Real Neurons Research Articles

Functional Connectivity Analysis of Neuronal Interactions

Advancements in Computational Neuroscience have resulted in the construction of physiologically realistic and computationally efficient neuronal networks. The developments of diverse neuronal models at various implementation scales have caused an affirmative leap in understanding the numerous causes of brain disorders and anomalies, yet the treatment of the brain is regarded as complicated. Observing the responses of a network of realistic neurons constructed using compartmental modelling makes it possible to overcome the shortcomings of descriptive models, such as their inability to highlight synaptic behaviour and their lack of biological realism in spiking behaviour, which in fact play a crucial role in understanding various cognitive mechanisms via the computational unit called neurons. Henceforth, the dynamics of a microscopic level of interacting neurons coupled hierarchically via feedback and feed forward connections are examined in this article. A tiny network of two and four coupled Hodgkin Huxley (HH) neurons integrated into Freeman KIII set topology have been built using the Spike Timing Dependent plasticity(STDP) algorithm. The entire network has been modelled in Neuron modelling environment using Python as the programming interface, and the functional connectivity between the layers has been analysed.

Multi-physics modeling and finite-element formulation of neuronal dendrite growth with electrical polarization

The neuron serves as the basic computational unit for the brain. Altered neuronal morphologies are usually found in various neurological diseases, such as Down syndrome, Williams syndrome, and idiopathic autism. Compelling biological evidence demonstrates that neuronal morphology can be dynamically regulated by neuronal activity through the mediation of calcium signaling pathways. Moreover, studies have revealed that exposure to an applied electric field can induce directional migration of neurites toward the cathode. In this study, we developed a coupled system that combines an advective Gray–Scott model with Gauss’s law to gain a better understanding of dendrite growth and response to electrical polarization. Our simulation results successfully capture key features such as dendrite branching, space-filling, self-avoidance, and electrical polarization. With the help of the convolutional neural network, we inversely identified model parameters of real dendrite morphologies from an online open source. Finally, we calibrated our model using experimental data on growing neurons under applied electric fields.Statement of Significance: The work sheds light on the underlying mechanisms that govern the growth of neuronal dendrites under electrical polarization via mathematical modeling and numerical simulations. We also use a machine-learning technique to calibrate the model against real neuron images. Our numerical implementations and machine-learning pipeline provided online would benefit researchers in understanding the development of various abnormal neuronal morphologies and related neurological diseases.

Reconstruction of Adaptive Leaky Integrate-and-Fire Neuron to Enhance the Spiking Neural Networks Performance by Establishing Complex Dynamics.

Since digital spiking signals can carry rich information and propagate with low computational consumption, spiking neural networks (SNNs) have received great attention from neuroscientists and are regarded as the future development object of neural networks. However, generating the appropriate spiking signals remains challenging, which is related to the dynamics property of neurons. Most existing studies imitate the biological neurons based on the correlation of synaptic input and output, but these models have only one time constant, thus ignoring the structural differentiation and versatility in biological neurons. In this article, we propose the reconstruction of adaptive leaky integrate-and-fire (R-ALIF) neuron to perform complex behaviors similar to real neurons. First, a synaptic cleft time constant is introduced into the membrane voltage charging equation to distinguish the leakage degree between the neuron membrane and the synaptic cleft, which can expand the representation space of spiking neurons to facilitate SNNs to obtain better information expression way. Second, R-ALIF constructs a voltage threshold adjustment equation to balance the firing rate of output signals. Third, three time constants are transformed into learnable parameters, enabling the adaptive adjustment of dynamics equation and enhancing the information expression ability of SNNs. Fourth, the computational graph of R-ALIF is optimized to improve the performance of SNNs. Moreover, we adopt a temporal dropout (TemDrop) method to solve the overfitting problem in SNNs and propose a data augmentation method for neuromorphic datasets. Finally, we evaluate our method on CIFAR10-DVS, ASL-DVS, and CIFAR-100, and achieve top1 accuracy of 81.0% , 99.8% , and 67.83% , respectively, with few time steps. We believe that our method will further promote the development of SNNs trained by spatiotemporal backpropagation (STBP).

