Fast Spiking Research Articles

Oligodendrocyte–axon metabolic coupling is mediated by extracellular K+ and maintains axonal health

The integrity of myelinated axons relies on homeostatic support from oligodendrocytes (OLs). To determine how OLs detect axonal spiking and how rapid axon–OL metabolic coupling is regulated in the white matter, we studied activity-dependent calcium (Ca2+) and metabolite fluxes in the mouse optic nerve. We show that fast axonal spiking triggers Ca2+ signaling and glycolysis in OLs. OLs detect axonal activity through increases in extracellular potassium (K+) concentrations and activation of Kir4.1 channels, thereby regulating metabolite supply to axons. Both pharmacological inhibition and OL-specific inactivation of Kir4.1 reduce the activity-induced axonal lactate surge. Mice lacking oligodendroglial Kir4.1 exhibit lower resting lactate levels and altered glucose metabolism in axons. These early deficits in axonal energy metabolism are associated with late-onset axonopathy. Our findings reveal that OLs detect fast axonal spiking through K+ signaling, making acute metabolic coupling possible and adjusting the axon–OL metabolic unit to promote axonal health.

Algorithms for Fast Spiking Neural Network Simulation on FPGAs

Algorithms for Fast Spiking Neural Network Simulation on FPGAs

Spike Timing Dependent Gradient for Direct Training of Fast and Efficient Binarized Spiking Neural Networks

Spike Timing Dependent Gradient for Direct Training of Fast and Efficient Binarized Spiking Neural Networks

Fast-SNN: Fast Spiking Neural Network by Converting Quantized ANN.

Spiking neural networks (SNNs) have shown advantages in computation and energy efficiency over traditional artificial neural networks (ANNs) thanks to their event-driven representations. SNNs also replace weight multiplications in ANNs with additions, which are more energy-efficient and less computationally intensive. However, it remains a challenge to train deep SNNs due to the discrete spiking function. A popular approach to circumvent this challenge is ANN-to-SNN conversion. However, due to the quantization error and accumulating error, it often requires lots of time steps (high inference latency) to achieve high performance, which negates SNN's advantages. To this end, this paper proposes Fast-SNN that achieves high performance with low latency. We demonstrate the equivalent mapping between temporal quantization in SNNs and spatial quantization in ANNs, based on which the minimization of the quantization error is transferred to quantized ANN training. With the minimization of the quantization error, we show that the sequential error is the primary cause of the accumulating error, which is addressed by introducing a signed IF neuron model and a layer-wise fine-tuning mechanism. Our method achieves state-of-the-art performance and low latency on various computer vision tasks, including image classification, object detection, and semantic segmentation. Codes are available at: https://github.com/yangfan-hu/Fast-SNN.

Special Memristor and Memristor-Based Compact Neuron Circuit

Implementing neuron circuit using memristor can be more effective thanks to some properties of memristor such as nonlinearity. For this reason, many memristor-based neuron circuits have been found in the literature. This study presents a memristor-based floating neuron circuit that exhibits different types of spikes. In the proposed circuit, two classical memristors and one special memristor are utilized. The special memristor is designed to exhibit various pinched hysteresis loops and plays an important role in producing different spike types. The proposed special memristor can also be controlled using only one voltage source. Using memristors, we implemented a neuron circuit which can produce regular spike, bursting spike and fast spike. Also, neuron circuit has a floating structure, and so there is no restriction on using it with other circuit elements. All simulations are implemented in the SPICE simulation program with TSMC 0.18-[Formula: see text]m CMOS process parameters.

Characterization of extracellular spike waveforms recorded in wallaby primary visual cortex.

