| Approximately fifty years ago,researchers in the field of neuroscience made a significant discovery regarding the neural basis of working memory.It was determined that sustained spikes in neural activity were a hallmark of this cognitive function,which had been observed across a diverse range of behavioral contexts and in various brain regions.In order to further study the mechanisms underlying working memory,scientists have turned to models of reverberating activity in cultured hippocampus networks,which serve as simplified representations of this complex phenomenon.The development of these theoretical models has facilitated a deeper understanding of the dynamics of working memory and enabled the conduct of more comprehensive,quantitative,and systematic investigations into the mechanisms of learning and working memory.In this study,a comprehensive review of the existing literature on both the biological mechanisms of working memory and mathematical modeling is presented in the introduction,along with an assessment of the limitations of current work in these areas.The application of neural culture techniques and numerical simulation as fundamental tools in this field is also discussed.In the methodology part,this paper discussed relevant modeling studies,and developed a differential equation-based model of reverberation neural networks based on our understanding of biological systems,which contains the dynamics of the cell body of neurons,the activity inside the synapse and synaptic plasticity effect.In contrast to existing models,our approach integrates NMDA currents and takes into account the STDP plasticity rule.By employing both the Morris-Lecar and Izhikevich models for simulation,we enhance the generality and applicability of our findings.We also established a statistical method to quantify the spatiotemporal conservatism of network activity by which we analysis the experiment dataset.For the acquisition of experimental data,electrophysiological techniques were applied to record reverberation activity,based on which we completed our reverberation model.In numerical simulation,our model showed the same active phenomenon in biological systems,and we clarify the mechanism of generation,maintenance and cessation of reverberation activity by our model.Through numerical simulations of different neuron models,It was found that the occurrence of reverberation activity is highly dependent on the dynamic properties of neuronal model.It was also revealing that the upper limit of neuron firing rate is necessary for the generation of reverberation activity.Furthermore,we found that spatiotemporal conservation of reverberation activity is also related to the network topology and synaptic strength distribution.Moreover,in biological experiments,some nascent circuits where reverberation were unable to occur could be transformed into mature circuits in which reverberation can occur with spatiotemporal firing sequences through specific electrical stimulation.In numerical simulation,through a combination of paired-pulse stimulation(PPS)and Spike-timing-dependent plasticity(STDP)training,the model showed the same maturation process.One of the most typical feature of reverberation is the spatiotemporal firing pattern,however the mechanism of it still unclear.We began with topology structure of the network and tested various model to dissect how the conservatism formation.Based on the finding,a workflow to train the network with STDP were proposed,which allows the network ’remember’ specific firing sequence.In addition,we also modeled the biphasic cholinergic modulation,and explained the modulation effects by acetylcholine at different concentrations by the model.In conclusion,this study sheds light on the basic features of reverberation.The occurrence of reverberation is determined by neuron firing rate upper limitation,at the same time the maintenance and cessation are related to the vesicle recycle situation.The spatiotemporal conservation is the result of a combination of network topology,strength of synaptic connections and asynchronous transmitter release.Our model also supports the idea that the regulation of acetylcholine is achieved mainly through the sensitivity of postsynaptic excitatory currents to neuronal cell membranes.This research investigates the reverberation activity systematically,reveals the mechanism of spatiotemporal conserved formation,and explores the effect of STDP learning.This work is enlightening for understanding the learning memory mechanism of real biological systems and developing brain-inspire computing. |