Responses of model cortical neurons to temporal interference stimulation and related transcranial alternating current stimulation modalities

Objective. Temporal interference stimulation (TIS) was proposed as a non-invasive, focal, and steerable deep brain stimulation method. However, the mechanisms underlying experimentally-observed suprathreshold TIS effects are unknown, and prior simulation studies had limitations in the representations of the TIS electric field (E-field) and cerebral neurons. We examined the E-field and neural response characteristics for TIS and related transcranial alternating current stimulation modalities. Approach. Using the uniform-field approximation, we simulated a range of stimulation parameters in biophysically realistic model cortical neurons, including different orientations, frequencies, amplitude ratios, amplitude modulation, and phase difference of the E-fields, and obtained thresholds for both activation and conduction block. Main results. For two E-fields with similar amplitudes (representative of E-field distributions at the target region), TIS generated an amplitude-modulated (AM) total E-field. Due to the phase difference of the individual E-fields, the total TIS E-field vector also exhibited rotation where the orientations of the two E-fields were not aligned (generally also at the target region). TIS activation thresholds (75–230 V m−1) were similar to those of high-frequency stimulation with or without modulation and/or rotation. For E-field dominated by the high-frequency carrier and with minimal amplitude modulation and/or rotation (typically outside the target region), TIS was less effective at activation and more effective at block. Unlike AM high-frequency stimulation, TIS generated conduction block with some orientations and amplitude ratios of individual E-fields at very high amplitudes of the total E-field (>1700 V m−1). Significance. The complex 3D properties of the TIS E-fields should be accounted for in computational and experimental studies. The mechanisms of suprathreshold cortical TIS appear to involve neural activity block and periodic activation or onset response, consistent with computational studies of peripheral axons. These phenomena occur at E-field strengths too high to be delivered tolerably through scalp electrodes and may inhibit endogenous activity in off-target regions, suggesting limited significance of suprathreshold TIS.

Nano-scale simulation of neuronal damage by galactic cosmic rays

The effects of realistic, deep space radiation environments on neuronal function remain largely unexplored. In silico modeling studies of radiation-induced neuronal damage provide important quantitative information about physico-chemical processes that are not directly accessible through radiobiological experiments. Here, we present the first nano-scale computational analysis of broad-spectrum galactic cosmic ray irradiation in a realistic neuron geometry. We constructed thousands of in silico realizations of a CA1 pyramidal neuron, each with over 3500 stochastically generated dendritic spines. We simulated the entire 33 ion-energy beam spectrum currently in use at the NASA Space Radiation Laboratory galactic cosmic ray simulator (GCRSim) using the TOol for PArticle Simulation (TOPAS) and TOPAS-nBio Monte Carlo-based track structure simulation toolkits. We then assessed the resulting nano-scale dosimetry, physics processes, and fluence patterns. Additional comparisons were made to a simplified 6 ion-energy spectrum (SimGCRSim) also used in NASA experiments. For a neuronal absorbed dose of 0.5 Gy GCRSim, we report an average of 250 ± 10 ionizations per micrometer of dendritic length, and an additional 50 ± 10, 7 ± 2, and 4 ± 2 ionizations per mushroom, thin, and stubby spine, respectively. We show that neuronal energy deposition by proton and -particle tracks declines approximately hyperbolically with increasing primary particle energy at mission-relevant energies. We demonstrate an inverted exponential relationship between dendritic segment irradiation probability and neuronal absorbed dose for each ion-energy beam. We also find that there are no significant differences in the average physical responses between the GCRSim and SimGCRSim spectra. To our knowledge, this is the first nano-scale simulation study of a realistic neuron geometry using the GCRSim and SimGCRSim spectra. These results may be used as inputs to theoretical models, aid in the interpretation of experimental results, and help guide future study designs.

Hypergraph-based spiking neural P systems for predicting the overall survival time of glioblastoma patients

Spiking neural P (SN P) systems are membrane computing models inspired by the information interaction of spikes among neurons. Although real neurons have complex structures, classical SN P systems are two-dimensional graph structures. Neurons can only communicate in plane, which limits the learning ability of SN P systems in solving practical problems. To solve this issue, we propose in this paper hypergraph-based SN P (HSN P) systems containing three new classes of neurons to describe higher-order relationships among neurons. Three new kinds of rules among neurons are also designed to expand the model into planar, hierarchical and transmembrane computations. Based on the hypergraph-based spiking neural P systems, a new model for predicting the overall survival (OS) time of glioblastoma (GBM) patients is developed. The proposed model is evaluated on GBM cohorts from The Cancer Genome Atlas (TCGA-GBM). The HSN P system achieves good performance compared to the six state-of-the-art methods, thereby verifying the effectiveness of the model in predicting the OS time of GBM patients.