Extracellular recordings were made from 642 units in the primary visual cortex (V1) of a highly visual marsupial, the Tammar wallaby. The receptive field (RF) characteristics of the cells were objectively estimated using the non-linear input model (NIM), and these were correlated with spike shapes. We found that wallaby cortical units had 68% regular spiking (RS), 12% fast spiking (FS), 4% triphasic spiking (TS), 5% compound spiking (CS) and 11% positive spiking (PS). RS waveforms are most often associated with recordings from pyramidal or spiny stellate cell bodies, suggesting that recordings from these cell types dominate in the wallaby cortex. In wallaby, 70-80% of FS and RS cells had orientation selective RFs and had evenly distributed linear and nonlinear RFs. We found that 47% of wallaby PS units were non-orientation selective and they were dominated by linear RFs. Previous studies suggest that the PS units represent recordings from the axon terminals of non-orientation selective cells originating in the lateral geniculate nucleus (LGN). If this is also true in wallaby, as strongly suggested by their low response latencies and bursty spiking properties, the results suggest that significantly more neurons in wallaby LGN are already orientation selective. In wallaby, less than 10% of recorded spikes had triphasic (TS) or sluggish compound spiking (CS) waveforms. These units had a mixture of orientation selective and non-oriented properties, and their cellular origins remain difficult to classify.

A Fast and Effective Spike Sorting Method Based on Multi-Frequency Composite Waveform Shapes.

Accurate spike sorting to the appropriate neuron is crucial for neural activity analysis. To improve spike sorting performance, it is essential to fully leverage each processing step, including filtering, spike detection, feature extraction, and clustering. However, compared to the latter two steps that were widely studied and optimized, the filtering process was largely neglected. In this study, we proposed a fast and effective spike sorting method (MultiFq) based on multi-frequency composite waveform shapes acquired through an optimized filtering process. When combined with the classical PCA-Km spiking sorting algorithm, our proposed MultiFq significantly improved its sorting performance and achieved as high performance as the complex Wave-clus did in both the simulated and in vivo datasets. But, the combined method was about 10 times faster than Wave-clus (0.16 s vs. 2.06 s in simulated datasets; 0.46 s vs. 2.03 s in in vivo datasets). Furthermore, we demonstrated the compatibility of our MultiFq by combining it with other sorting algorithms, which consistently resulted in significant improvement in sorting accuracy with the maximum improvement at 35.04%. The above results demonstrated that our proposed method could significantly improve the sorting performance with low computation cost and good compatibility by leveraging the multi-frequency composite waveform shapes.

Routing states transition during oscillatory bursts and attentional selection

Brain-wide information routing relies on the spatio-temporal dynamics of neural activity, but it remains unclear how routing states emerge at fast spiking timescales and relate to slower activity dynamics during cognitive processes. Here, we show that localized spiking events participate in directional routing states with spiking activity in distant brain areas that dynamically switch or amplify states during oscillatory bursts, attentional selection, and decision-making. Modeling and neural recordings from lateral prefrontal cortex (LPFC), anterior cingulate cortex (ACC), and striatum of nonhuman primates revealed that cross-regional routing states arise within 20ms following individual neuron spikes, with LPFC spikes leading the activity in ACC and striatum. The baseline routing state amplified during LPFC beta bursts between LPFC and striatum and switched direction during ACC theta/alpha bursts between ACC and LPFC. Selective attention amplified theta-/alpha-band-specific lead ensembles in ACC, while decision-making increased the lead of ACC and LPFC spikes to the striatum.

Rhythmic oscillations in the midbrain dopaminergic nuclei in mice.

Dopamine release in the forebrain by midbrain ventral tegmental nucleus (VTA) and substantia nigra pars compacta (SNc) neurons is implicated in reward processing, goal-directed learning, and decision-making. Rhythmic oscillations of neural excitability underlie coordination of network processing, and have been reported in these dopaminergic nuclei at several frequency bands. This paper provides a comparative characterization of several frequencies of oscillations of local field potential and single unit activity, highlighting some behavioral correlates. We recorded from optogenetically identified dopaminergic sites in four mice training in operant olfactory and visual discrimination tasks. Rayleigh and Pairwise Phase Consistency (PPC) analyses revealed some VTA/SNc neurons phase-locked to each frequency range, with fast spiking interneurons (FSIs) prevalent at 1-2.5 Hz (slow) and 4 Hz bands, and dopaminergic neurons predominant in the theta band. More FSIs than dopaminergic neurons were phase-locked in the slow and 4 Hz bands during many task events. The highest incidence of phase-locking in neurons was in the slow and 4 Hz bands, and occurred during the delay between the operant choice and trial outcome (reward or punishment) signals. These data provide a basis for further examination of rhythmic coordination of activity of dopaminergic nuclei with other brain structures, and its impact for adaptive behavior.

