Real Neurons Research Articles

Deep-Learning-Based Automated Neuron Reconstruction From 3D Microscopy Images Using Synthetic Training Images.

Digital reconstruction of neuronal structures from 3D microscopy images is critical for the quantitative investigation of brain circuits and functions. It is a challenging task that would greatly benefit from automatic neuron reconstruction methods. In this paper, we propose a novel method called SPE-DNR that combines spherical-patches extraction (SPE) and deep-learning for neuron reconstruction (DNR). Based on 2D Convolutional Neural Networks (CNNs) and the intensity distribution features extracted by SPE, it determines the tracing directions and classifies voxels into foreground or background. This way, starting from a set of seed points, it automatically traces the neurite centerlines and determines when to stop tracing. To avoid errors caused by imperfect manual reconstructions, we develop an image synthesizing scheme to generate synthetic training images with exact reconstructions. This scheme simulates 3D microscopy imaging conditions as well as structural defects, such as gaps and abrupt radii changes, to improve the visual realism of the synthetic images. To demonstrate the applicability and generalizability of SPE-DNR, we test it on 67 real 3D neuron microscopy images from three datasets. The experimental results show that the proposed SPE-DNR method is robust and competitive compared with other state-of-the-art neuron reconstruction methods.

A hybrid artificial neural network for the generation of critical fluctuations and inter-spike intervals

The recently introduced [Contoyiannis et al., 2021] hybrid artificial neural network can simulate the dynamics of membrane potential fluctuations of real neurons based on fundamental principles of Physics. Here, we propose a temporal description of the membrane potential fluctuations, which resembles the soliton solutions in φ4 field theory. Within this framework, kink-antikink dynamics are associated with spike generation. Furthermore, we show that the simulation can also reproduce the distribution of inter-spike intervals of biological neurons in their critical state [Kosmidis et al., 2018]. A proposal for the intermittency origin of these fluctuations is discussed.

Energy analysis of bursting Hindmarsh-Rose neurons with time-delayed coupling

Mathematical modeling is an important tool to study the role of delay in neural systems and to evaluate its effects on the signaling activity of coupled neurons. Models for delayed neurons are often used to represent the dynamics of real neurons, but rarely to assess the energy required to maintain these dynamics. In this work, we address these questions from an energy perspective by considering a pair of Hindmarsh-Rose burst neurons coupled by reciprocal time-delayed coupling with electrical and chemical synapses. We examine the average energy consumption required to maintain cooperative behavior and quantify the contribution of synapses to total energy consumption. We show that unlike electrical coupling, where the time delay appears to reduce the instantaneous average relative weight of the synaptic contribution, in chemical coupling this average synaptic contribution appears to be much higher in delayed coupling than in instantaneous coupling, except at certain values of coupling strength where the instantaneous synaptic contribution is more important.

Optical dendrites for spatio-temporal computing with few-mode fibers [Invited

Multimode fibers have been extensively used in short-range communication systems and optical imaging. More recently, they have been considered for optical computing by exploiting the complex spatial and spatio-temporal transformation at the fiber end. Mimicking the dendrites of real neurons, we consider here the spatial modes that propagate with different group velocities along the fiber as different dendritic branches. The multimode fiber plays the role of an optical dendritic unit and the signals obtained from the different dendritic branches are temporally mixed and used for spatio-temporal information processing. We numerically demonstrate the use of a few-mode, step-index fiber as a linear computing element in an ultra-fast spatio-temporal coincidence detector that operates at 40 Gb/s data encoding rate. We evaluate this detector as a linear classifier in header recognition and bit counting tasks.