Model simulations unveil the structure-function-dynamics relationship of the cerebellar cortical microcircuit

The cerebellar network is renowned for its regular architecture that has inspired foundational computational theories. However, the relationship between circuit structure, function and dynamics remains elusive. To tackle the issue, we developed an advanced computational modeling framework that allows us to reconstruct and simulate the structure and function of the mouse cerebellar cortex using morphologically realistic multi-compartmental neuron models. The cerebellar connectome is generated through appropriate connection rules, unifying a collection of scattered experimental data into a coherent construct and providing a new model-based ground-truth about circuit organization. Naturalistic background and sensory-burst stimulation are used for functional validation against recordings in vivo, monitoring the impact of cellular mechanisms on signal propagation, inhibitory control, and long-term synaptic plasticity. Our simulations show how mossy fibers entrain the local neuronal microcircuit, boosting the formation of columns of activity travelling from the granular to the molecular layer providing a new resource for the investigation of local microcircuit computation and of the neural correlates of behavior.

Ideal current dipoles are appropriate source representations for simulating neurons for intracranial recordings

ObjectiveTo determine whether dipoles are an appropriate simplified representation of neural sources for stereo-EEG (sEEG). MethodsWe compared the distributions of voltages generated by a dipole, biophysically realistic cortical neuron models, and extended regions of cortex to determine how well a dipole represented neural sources at different spatial scales and at electrode to neuron distances relevant for sEEG. We also quantified errors introduced by the dipole approximation of neural sources in sEEG source localization using standardized low-resolution electrotomography (sLORETA). ResultsFor pyramidal neurons, the coefficient of correlation between voltages generated by a dipole and neuron model were > 0.9 for distances > 1 mm. For small regions of cortex (∼0.1 cm2), the error in voltages between a dipole and region was < 100 µV for all distances. However, larger regions of active cortex (>5 cm2) yielded > 50 µV errors within 1.5 cm of an electrode when compared to single dipoles. Finally, source localization errors were < 5 mm when using dipoles to represent realistic neural sources. ConclusionsSingle dipoles are an appropriate source model to represent both single neurons and small regions of active cortex, while multiple dipoles are required to represent large regions of cortex. SignificanceDipoles are computationally tractable and valid source models for sEEG.

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.

The Butterfly Effect in Primary Visual Cortex

Exploring and establishing artificial neural networks with electrophysiological characteristics and high computational efficiency is a popular topic in the field of computer vision. Inspired by the working mechanism of primary visual cortex, pulse-coupled neural network (PCNN) can exhibit the characteristics of synchronous oscillation, refractory period, and exponential decay. However, electrophysiological evidence shows that the neurons exhibit highly complex non-linear dynamics when stimulated by external periodic signals. This chaos phenomenon, also known as the " butterfly effect", cannot be explained by all PCNN models. In this work, we analyze the main obstacle preventing PCNN models from imitating real primary visual cortex. We consider neuronal excitation as a stochastic process. We then propose a novel neural network, called continuous-coupled neural network (CCNN). Theoretical analysis indicates that the dynamic behavior of CCNN is distinct from PCNN. Numerical results show that the CCNN model exhibits periodic behavior under DC stimulus, and exhibits chaotic behavior under AC stimulus, which is consistent with the results of real neurons. Furthermore, the image and video processing mechanisms of the CCNN model are analyzed. Experimental results on image segmentation indicate that the CCNN model has better performance than the state-of-the-art of visual cortex neural network models.

Transport between im/mobile fractions shapes the speed and profile of cargo distribution in neurons

Neuronal function requires continuous distribution of ion channels and other proteins throughout large cell morphologies. Protein distribution is complicated by immobilization of freely diffusing subunits such as on lipid rafts or in postsynaptic densities. Here, we infer rates of immobilization for the voltage-gated potassium channel Kv4.2. Fluorescence recovery after photobleaching quantifies protein diffusion kinetics, typically reported as a recovery rate and mobile fraction. We show that, implicit in the fluorescence recovery, are rates of particle transfer between mobile and immobile fractions (im/mobilization). We performed photobleaching of fluorescein-tagged ion channel Kv4.2-sGFP2 in over 450 dendrites of rat hippocampal cells. Using mass-action models, we infer rates of Kv4.2-sGFP2 im/mobilization. Using a realistic neuron morphology, we show how these rates shape the speed and profile of subunit distribution. The experimental protocol and model inference introduced here is widely applicable to other cargo and experimental systems.

