that can arise from a MN pool under the action of PICs.

Parameter estimation in spiking neural networks: a reverse-engineering approach

This paper presents a reverse engineering approach for parameter estimation in spiking neural networks (SNNs). We consider the deterministic evolution of a time-discretized network with spiking neurons, where synaptic transmission has delays, modeled as a neural network of the generalized integrate and fire type. Our approach aims at by-passing the fact that the parameter estimation in SNN results in a non-deterministic polynomial-time hard problem when delays are to be considered. Here, this assumption has been reformulated as a linear programming (LP) problem in order to perform the solution in a polynomial time. Besides, the LP problem formulation makes the fact that the reverse engineering of a neural network can be performed from the observation of the spike times explicit. Furthermore, we point out how the LP adjustment mechanism is local to each neuron and has the same structure as a 'Hebbian' rule. Finally, we present a generalization of this approach to the design of input-output (I/O) transformations as a practical method to 'program' a spiking network, i.e. find a set of parameters allowing us to exactly reproduce the network output, given an input. Numerical verifications and illustrations are provided.

Tell me something interesting: context dependent adaptation in somatosensory cortex

It is widely accepted that through a process of adaptation cells adjust their sensitivity in accordance with prevailing stimulus conditions. However, in two recent studies exploring adaptation in the rodent inferior colliculus and somatosensory cortex, neurons did not adapt towards global mean, but rather became most sensitive to inputs that were located towards the edge of the stimulus distribution with greater intensity than the mean. We re-examined electrophysiological data from the somatosensory study with the purpose of exploring the underlying encoding strategies. We found that neural gain tended to decrease as stimulus variance increased. Following adaptation to changes in global mean, neuronal output was scaled such that the relationship between firing rate and local, rather than global, differences in stimulus intensity was maintained. The majority of cells responded to large, positive deviations in stimulus amplitude; with a small number responding to both positive and negative changes in stimulus intensity. Adaptation to global mean was replicated in a model neuron by incorporating both spike-rate adaptation and tonic-inhibition, which increased in proportion to stimulus mean. Adaptation to stimulus variance was replicated by approximating the output of a population of neurons adapted to global mean and using it to drive a layer of recurrently connected depressing synapses. Within the barrel cortex, adaptation ensures that neurons are able to encode both overall levels of variance and large deviations in the input. This is achieved through a combination of gain modulation and a shift in sensitivity to intensity levels that are greater than the mean.

Comparison of different neuron models to conductance-based post-stimulus time histograms obtained in cortical pyramidal cells using dynamic-clamp in vitro

A wide diversity of models have been proposed to account for the spiking response of central neurons, from the integrate-and-fire (IF) model and its quadratic and exponential variants, to multiple-variable models such as the Izhikevich (IZ) model and the well-known Hodgkin-Huxley (HH) type models. Such models can capture different aspects of the spiking response of neurons, but there is few objective comparison of their performance. In this article, we provide such a comparison in the context of well-defined stimulation protocols, including, for each cell, DC stimulation, and a series of excitatory conductance injections, arising in the presence of synaptic background activity. We use the dynamic-clamp technique to characterize the response of regular-spiking neurons from guinea-pig visual cortex by computing families of post-stimulus time histograms (PSTH), for different stimulus intensities, and for two different background activities (low- and high-conductance states). The data obtained are then used to fit different classes of models such as the IF, IZ, or HH types, which are constrained by the whole data set. This analysis shows that HH models are generally more accurate to fit the series of experimental PSTH, but their performance is almost equaled by much simpler models, such as the exponential or pulse-based IF models. Similar conclusions were also reached by performing partial fitting of the data, and examining the ability of different models to predict responses that were not used for the fitting. Although such results must be qualified by using more sophisticated stimulation protocols, they suggest that nonlinear IF models can capture surprisingly well the response of cortical regular-spiking neurons and appear as useful candidates for network simulations with conductance-based synaptic interactions.