Experimental and simulation studies of localization and decoding of single and double dipoles

Objective. Electroencephalography is a technique for measuring normal or abnormal neuronal activity in the human brain, but its low spatial resolution makes it difficult to locate the precise locations of neurons due to the volume conduction effect of brain tissue. Approach. The acoustoelectric (AE) effect has the advantage of detecting electrical signals with high temporal resolution and focused ultrasound with high spatial resolution. In this paper, we use dipoles to simulate real single and double neurons, and further investigate the localization and decoding of single and double dipoles based on AE effects from numerical simulations, brain tissue phantom experiments, and fresh porcine brain tissue experiments. Main results. The results show that the localization error of a single dipole is less than 0.3 mm, the decoding signal is highly correlated with the source signal, and the decoding accuracy is greater than 0.94; the location of double dipoles with an interval of 0.4 mm or more can be localized, the localization error tends to increase as the interval of dipoles decreases, and the decoding accuracy tends to decrease as the frequency of dipoles decreases. Significance. This study localizes and decodes dipole signals with high accuracy, and provides a technical method for the development of EEG.

Analysis of the electronic integrate and fire neuron model

Nano-scale devices are thought to intervene in natural life for a variety of responsibilities. For understanding the intrinsic communication of such nano-scale devices, software and hardware modalities have been introduced. Some of these models are of neuro-spike communication systems which employ spiking neuron circuits. In this study, the previously designed electronic integrate and fire circuit inspired by Hodgkin Huxley membrane model is analyzedand interrelated to the Izhikevich’s systematic integrate and fire model. The generated action potentials with this model are very similar to the ones generated by real biophysical neurons which are thought as the inter-neuronal ionic transporters of information. The superiority of the analyzed model to the existing models is that it can show pulse trains whose characteristics are almost similar to those produced by nerve cells. The analytical, hardware and simulation results have shown that the model has the potential of employment in the smart nano-scale systems and medical treatment strategies.

Diffusion time dependence, power-law scaling, and exchange in gray matter

Characterizing neural tissue microstructure is a critical goal for future neuroimaging. Diffusion MRI (dMRI) provides contrasts that reflect diffusing spins’ interactions with myriad microstructural features of biological systems. However, the specificity of dMRI remains limited due to the ambiguity of its signals vis-à-vis the underlying microstructure. To improve specificity, biophysical models of white matter (WM) typically express dMRI signals according to the Standard Model (SM) and have more recently in gray matter (GM) taken spherical compartments into account (the SANDI model) in attempts to represent cell soma. The validity of the assumptions underlying these models, however, remains largely undetermined, especially in GM. To validate these assumptions experimentally, observing their unique, functional properties, such as the b−1/2 power-law associated with one-dimensional diffusion, has emerged as a fruitful strategy. The absence of this signature in GM, in turn, has been explained by neurite water exchange, non-linear morphology, and/or by obscuring soma signal contributions. Here, we present diffusion simulations in realistic neurons demonstrating that curvature and branching does not destroy the stick power-law behavior in impermeable neurites, but also that their signal is drowned by the soma signal under typical experimental conditions. Nevertheless, by studying the GM dMRI signal's behavior as a function of diffusion weighting as well as time, we identify an attainable experimental regime in which the neurite signal dominates. Furthermore, we find that exchange-driven time dependence produces a signal behavior opposite to that which would be expected from restricted diffusion, thereby providing a functional signature that disambiguates the two effects. We present data from dMRI experiments in ex vivo rat brain at ultrahigh field of 16.4T and observe a time dependence that is consistent with substantial exchange but also with a GM stick power-law. The first finding suggests significant water exchange between neurites and the extracellular space while the second suggests a small sub-population of impermeable neurites. To quantify these observations, we harness the Kärger exchange model and incorporate the corresponding signal time dependence in the SM and SANDI models.

Effects of transcranial alternating current stimulation on spiking activity in computational models of single neocortical neurons

Neural oscillations are a key mechanism for information transfer in brain circuits. Rhythmic fluctuations of local field potentials control spike timing through cyclic membrane de- and hyperpolarization. Transcranial alternating current stimulation (tACS) is a non-invasive neuromodulation method which can directly interact with brain oscillatory activity by imposing an oscillating electric field on neurons. Despite its increasing use, the basic mechanisms of tACS are still not fully understood. Here, we investigate in a computational study the effects of tACS on morphologically realistic neurons with ongoing spiking activity. We characterize the membrane polarization as a function of electric field strength and subsequent effects on spiking activity in a set of 25 neurons from different neocortical layers. We find that tACS does not affect the firing rate of investigated neurons for electric field strengths applicable to human studies. However, we find that the applied electric fields entrain the spiking activity of large pyramidal neurons and large basket neurons at < 1 mV/mm field strengths. Our model results are in line with recent experimental studies and can provide a mechanistic framework to understand the effects of oscillating electric fields on single neuron activity. They highlight the importance of neuron morphology and biophysics in responsiveness to electrical stimulation.