Spintronic Integrate-Fire-Reset Neuron with Stochasticity for Neuromorphic Computing.

Spintronics has been recently extended to neuromorphic computing because of its energy efficiency and scalability. However, a biorealistic spintronic neuron with probabilistic "spiking" and a spontaneous reset functionality has not been demonstrated yet. Here, we propose a biorealistic spintronic neuron device based on the heavy metal (HM)/ferromagnet (FM)/antiferromagnet (AFM) spin-orbit torque (SOT) heterostructure. The spintronic neuron can autoreset itself after firing due to the exchange bias of the AFM. The firing process is inherently stochastic because of the competition between the SOT and AFM pinning effects. We also implement a restricted Boltzmann machine (RBM) and stochastic integration multilayer perceptron (SI-MLP) using our proposed neuron. Despite the bit-width limitation, the proposed spintronic model can achieve an accuracy of 97.38% in pattern recognition, which is even higher than the baseline accuracy (96.47%). Our results offer a spintronic device solution to emulate biologically realistic spiking neurons.

Generation and Transmission of Action Potential in Nerve Cells and Neuron Populations Based on the Realistic Hodgkin-Huxley Neuron Model

There are several types of nerve cells in the central nervous system. Thanks to the synaptic connections, these cells form large and complicated networks. However, these cells have a stereotypical electrical activity called action potential (AP) or spike. In this work, the mechanisms of formation of this typical electrical signal and the methods of transferring from one cell to another were investigated using Hodgkin-Huxley neuron model simulations. It has been seen that the formation of AP is based on the principle of "all or nothing" and that ion channel dynamics are critical in the typical form of AP. It has been shown that signal transduction between nerve cells is transmitted by post-synaptic potential and that these signals may be cell depolarizing or polarizing. Finally, it is discussed that these electrical activities are quantities that can be measured at micro and macro levels, and various methods are used for this purpose.

Brain-inspired chaotic backpropagation for MLP

Backpropagation (BP) algorithm is one of the most basic learning algorithms in deep learning. Although BP has been widely used, it still suffers from the problem of easily falling into the local minima due to its gradient dynamics. Inspired by the fact that the learning of real brains may exploit chaotic dynamics, we propose the chaotic backpropagation (CBP) algorithm by integrating the intrinsic chaos of real neurons into BP. By validating on multiple datasets (e.g. cifar10), we show that, for multilayer perception (MLP), CBP has significantly better abilities than those of BP and its variants in terms of optimization and generalization from both computational and theoretical viewpoints. Actually, CBP can be regarded as a general form of BP with global searching ability inspired by the chaotic learning process in the brain. Therefore, CBP not only has the potential of complementing or replacing BP in deep learning practice, but also provides a new way for understanding the learning process of the real brain.

Photochromic and luminescent materials for the development of Chemical Artificial Intelligence

Artificial Intelligence (AI) is an interdisciplinary research line that aims to devise intelligent machines. Chemical Artificial Intelligence (CAI) focuses on designing intelligent chemical systems, primarily in wetware. This work shows that intelligent chemical systems can be implemented by devising neural surrogates that, by interplaying, can give rise to synchronization phenomena analogous to those of real neurons in the brain. When the signals employed in the communication are UV–visible radiations, the artificial neuron models can be implemented through photochromic and luminescent compounds, as this work demonstrates. Furthermore, this article describes how photochromic and luminescent materials contribute to processing Fuzzy logic, which is a good model of human power to compute by using words. Perspectives on how photoswitchable materials will be employed in further developments of CAI are outlined.