A spiking neural network model of the medial superior olive using spike timing dependent plasticity for sound localization

Sound localization can be defined as the ability to identify the position of an input sound source and is considered a powerful aspect of mammalian perception. For low frequency sounds, i.e., in the range 270 Hz–1.5 KHz, the mammalian auditory pathway achieves this by extracting the Interaural Time Difference between sound signals being received by the left and right ear. This processing is performed in a region of the brain known as the Medial Superior Olive (MSO). This paper presents a Spiking Neural Network (SNN) based model of the MSO. The network model is trained using the Spike Timing Dependent Plasticity learning rule using experimentally observed Head Related Transfer Function data in an adult domestic cat. The results presented demonstrate how the proposed SNN model is able to perform sound localization with an accuracy of 91.82% when an error tolerance of ±10° is used. For angular resolutions down to 2.5°, it will be demonstrated how software based simulations of the model incur significant computation times. The paper thus also addresses preliminary implementation on a Field Programmable Gate Array based hardware platform to accelerate system performance.

AnimatLab: a 3D graphics environment for neuromechanical simulations

The nervous systems of animals evolved to exert dynamic control of behavior in response to the needs of the animal and changing signals from the environment. To understand the mechanisms of dynamic control requires a means of predicting how individual neural and body elements will interact to produce the performance of the entire system. AnimatLab is a software tool that provides an approach to this problem through computer simulation. AnimatLab enables a computational model of an animal's body to be constructed from simple building blocks, situated in a virtual 3D world subject to the laws of physics, and controlled by the activity of a multicellular, multicompartment neural circuit. Sensor receptors on the body surface and inside the body respond to external and internal signals and then excite central neurons, while motor neurons activate Hill muscle models that span the joints and generate movement. AnimatLab provides a common neuromechanical simulation environment in which to construct and test models of any skeletal animal, vertebrate or invertebrate. The use of AnimatLab is demonstrated in a neuromechanical simulation of human arm flexion and the myotactic and contact-withdrawal reflexes.

Overview of facts and issues about neural coding by spikes

In the present overview, our wish is to demystify some aspects of coding with spike-timing, through a simple review of well-understood technical facts regarding spike coding. Our goal is a better understanding of the extent to which computing and modeling with spiking neuron networks might be biologically plausible and computationally efficient. We intentionally restrict ourselves to a deterministic implementation of spiking neuron networks and we consider that the dynamics of a network is defined by a non-stochastic mapping. By staying in this rather simple framework, we are able to propose results, formula and concrete numerical values, on several topics: (i) general time constraints, (ii) links between continuous signals and spike trains, (iii) spiking neuron networks parameter adjustment. Beside an argued review of several facts and issues about neural coding by spikes, we propose new results, such as a numerical evaluation of the most critical temporal variables that schedule the progress of realistic spike trains. When implementing spiking neuron networks, for biological simulation or computational purpose, it is important to take into account the indisputable facts here unfolded. This precaution could prevent one from implementing mechanisms that would be meaningless relative to obvious time constraints, or from artificially introducing spikes when continuous calculations would be sufficient and more simple. It is also pointed out that implementing a large-scale spiking neuron network is finally a simple task.

A novel role of dendritic gap junction and mechanisms underlying its interaction with thalamocortical conductance in fast spiking inhibitory neurons

BACKGROUND

Little is known about the roles of dendritic gap junctions (GJs) of inhibitory interneurons in modulating temporal properties of sensory induced responses in sensory cortices. Electrophysiological dual patch-clamp recording and computational simulation methods were used in combination to examine a novel role of GJs in sensory mediated feed-forward inhibitory responses in barrel cortex layer IV and its underlying mechanisms.