A superconducting synapse exhibiting spike-timing dependent plasticity

The recent success of artificial intelligence (AI) systems has been accompanied by a rapid increase in the computational resources needed to successfully train them. This rate of increase threatens the future development of AI systems as they are presently configured. Unsupervised learning, where systems are trained online instead of through offline computation, offers a possible way forward. Here, we present the design of a synaptic circuit made from superconducting electronics capable of spike-timing dependent plasticity (STDP), a form of unsupervised learning. The synapse is constructed from three sub-circuits, each responsible for a part of the synaptic action. We demonstrate the operation of the synapse through numerical simulation and show that it reproduces the hallmark behaviors of STDP. Combined with existing superconducting neuromorphic components like neurons and axons, this synaptic structure could help form a fast, powerful, and energy-efficient Spiking Neural Network.

Ultrafast (400 Hz) network oscillations induced in mouse barrel cortex by optogenetic activation of thalamocortical axons.

Oscillations of extracellular voltage, reflecting synchronous, rhythmic activity in large populations of neurons, are a ubiquitous feature in the mammalian brain, and are thought to subserve important, if not fully understood roles in normal and abnormal brain function. Oscillations at different frequency bands are hallmarks of specific brain and behavioral states. At the higher end of the spectrum, 150-200 Hz ripples occur in the hippocampus during slow-wave sleep, and ultrafast (400-600 Hz) oscillations arise in the somatosensory cortices of humans and several other mammalian species in response to peripheral nerve stimulation or punctate sensory stimuli. Here we report that brief optogenetic activation of thalamocortical axons, in brain slices from mouse somatosensory (barrel) cortex, elicited in the thalamorecipient layer local field potential (LFP) oscillations which we dubbed "ripplets". Ripplets originated in the postsynaptic cortical network and consisted of a precisely repeating sequence of 2‑5 negative transients, closely resembling hippocampal ripples but, at ~400 Hz, over twice as fast. Fast-spiking (FS) inhibitory interneurons fired highly synchronous 400 Hz spike bursts entrained to the LFP oscillation, while regular-spiking (RS), excitatory neurons typically fired only 1-2 spikes per ripplet, in antiphase to FS spikes, and received synchronous sequences of alternating excitatory and inhibitory inputs. We suggest that ripplets are an intrinsically generated cortical response to a strong, synchronous thalamocortical volley, and could provide increased bandwidth for encoding and transmitting sensory information. Importantly, optogenetically induced ripplets are a uniquely accessible model system for studying synaptic mechanisms of fast and ultrafast cortical and hippocampal oscillations.

Dopamine depletion leads to pathological synchronization of distinct basal ganglia loops in the beta band.

Motor symptoms of Parkinson's Disease (PD) are associated with dopamine deficits and pathological oscillation of basal ganglia (BG) neurons in the β range ([12-30] Hz). However, how dopamine depletion affects the oscillation dynamics of BG nuclei is still unclear. With a spiking neurons model, we here capture the features of BG nuclei interactions leading to oscillations in dopamine-depleted condition. We highlight that both the loop between subthalamic nucleus (STN) and Globus Pallidus pars externa (GPe) and the loop between striatal fast spiking and medium spiny neurons and GPe display resonances in the β range, and synchronize to a common β frequency through interaction. Crucially, the synchronization depends on dopamine depletion: the two loops are largely independent for high levels of dopamine, but progressively synchronize as dopamine is depleted due to the increased strength of the striatal loop. The model is validated against recent experimental reports on the role of cortical inputs, STN and GPe activity in the generation of β oscillations. Our results highlight the role of the interplay between the GPe-STN and the GPe-striatum loop in generating sustained β oscillations in PD subjects, and explain how this interplay depends on the level of dopamine. This paves the way to the design of therapies specifically addressing the onset of pathological β oscillations.