Theory of Gating in Recurrent Neural Networks.

Recurrent neural networks (RNNs) are powerful dynamical models, widely used in machine learning (ML) and neuroscience. Prior theoretical work has focused on RNNs with additive interactions. However gating i.e., multiplicative interactions are ubiquitous in real neurons and also the central feature of the best-performing RNNs in ML. Here, we show that gating offers flexible control of two salient features of the collective dynamics: (i) timescales and (ii) dimensionality. The gate controlling timescales leads to a novel marginally stable state, where the network functions as a flexible integrator. Unlike previous approaches, gating permits this important function without parameter fine-tuning or special symmetries. Gates also provide a flexible, context-dependent mechanism to reset the memory trace, thus complementing the memory function. The gate modulating the dimensionality can induce a novel, discontinuous chaotic transition, where inputs push a stable system to strong chaotic activity, in contrast to the typically stabilizing effect of inputs. At this transition, unlike additive RNNs, the proliferation of critical points (topological complexity) is decoupled from the appearance of chaotic dynamics (dynamical complexity). The rich dynamics are summarized in phase diagrams, thus providing a map for principled parameter initialization choices to ML practitioners.

Distribution of spiking and bursting in Rulkov’s neuron model

Large-scale brain simulations require the investigation of large networks of realistic neuron models, usually represented by sets of differential equations. Here we report a detailed fine-scale study of the dynamical response over extended parameter ranges of a computationally inexpensive model, the two-dimensional Rulkov map, which reproduces well the spiking and spiking-bursting activity of real biological neurons. In addition, we provide evidence of the existence of nested arithmetic progressions among periodic pulsing and bursting phases of Rulkov’s neuron. We find that specific remarkably complex nested sequences of periodic neural oscillations can be expressed as simple linear combinations of pairs of certain basal periodicities. Moreover, such nested progressions are robust and can be observed abundantly in diverse control parameter planes which are described in detail. We believe such findings to add significantly to the knowledge of Rulkov neuron dynamics and to be potentially helpful in large-scale simulations of the brain and other complex neuron networks.

The Engineered CRISPR-CAS System is a Beneficial Biological Tool for Detecting and Combating Antibiotic Resistance Microbes

Nowadays, the quick detection of antibiotic resistance bacteria causes a major problem in the field of the development of new antibiotics against the resistant bacteria. To overcome this problem, genome editing tools like clustered regularly interspaced short palindromic repeats (CRISPR) can be used. The CRISPR-CAS system is useful for targeting and killing antibiotic-resistant bacteria by cleaving resistance genes. It is also used to detect antibioticresistant bacteria. CRISPR is made up of a single guide RNA and the CAS 9 protein. The single guide RNA is used to guide toward the target sequence, and the CAS 9 protein is an enzyme that cuts DNA and is used in conjunction with the guide RNA. This modified sgRNA contains a complementary sequence to that of the target resistance gene and recognizes the target resistance sequence; therefore, it is cleaved by CAS-9 protein, and the removal of the resistance gene turns bacteria into antibiotic-sensitive ones. One of the delivery systems of CRISPR into bacteria is via bacteriophage.

Two-Argument Activation Functions Learn Soft XOR Operations Like Cortical Neurons.

Neurons in the brain are complex machines with distinct functional compartments that interact nonlinearly. In contrast, neurons in artificial neural networks abstract away this complexity, typically down to a scalar activation function of a weighted sum of inputs. Here we emulate more biologically realistic neurons by learning canonical activation functions with two input arguments, analogous to basal and apical dendrites. We use a network-in-network architecture where each neuron is modeled as a multilayer perceptron with two inputs and a single output. This inner perceptron is shared by all units in the outer network. Remarkably, the resultant nonlinearities often produce soft XOR functions, consistent with recent experimental observations about interactions between inputs in human cortical neurons. When hyperparameters are optimized, networks with these nonlinearities learn faster and perform better than conventional ReLU nonlinearities with matched parameter counts, and they are more robust to natural and adversarial perturbations.