RESULTS

Under physiological conditions, excitatory post-junctional potentials (EPJPs) interact with thalamocortical (TC) inputs within an unprecedented few milliseconds (i.e. over 200 Hz) to enhance the firing probability and synchrony of coupled fast-spiking (FS) cells. Dendritic GJ coupling allows fourfold increase in synchrony and a significant enhancement in spike transmission efficacy in excitatory spiny stellate cells. The model revealed the following novel mechanisms: 1) rapid capacitive current (Icap) underlies the activation of voltage-gated sodium channels; 2) there was less than 2 milliseconds in which the Icap underlying TC input and EPJP was coupled effectively; 3) cells with dendritic GJs had larger input conductance and smaller membrane response to weaker inputs; 4) synchrony in inhibitory networks by GJ coupling leads to reduced sporadic lateral inhibition and increased TC transmission efficacy.

CONCLUSION

Dendritic GJs of neocortical inhibitory networks can have very powerful effects in modulating the strength and the temporal properties of sensory induced feed-forward inhibitory and excitatory responses at a very high frequency band (>200 Hz). Rapid capacitive currents are identified as main mechanisms underlying interaction between two transient synaptic conductances.

Network-state modulation of power-law frequency-scaling in visual cortical neurons

Various types of neural-based signals, such as EEG, local field potentials and intracellular synaptic potentials, integrate multiple sources of activity distributed across large assemblies. They have in common a power-law frequency-scaling structure at high frequencies, but it is still unclear whether this scaling property is dominated by intrinsic neuronal properties or by network activity. The latter case is particularly interesting because if frequency-scaling reflects the network state it could be used to characterize the functional impact of the connectivity. In intracellularly recorded neurons of cat primary visual cortex in vivo, the power spectral density of V_m activity displays a power-law structure at high frequencies with a fractional scaling exponent. We show that this exponent is not constant, but depends on the visual statistics used to drive the network. To investigate the determinants of this frequency-scaling, we considered a generic recurrent model of cortex receiving a retinotopically organized external input. Similarly to the in vivo case, our in computo simulations show that the scaling exponent reflects the correlation level imposed in the input. This systematic dependence was also replicated at the single cell level, by controlling independently, in a parametric way, the strength and the temporal decay of the pairwise correlation between presynaptic inputs. This last model was implemented in vitro by imposing the correlation control in artificial presynaptic spike trains through dynamic-clamp techniques. These in vitro manipulations induced a modulation of the scaling exponent, similar to that observed in vivo and predicted in computo. We conclude that the frequency-scaling exponent of the V_m reflects stimulus-driven correlations in the cortical network activity. Therefore, we propose that the scaling exponent could be used to read-out the “effective” connectivity responsible for the dynamical signature of the population signals measured at different integration levels, from Vm to LFP, EEG and fMRI.

Intracellular recording of neocortical neurons provides an opportunity of characterizing the statistical signature of the synaptic bombardment to which it is submitted. Indeed the membrane potential displays intense fluctuations which reflect the cumulative activity of thousands of input neurons. In sensory cortical areas, this measure could be used to estimate the correlational structure of the external drive. We show that changes in the statistical properties of network activity, namely the local correlation between neurons, can be detected by analyzing the power spectrum density (PSD) of the subthreshold membrane potential. These PSD can be fitted by a power-law function 1/f^α in the upper temporal frequency range. In vivo recordings in primary visual cortex show that the α exponent varies with the statistics of the sensory input. Most remarkably, the exponent observed in the ongoing activity is indistinguishable from that evoked by natural visual statistics. These results are emulated by models which demonstrate that the exponent α is determined by the local level of correlation imposed in the recurrent network activity. Similar relationships are also reproduced in cortical neurons recorded in vitro with artificial synaptic inputs by controlling in computo the level of correlation in real time.

Optimality and robustness of a biophysical decision-making model under norepinephrine modulation

The locus ceruleus (LC) can exhibit tonic or phasic activity and release norepinephrine (NE) throughout the cortex, modulating cellular excitability and synaptic efficacy and thus influencing behavioral performance. We study the effects of LC-NE modulation on decision making in two-alternative forced-choice tasks by changing conductances in a biophysical neural network model, and we investigate how it affects performance measured in terms of reward rate. We find that low tonic NE levels result in unmotivated behavior and high levels in impulsive, inaccurate choices, but that near-optimal performance can occur over a broad middle range. Robustness is greatest when pyramidal cells are less strongly modulated than interneurons, and superior performance can be achieved with phasic NE release, provided only glutamatergic synapses are modulated. We also show that network functions such as sensory information accumulation and short-term memory can be modulated by tonic NE levels, and that previously observed diverse evoked cell responses may be due to network effects.