Reverse engineering information processing in lateral amygdala during auditory tones.

Learning in the mammalian lateral amygdala (LA) during auditory fear conditioning (tone - foot shock pairing), one form of associative learning, requires N-methyl-D-aspartate (NMDA) receptor-dependent plasticity. Despite this fact being known for more than two decades, the biophysical details related to signal flow and the involvement of the coincidence detector, NMDAR, in this learning, remain unclear. Here we use a 4000-neuron computational model of the LA (containing two types of pyramidal cells, types A and C, and two types of interneurons, fast spiking FSI and low-threshold spiking LTS) to reverse engineer changes in information flow in the amygdala that underpin such learning; with a specific focus on the role of the coincidence detector NMDAR. The model also included a Ca2s based learning rule for synaptic plasticity. The physiologically constrained model provides insights into the underlying mechanisms that implement habituation to the tone, including the role of NMDARs in generating network activity which engenders synaptic plasticity in specific afferent synapses. Specifically, model runs revealed that NMDARs in tone-FSI synapses were more important during the spontaneous state, although LTS cells also played a role. Training trails with tone only also suggested long term depression in tone-PN as well as tone-FSI synapses, providing possible hypothesis related to underlying mechanisms that might implement the phenomenon of habituation.

Pharmacogenetic activation of parvalbumin interneurons in the prefrontal cortex rescues cognitive deficits induced by adolescent MK801 administration.

The cognitive symptoms of schizophrenia (SZ) present a significant clinical burden. They are treatment resistant and are the primary predictor of functional outcomes. Although the neural mechanisms underlying these deficits remain unclear, pathological GABAergic signaling likely plays an essential role. Perturbations with parvalbumin (PV)-expressing fast-spiking (FS) interneurons in the prefrontal cortex (PFC) are consistently found in post-mortem studies of patients with SZ, as well as in animal models. Our studies have shown decreased prefrontal synaptic inhibition and PV immunostaining, along with working memory and cognitive flexibility deficits in the MK801 model. To test the hypothesized association between PV cell perturbations and impaired cognition in SZ, we activated prefrontal PV cells by using an excitatory DREADD viral vector with a PV promoter to rescue the cognitive deficits induced by adolescent MK801 administration in female rats. We found that targeted pharmacogenetic upregulation of prefrontal PV interneuron activity can restore E/I balance and improve cognition in the MK801 model. Our findings support the hypothesis that the reduced PV cell activity levels disrupt GABA transmission, resulting in the disinhibition of excitatory pyramidal cells. This disinhibition leads to an elevated prefrontal excitation/inhibition (E/I) balance that could be causal for cognitive impairments. Our study provides novel insights into the causal role of PV cells in cognitive function and has clinical implications for understanding the pathophysiology and management of SZ.

GABAergic synapses onto SST and PV interneurons in the CA1 hippocampal region show cell-specific and integrin-dependent plasticity