Face identity coding in the deep neural network and primate brain

A central challenge in face perception research is to understand how neurons encode face identities. This challenge has not been met largely due to the lack of simultaneous access to the entire face processing neural network and the lack of a comprehensive multifaceted model capable of characterizing a large number of facial features. Here, we addressed this challenge by conducting in silico experiments using a pre-trained face recognition deep neural network (DNN) with a diverse array of stimuli. We identified a subset of DNN units selective to face identities, and these identity-selective units demonstrated generalized discriminability to novel faces. Visualization and manipulation of the network revealed the importance of identity-selective units in face recognition. Importantly, using our monkey and human single-neuron recordings, we directly compared the response of artificial units with real primate neurons to the same stimuli and found that artificial units shared a similar representation of facial features as primate neurons. We also observed a region-based feature coding mechanism in DNN units as in human neurons. Together, by directly linking between artificial and primate neural systems, our results shed light on how the primate brain performs face recognition tasks.

Lag Synchronization of Noisy and Nonnoisy Multiple Neurobiological Coupled FitzHugh-Nagumo Networks with and without Delayed Coupling.

This paper presents a methodology for synchronizing noisy and nonnoisy multiple coupled neurobiological FitzHugh–Nagumo (FHN) drive and slave neural networks with and without delayed coupling, under external electrical stimulation (EES), external disturbance, and variable parameters for each state of both FHN networks. Each network of neurons was configured by considering all aspects of real neurons communications in the brain, i.e., synapse and gap junctions. Novel adaptive control laws were developed and proposed that guarantee the synchronization of FHN neural networks in different configurations. The Lyapunov stability theory was utilized to analytically derive the sufficient conditions that ensure the synchronization of the FHN networks. The effectiveness and robustness of the proposed control laws were shown through different numerical simulations.

A New Family of bioSFQ Logic/Memory Cells

Superconductor electronics (SCE) is competing to become a platform for efficient implementations of neuromorphic computing and deep learning algorithms (DLAs) with projects mostly concentrating on searching for gates that would better mimic behavior of real neurons. In contrast, we believe that most of the required components have already been demonstrated during the long history of SCE, whereas the missing part is how to organize these components to efficiently implement DLAs. We propose a family of logic/memory cells in which stored multi-bit data are encoded by quasi-analog currents or magnetic flux in superconductor loops while transmitted data are encoded as the rate of SFQ pulses. We designed, fabricated, and tested some of the basic cells to demonstrate a proof of concept, e.g., a unipolar and bipolar multipliers based on Josephson junction comparators. We coined the term bioSFQ to clearly connote close but distinguishable relations between the conventional SFQ electronics and its new neuromorphic paradigm.

Bi-directional Control of Synaptic Input Summation and Spike Generation by GABAergic Inputs at the Axon Initial Segment.

Differing from other subtypes of inhibitory interneuron, chandelier or axo-axoniccells form depolarizing GABAergic synapses exclusively onto the axon initial segment (AIS) of targeted pyramidal cells (PCs). However, the debate whether these AIS-GABAergic inputs produce excitation or inhibition in neuronal processing is not resolved. Using realistic NEURON modeling and electrophysiological recording of cortical layer-5 PCs, we quantitatively demonstrate that the onset-timing of AIS-GABAergic input, relative to dendritic excitatory glutamatergic inputs, determines its bi-directional regulation of the efficacy of synaptic integration and spike generation in a PC. More specifically, AIS-GABAergic inputs promote the boosting effect of voltage-activated Na+ channels on summed synaptic excitation when they precede glutamatergic inputs by >15 ms, while for nearly concurrent excitatory inputs, they primarily produce a shunting inhibition at the AIS. Thus, our findings offer an integrative mechanism by which AIS-targeting interneurons exert sophisticated regulation of the input-output function in targeted PCs.

A varying-radius cable equation for the modelling of impulse propagation in excitable fibres.

In this study, we present a varying-radius cable equation for nerve fibres taking into account the varying diameter along the neuronal segments. Finite element neuronal models utilising the classical (fixed-radius) and varying-radius cable formulations were compared using simple and realistic morphologies under intra- and extracellular electrical stimulation protocols. We found that the use of the classical cable equation to model intracellular neural electrical stimulation exhibited an error of 17% in a passive resistive cable model with abrupt change in radius from 1 to 2 μm, when compared to the known analytical solution and varying-radius cable formulation. This error was observed to increase substantially using more realistic neuron morphologies and branching structures. In the case of extracellular stimulation however, the difference between the classical and varying-radius formulations was less pronounced, but we expect this difference will increase under more complex stimulation paradigms such as high-frequency stimulation. We conclude that for computational neuroscience applications, it is essential to use the varying-radius cable equation for accurate prediction of neuronal responses under electrical stimulation.