Functional consequences of model complexity in rhythmic systems: I. Systematic reduction of a bursting neuron model

Neural models are increasingly being used as design components of physical systems. In order to most effectively utilize neuronal models in these novel contexts, we need to develop design rules for neuronal systems that relate how model design affects overall system performance. In this paper and a companion article, we investigate how the complexity of a neural model affects the performance of a two-cell oscillator built from the model. In this paper, we create a series of related neuron models with different mathematical complexity. Starting with a complex mechanistic model of a bursting neuron, we use a variety of techniques to create a series of simplified neuron models. These three reduced models produce bursting activity that is qualitatively very similar to the original model. In the following companion article, we investigate the functional performance of oscillators built from these models.

A review of the integrate-and-fire neuron model: I. Homogeneous synaptic input

The integrate-and-fire neuron model is one of the most widely used models for analyzing the behavior of neural systems. It describes the membrane potential of a neuron in terms of the synaptic inputs and the injected current that it receives. An action potential (spike) is generated when the membrane potential reaches a threshold, but the actual changes associated with the membrane voltage and conductances driving the action potential do not form part of the model. The synaptic inputs to the neuron are considered to be stochastic and are described as a temporally homogeneous Poisson process. Methods and results for both current synapses and conductance synapses are examined in the diffusion approximation, where the individual contributions to the postsynaptic potential are small. The focus of this review is upon the mathematical techniques that give the time distribution of output spikes, namely stochastic differential equations and the Fokker-Planck equation. The integrate-and-fire neuron model has become established as a canonical model for the description of spiking neurons because it is capable of being analyzed mathematically while at the same time being sufficiently complex to capture many of the essential features of neural processing. A number of variations of the model are discussed, together with the relationship with the Hodgkin-Huxley neuron model and the comparison with electrophysiological data. A brief overview is given of two issues in neural information processing that the integrate-and-fire neuron model has contributed to - the irregular nature of spiking in cortical neurons and neural gain modulation.

Essential role of the persistent sodium current in spike initiation during slowly rising inputs in mouse spinal neurones

Spinal motoneurons, like many neurons, respond with repetitive spiking to sustained inputs. The afterhyperpolarization (AHP) that follows each spike, however, decays relatively slowly in motoneurons. The slow depolarization during this decay should allow sodium (Na+) channel inactivation to keep up with its activation and thus should prevent initiation of the next spike. We hypothesized that the persistent component of the total Na+ current provides the mechanism that generates a rate of rise sufficiently rapid to generate a spike. In large cultured spinal neurons, presumed to be primarily motoneurons, inhibition of persistent sodium current (NaP) by the drug riluzole at low concentrations resulted in a loss of repetitive firing. However, cells remained fully capable of producing spikes to transient inputs. These effects of riluzole were not due to insufficient depolarization, enhancement of the AHP, or sustained Na+ channel inactivation. To further test this hypothesis, computer simulations were performed with a kinetic Na+ channel model that provided greater independent control of NaP relative to transient Na+ current (NaT) than that provided by riluzole administration. The model was tuned to generate substantial NaP and exhibited good repetitive firing to slowly rising inputs. When NaP was sharply reduced without significantly altering NaT, the model reproduced the effects of riluzole administration, inducing failure of repetitive firing but allowing single spikes in response to sharp transients. These results strongly support the essential role of NaP in spike initiation to slow inputs in spinal neurons. NaP may play a fundamental role in determining how a neuron responds to sustained inputs.