It is known that GABAergic transmission onto pyramidal neurons shows different forms of plasticity. However, GABAergic cells innervate also other inhibitory interneurons and plasticity phenomena at these projections remain largely unknown. Several mechanisms underlying plastic changes, both at inhibitory and excitatory synapses, show dependence on integrins, key proteins mediating interaction between intra- and extracellular environment. We thus used hippocampal slices to address the impact of integrins on long-term plasticity of GABAergic synapses on specific inhibitory interneurons (containing parvalbumin, PV + or somatostatin, SST +) known to innervate distinct parts of principal cells. Administration of RGD sequence-containing peptide induced inhibitory long-term potentiation (iLTP) at fast-spiking (FS) PV + as well as on SST + interneurons. Interestingly, treatment with a more specific peptide GA(C)RRETAWA(C)GA (RRETAWA), affecting α5β1 integrins, resulted in iLTP in SST + and iLTD in FS PV + interneurons. Brief exposure to NMDA is known to induce iLTP at GABAergic synapses on pyramidal cells. Intriguingly, application of this protocol for considered interneurons evoked iLTP in SST + and iLTD in PV + interneurons. Moreover, we showed that in SST + cells, NMDA-evoked iLTP depends on the incorporation of GABAA receptors containing α5 subunit to the synapses, and this iLTP is occluded by RRETAWA peptide, indicating a key role of α5β1 integrins. Altogether, our results revealed that plasticity of inhibitory synapses at GABAergic cells shows interneuron-specificity and show differences in the underlying integrin-dependent mechanisms. This is the first evidence that neuronal disinhibition may be a highly plastic process depending on interneuron type and integrins’ activity.

Spectrum of motor responses elicited by electrical stimulation of primary motor cortex: A polygraphic study in patients with epilepsy

ObjectiveTo study the neurophysiology of motor responses elicited by electrical stimulation of the primary motor cortex. MethodsWe studied motor responses in four patients undergoing invasive epilepsy monitoring and functional cortical mapping via electrical cortical stimulation using surface EMG electrodes. In addition, polygraphic analysis of intracranial EEG and EMG during bilateral tonic-clonic seizures, induced by cortical stimulation, was performed in two patients. Results(a) Electrical cortical stimulation: The motor responses were classified as clonic, jittery, and tonic. The clonic responses were characterized by synchronous EMG bursts of agonist and antagonistic muscles, alternating with silent periods. At stimulation frequencies of <20 Hz, EMG bursts were of ≤50 ms duration (Type I clonic). At stimulation frequencies of 20–50 Hz, EMG bursts were of >50 ms duration and had a complex morphology (Type II clonic). Increasing the current intensity at a constant frequency converted clonic responses into jittery and tonic contractions.(b) Bilateral tonic-clonic seizures: The intracranial EEG showed continuous fast spiking activity during the tonic phase along with interference pattern on surface EMG. The clonic phase was characterized by a polyspike-and-slow wave pattern. The polyspikes were time-locked with the synchronous EMG bursts of agonists and antagonists and the slow waves were time-locked with silent periods. InterpretationThese results suggest that epileptic activity involving the primary motor cortex can produce a continuum of motor responses ranging from type I clonic, type II clonic, and tonic responses to bilateral tonic-clonic seizures. This continuum is related to the frequency and intensity of the epileptiform discharges with tonic seizures representing the highest end of the spectrum.

Bayesian prediction of psychophysical detection responses from spike activity in the rat sensorimotor cortex.

Decoding of sensorimotor information is essential for brain-computer interfaces (BCIs) as well as in normal functioning organisms. In this study, Bayesian models were developed for the prediction of binary decisions of 10 awake freely-moving male/female rats based on neural activity in a vibrotactile yes/no detection task. The vibrotactile stimuli were 40-Hz sinusoidal displacements (amplitude: 200 µm, duration: 0.5 s) applied on the glabrous skin. The task was to depress the right lever for stimulus detection and left lever for stimulus-off condition. Spike activity was recorded from 16-channel microwire arrays implanted in the hindlimb representation of primary somatosensory cortex (S1), overlapping also with the associated representation in the primary motor cortex (M1). Single-/multi-unit average spike rate (Rd) within the stimulus analysis window was used as the predictor of the stimulus state and the behavioral response at each trial based on a Bayesian network model. Due to high neural and psychophysical response variability for each rat and also across subjects, mean Rd was not correlated with hit and false alarm rates. Despite the fluctuations in the neural data, the Bayesian model for each rat generated moderately good accuracy (0.60-0.90) and good class prediction scores (recall, precision, F1) and was also tested with subsets of data (e.g. regular vs. fast spike groups). It was generally observed that the models were better for rats with lower psychophysical performance (lower sensitivity index A'). This suggests that Bayesian inference and similar machine learning techniques may be especially helpful during the training phase of BCIs or for rehabilitation with neuroprostheses.