Critical behavior in the artificial axon

The Artificial Axon is a unique synthetic system, based on biomolecular components, which supports action potentials. Here we examine, experimentally and theoretically, the properties of the threshold for firing in this system. As in real neurons, this threshold corresponds to the critical point of a saddle-node bifurcation. We measure the delay time for firing as a function of the distance to threshold, recovering the expected scaling exponent of −1/2. We introduce a minimal model of the Morris-Lecar type, validate it on the experiments, and use it to extend analytical results obtained in the limit of ‘fast’ ion channel dynamics. In particular, we discuss the dependence of the firing threshold on the number of channels. The Artificial Axon is a simplified system, an Ur-neuron, relying on only one ion channel species for functioning. Nonetheless, universal properties such as the action potential behavior near threshold are the same as in real neurons. Thus we may think of the Artificial Axon as a cell-free breadboard for electrophysiology research.

Analysis of the limiting behavior of a biological neurons system with delay

In this work an analytical and numerical analysis of the limiting behaviors of a system consisting of a pair of biological neurons was carried out. In this case connection between neurons will occur with a delay. As a neuron model, the FitzHugh-Nagumo model was chosen as a model that can reproduce many dynamic behaviors of a real neuron and, at the same time, is not very complex computationally.

Development of new deep neural network architecture based on hodgkin huxley model

Abstract There have been many studies to simulate the functions of the brain from the neural level to the cognitive function. Since the deep neural network has shown high performance for many tasks, researches for artificial intelligence are focused on deep neural networks. In terms of neurophysiology, the configuration of nodes used for a deep neural network is too simplistic to represent real neurons. Therefore, simulation of the brain activities in response to stimuli from the external environment using the deep neural networks is inappropriate. For that reason, our group makes a new deep neural network configuration based on the Hodgkin Huxley model. The Hodgkin Huxley model has high biological plausibility for representing the firing pattern of real neurons. Therefore, the deep neural network architecture based on the Hodgkin Huxley model successfully mimics human cognitive process and is suitable for simulation assuming external stimuli. The most important point for constructing the new architecture is setting the proper learning rule that updates the synaptic information for solving cognitive problems. We used the idea of reinforcement learning so that each neuron composing the neural layer does the synaptic information update as an agent. This learning method enables to increase the computing time linearly proportional to the number of neurons, unlike convolution neural networks. Also, we introduce an algorithm that optimizes the firing rate of neurons to generate synaptic information using learning. The algorithm represents the characteristics of brain waves found in brain imaging techniques. We trained the new architecture to classify 10 class mnist input data. The classification performance is very lower than other existing deep neural network architecture. However, the new architecture shows the potential to connect from neuronal level to cognitive level. Keywords: deep neural network, Hogkin Huxely, single neuron model, cognitive model

Spontaneous representation of numerosity zero in a deep neural network for visual object recognition

SummaryConceiving “nothing” as a numerical value zero is considered a sophisticated numerical capability that humans share with cognitively advanced animals. We demonstrate that representation of zero spontaneously emerges in a deep learning neural network without any number training. As a signature of numerical quantity representation, and similar to real neurons from animals, numerosity zero network units show maximum activity to empty sets and a gradual decrease in activity with increasing countable numerosities. This indicates that the network spontaneously ordered numerosity zero as the smallest numerical value along the number line. Removal of empty-set network units caused specific deficits in the network's judgment of numerosity zero, thus reflecting these units' functional relevance. These findings suggest that processing visual information is sufficient for a visual number sense that includes zero to emerge and explains why cognitively advanced animals with whom we share a nonverbal number system exhibit rudiments of numerosity zero.