Generalized Integrate-and-Fire Models of Neuronal Activity Approximate Spike Trains of a Detailed Model to a High Degree of Accuracy

We demonstrate that single-variable integrate-and-fire models can quantitatively capture the dynamics of a physiologically detailed model for fast-spiking cortical neurons. Through a systematic set of approximations, we reduce the conductance-based model to 2 variants of integrate-and-fire models. In the first variant (nonlinear integrate-and-fire model), parameters depend on the instantaneous membrane potential, whereas in the second variant, they depend on the time elapsed since the last spike [Spike Response Model (SRM)]. The direct reduction links features of the simple models to biophysical features of the full conductance-based model. To quantitatively test the predictive power of the SRM and of the nonlinear integrate-and-fire model, we compare spike trains in the simple models to those in the full conductance-based model when the models are subjected to identical randomly fluctuating input. For random current input, the simple models reproduce 70–80 percent of the spikes in the full model (with temporal precision of ±2 ms) over a wide range of firing frequencies. For random conductance injection, up to 73 percent of spikes are coincident. We also present a technique for numerically optimizing parameters in the SRM and the nonlinear integrate-and-fire model based on spike trains in the full conductance-based model. This technique can be used to tune simple models to reproduce spike trains of real neurons.

Hybrid integrate-and-fire model of a bursting neuron

We present a reduction of a Hodgkin-Huxley (HH)--style bursting model to a hybridized integrate-and-fire (IF) formalism based on a thorough bifurcation analysis of the neuron's dynamics. The model incorporates HH--style equations to evolve the subthreshold currents and includes IF mechanisms to characterize spike events and mediate interactions between the subthreshold and spiking currents. The hybrid IF model successfully reproduces the dynamic behavior and temporal characteristics of the full model over a wide range of activity, including bursting and tonic firing. Comparisons of timed computer simulations of the reduced model and the original model for both single neurons and moderately sized networks (n < or = 500) show that this model offers improvement in computational speed over the HH--style bursting model.

Synchronization of the neural response to noisy periodic synaptic input

The timing information contained in the response of a neuron to noisy periodic synaptic input is analyzed for the leaky integrate-and-fire neural model. We address the question of the relationship between the timing of the synaptic inputs and the output spikes. This requires an analysis of the interspike interval distribution of the output spikes, which is obtained in the gaussian approximation. The conditional output spike density in response to noisy periodic input is evaluated as a function of the initial phase of the inputs. This enables the phase transition matrix to be calculated, which relates the phase at which the output spike is generated to the initial phase of the inputs. The interspike interval histogram and the period histogram for the neural response to ongoing periodic input are then evaluated by using the leading eigenvector of this phase transition matrix. The synchronization index of the output spikes is found to increase sharply as the inputs become synchronized. This enhancement of synchronization is most pronounced for large numbers of inputs and lower frequencies of modulation and also for rates of input near the critical input rate. However, the mutual information between the input phase of the stimulus and the timing of output spikes is found to decrease at low input rates as the number of inputs increases. The results show close agreement with those obtained from numerical simulations for large numbers of inputs.

NEURON: a tool for neuroscientists

NEURON is a simulation environment for models of individual neurons and networks of neurons that are closely linked to experimental data. NEURON provides tools for conveniently constructing, exercising, and managing models, so that special expertise in numerical methods or programming is not required for its productive use. This article describes two tools that address the problem of how to achieve computational efficiency and accuracy.

Synthesis of generalized algorithms for the fast computation of synaptic conductances with Markov kinetic models in large network simulations

Markovkinetic models constitute a powerful framework to analyze patch-clamp data from single-channel recordings and model the dynamics of ion conductances and synaptic transmission between neurons. In particular, the accurate simulation of a large number of synaptic inputs in wide-scale network models may result in a computationally highly demanding process. We present a generalized consolidating algorithm to simulate efficiently a large number of synaptic inputs of the same kind (excitatory or inhibitory), converging on an isopotential compartment, independently modeling each synaptic current by a generic n-state Markov model characterized by piece-wise constant transition probabilities. We extend our findings to a class of simplified phenomenological descriptions of synaptic transmission that incorporate higher-order dynamics, such as short-term facilitation, depression, and synaptic plasticity.