Maturation of GABAergic Synaptic Transmission From Neocortical Parvalbumin Interneurons Involves N-methyl-D-aspartate Receptor Recruitment of Cav2.1 Channels

N-methyl-D-aspartate receptor (NMDAR) hypofunction during brain development is likely to contribute to the manifestation of schizophrenia (SCZ) in young adulthood. The cellular targets of NMDAR hypofunction appear to be at least in part corticolimbic fast-spiking (FS) interneurons. However, functional alterations in parvalbumin (PV)-positive FS interneurons following NMDAR hypofunction are poorly understood. Paired patch-clamp recordings from murine cortical PV interneurons and pyramidal neurons revealed that genetic deletion of NMDAR subunit Grin1 in prospective PV interneurons before the second postnatal week impaired evoked- and synchronized-GABA release. Whereas intrinsic excitability and spiking characteristics were also disturbed by Grin1 deletion, neither restoring their excitability by K+ channel blockade nor increasing extracellular Ca2+ rescued the GABA release. GABA release was also insensitive to the Cav2.1 channel antagonist ω-agatoxin IVA. Heterozygous deletion of Cacna1a gene (encoding Cav2.1) in PV interneurons produced a similar GABA release phenotype as the Grin1 mutants. Treatment with the Cav2.1/2.2 channel agonist GV-58 augmented somatic Ca2+ currents and GABA release in Cacna1a-haploinsufficient PV interneurons, but failed to enhance GABA release in the Grin1-deleted PV interneurons. Taken together, our results suggest that Grin1 deletion in prospective PV interneurons impairs proper maturation of membrane excitability and Cav2.1-recruited evoked GABA release. This may increase synaptic excitatory/inhibitory ratio in principal neurons, contributing to the emergence of SCZ-like phenotypes.

Rich spike patterns from a periodically forced Izhikevich neuron model

Physiological experiments have described the electrophysiological properties of a single neuron, and mathematical models formulating neuronal activity have been proposed to elucidate the mechanism of information processing by the brain. In the present study, we investigated one such model, the Izhikevich neuron model, stimulated by sinusoidal inputs, with the parameter sets of four principal neuron types: regular spiking, fast spiking, intrinsically bursting, and chattering neurons. We adopted three measures: the diversity index, the coefficient of variation, and the local variation, to quantify interspike intervals from different viewpoints. The combined evaluation of these three measures clarified that the positional relationship of the nullclines, which is determined by the amplitude of sinusoidal forcing, plays a crucial role in the intrinsic properties of a periodically forced Izhikevich neuron model. Moreover, we used stroboscopic plots to clarify qualitative differences between attractors. The results also imply that such combined evaluation is applicable to the classification of neurons.

A Novel and Efficient Surface Electromyography Decomposition Algorithm Using Local Spatial Information.

Motor unit spike trains (MUSTs) decomposed from surface electromyography (sEMG) have been an emerging solution for neural interfacing, especially for the control of upper limb prosthetics. Accurate and efficient decomposition techniques are essential and desirable. However, most decomposition methods are designed for motor units (MUs) with global maximum of single or large muscle, while in general forearm muscles are usually small and slender with low global energy. Thus, we propose a novel approach using local spatial information towards more accurate and efficient sEMG decomposition of forearm muscles. A fast spatial spike detection method is proposed to replace the time-consuming iteration process of blind source separation (BSS) methods. Here, spatial distribution characteristics of motor unit action potential are leveraged to pre-classify the candidate MUs, and further to create initial MU templates, aiming to avoid repeating convergence to high-energy MUs. The results of both simulated and experimental sEMG signals show that low-energy MUs from small muscles are more easily found compared with conventional BSS algorithm. Specifically, the proposed method can identify more 40% reliable MUs while only 30% consuming time are needed. The outcomes provide a novel solution for more efficient sEMG decomposition, potentially paving the way of MUST-based non-invasive neural interface.