Exploiting the electrothermal timescale in PrMnO3 RRAM for a compact, clock-less neuron exhibiting biological spiking patterns

Spiking neural networks (SNNs) are gaining widespread momentum in the field of neuromorphic computing. These network systems integrated with neurons and synapses provide computational efficiency by mimicking the human brain. It is desired to incorporate the biological neuronal dynamics, including complex spiking patterns which represent diverse brain activities within the neural networks. Earlier hardware realization of neurons was (a) area intensive because of large capacitors in the circuit design, (b) neuronal spiking patterns were demonstrated with clocked neurons at the device level. To achieve more realistic biological neuron spiking behavior, emerging memristive devices are considered promising alternatives. In this paper, we propose, PrMnO3 (PMO)-resistive random-access memory (RRAM) device-based neuron. The voltage-controlled electrothermal timescales of the compact PMO RRAM device replace the electrical timescales of charging a large capacitor. The electrothermal timescale is used to implement an integration block with multiple voltage-controlled timescales coupled with a refractory block to generate biological neuronal dynamics. Here, first, a Verilog-A implementation of the thermal device model is demonstrated, which captures the current-temperature dynamics of the PMO device. Second, a driving circuitry is designed to mimic different spiking patterns of cortical neurons, including intrinsic bursting and chattering. Third, a neuron circuit model is simulated, which includes the PMO RRAM device model and the driving circuitry to demonstrate the asynchronous neuron behavior. Finally, a hardware-software hybrid analysis is done in which the PMO RRAM device is experimentally characterized to mimic neuron spiking dynamics. The work presents a realizable and more biologically comparable hardware-efficient solution for large-scale SNNs.

Active intrinsic conductances in recurrent networks allow for long-lasting transients and sustained activity with realistic firing rates as well as robust plasticity.

Recurrent neural networks of spiking neurons can exhibit long lasting and even persistent activity. Such networks are often not robust and exhibit spike and firing rate statistics that are inconsistent with experimental observations. In order to overcome this problem most previous models had to assume that recurrent connections are dominated by slower NMDA type excitatory receptors. Usually, the single neurons within these networks are very simple leaky integrate and fire neurons or other low dimensional model neurons. However real neurons are much more complex, and exhibit a plethora of active conductances which are recruited both at the sub and supra threshold regimes. Here we show that by including a small number of additional active conductances we can produce recurrent networks that are both more robust and exhibit firing-rate statistics that are more consistent with experimental results. We show that this holds both for bi-stable recurrent networks, which are thought to underlie working memory and for slowly decaying networks which might underlie the estimation of interval timing. We also show that by including these conductances, such networks can be trained to using a simple learning rule to predict temporal intervals that are an order of magnitude larger than those that can be trained in networks of leaky integrate and fire neurons.

Multibranch Formal Neuron: An Internally Nonlinear Learning Unit.

The transformation of synaptic input into action potential in nerve cells is strongly influenced by the morphology of the dendritic arbor as well as the synaptic efficacy map. The multiplicity of dendritic branches strikingly enables a single cell to act as a highly nonlinear processing element. Studies have also found functional synaptic clustering whereby synapses that encode a common sensory feature are spatially clustered together on the branches. Motivated by these findings, here we introduce a multibranch formal model of the neuron that can integrate synaptic inputs nonlinearly through collective action of its dendritic branches and yields synaptic clustering. An analysis in support of its use as a computational building block is offered. Also offered is an accompanying gradient descent-based learning algorithm. The model unit spans a wide spectrum of nonlinearities, including the parity problem, and can outperform the multilayer perceptron in generalizing to unseen data. The occurrence of synaptic clustering boosts the generalization efficiency of the unit, which may also be the answer for the puzzling ubiquity of synaptic clustering in the real neurons. Our theoretical analysis is backed up by simulations. The study could pave the way to new artificial neural networks.

Neural Network Based on the Special Potential Function for the Recognition of the Multifunctional Signal

The article suggests using elements described by a special potential function as elements of a neural network. The article describes the features and characteristics of a potential function that has an output signal close to a real physiological neuron. The description and structure of a neural network is given, the elements of which are described by a special potential function. To study the neural network and configure its elements, a mathematical apparatus of tensor calculus is proposed. The article describes the possibility of using the proposed neural network for recognizing and finding areas with specified parameters on a complex spatial surface. As an example, we describe the procedure for determining an area on a complex surface that has the specified geometric and qualitative characteristics.