Mei Hong Zheng

A model describing collective behaviors of pedestrians with various personalities in danger situations

We present a model describing motions of individual pedestrians to study collective behaviors of pedestrians in various situations such as normal or emergency situations. Each individual pedestrian has ability to decide own desired action depending on circumstances and own personality. In the present work the personality of pedestrian is expressed as patient and impatient character. The main frame of the present model is constructed based on the social force model, and the individual ability to decide own desired action is represented by the counterpropagation neural network. The combination of neural network and dynamic model makes motions of pedestrians more natural and reliable in various situations. It has been shown from the computer simulation of passing of many pedestrians through a road that there exists an optimal ratio of patient persons number to impatient persons number for the passage time and amenity of pedestrians.

Insufficient augmentation of ambient GABA responsible for age-related cognitive deficit

Age-related degeneration of intracortical inhibition could underlie declines in cognitive function during senescence. Based on a hypothesis that a decrease in basal concentration of ambient (extrasynaptic) GABA with aging leads to depressing intracortical inhibition, we investigated how the basal concentration affects stimulus-evoked activity (as signal), ongoing-spontaneous activity (as noise) of neurons and their (signal-to-noise) ratio S/N. We simulated a neural network model equipped with a GABA transport system that regulates ambient GABA concentration in a neuronal activity-dependent manner. An increase in basal concentration augmented ambient GABA, increased GABA-mediated inhibitory current, and depressed ongoing-spontaneous activity while still keeping stimulus-evoked activity. This led to S/N improvement, for which it was necessary for the reversal potential of GABA transporter to be close to the resting potential of neurons. Above the resting potential, ongoing-spontaneous activity was predominantly enhanced due to excessive GABA-uptake from the extracellular space by transporters. Below the resting potential, stimulus-evoked activity was predominantly depressed, caused by excessive GABA-release. We suggest that the insufficient augmentation of ambient GABA due to a decrease in its basal concentration may be one of the possible causes of cognitive deficit with aging, increasing ongoing-spontaneous neuronal activity as noise. GABA transporter may contribute to improving S/N, provided that its reversal potential is close to the resting potential.

A neural network model for encoding and perception of vowel sounds

Abstract By simulating a hierarchical neural network model, we investigated neuronal bases for encoding and perception of vowel sounds. The lower network detects spectral peaks called formant frequencies of a vowel sound. The higher network detects the combinatory information of the first (F1) and second (F2) formant frequencies. We trained the model with five Japanese vowels spoken by different people and modified synaptic connections according to the Hebbian rule. The present model can recognize not only learned vowel sounds but also vowel sounds that model hears for the first time. We suggest that an ‘unknown’ vowel sound can be perceived if they activate portion of a cell assembly whose population activation encodes information about the category of the unknown vowel sound.

Roles of dynamic linkage of stable attractors across cortical networks in recalling long-term memory.

Abstract. We propose a neural network model for a category-association task. By simulating the model, neuronal relevance of cortical interactions to recalling long-term memory was investigated. The model consists of the left and right hemispheres, each of which has IT (inferotemporal cortex) and PC (prefrontal cortex) networks. Information about visual features and their categories were encoded into point attractors of the IT and PC networks, respectively. In the task, the IT network of the right hemisphere was stimulated with a cue feature. After a delay period, the IT network of the left hemisphere was simultaneously stimulated with the choice feature and an irrelevant feature. The cue and choice features belong to the same category, while the irrelevant feature belongs to another category. To complete the task, the IT network must select the point attractor corresponding to the choice feature. We demonstrate that the top-down pathway (PC-to-IT) triggers the retrieval of long-term memory of the choice feature from the IT, and the bottom-up pathway (IT-to-PC) contributes to the maintenance of the retrieved memory during the delay period. The key mechanism for the retrieval and maintenance of that memory is the dynamic linkage of attractors across separate cortical networks. We show that a single hemisphere is sufficient for the memory retrieval, but it is advantageous to use the two hemispheres because the retrieved memory is thereby retained with greater reliability until the brain chooses the choice feature.

Tonically balancing intracortical excitation and inhibition by gabaergic gliotransmission

For sensory cortices to respond reliably to feature stimuli, the balancing of neuronal excitation and inhibition is crucial. A typical example might be the balancing of phasic excitation within cell assemblies and phasic inhibition between cell assemblies. The former controls the gain of and the latter the tuning of neuronal responses. A change in ambient GABA concentration might affect the dynamic behavior of neurons in a tonic manner. For instance, an increase in ambient GABA concentration enhances the activation of extrasynaptic receptors, augments an inhibitory current, and thus inhibits neurons. When a decrease in ambient GABA concentration occurs, the tonic inhibitory current is reduced, and thus the neurons are relatively excited. We simulated a neural network model in order to examine whether and how such a tonic excitatory-inhibitory mechanism could work for sensory information processing. The network consists of cell assemblies. Each cell assembly, comprising principal cells (P), GABAergic interneurons (Ia, Ib), and glial cells (glia), responds to one particular feature stimulus. GABA transporters, embedded in glial plasma membranes, regulate ambient GABA levels. Hypothetical neuron-glia signaling via inhibitory (Ia-to-glia) and excitatory (P-to-glia) synaptic contacts was assumed. The former let transporters import (remove) GABA from the extracellular space and excited stimulus-relevant P cells. The latter let them export GABA into the extracellular space and inhibited stimulus-irrelevant P cells. The main finding was that the glial membrane transporter gave a combinatorial excitatory-inhibitory effect on P cells in a tonic manner, thereby improving the gain and tuning of neuronal responses. Interestingly, it worked cooperatively with the conventional, phasic excitatory-inhibitory mechanism. We suggest that the GABAergic gliotransmission mechanism may provide balanced intracortical excitation and inhibition so that the best perceptual performance of the cortex can be achieved.

Facilitation of neuronal responses by intrinsic default mode network activity

Default mode network (DMN) shows intrinsic, high-level activity at rest. We tested a hypothesis proposed for its role in sensory information processing: Intrinsic DMN activity facilitates neural responses to sensory input. A neural network model, consisting of a sensory network (Nsen) and a DMN, was simulated. The Nsen contained cell assemblies. Each cell assembly comprised principal cells, GABAergic interneurons (Ia, Ib), and glial cells. We let the Nsen carry out a perceptual task: detection of sensory stimuli. During DMN activation, glial cells were hyperpolarized by Ia-to-glia circuitry, by which glial membrane transporters imported GABA molecules from the extracellular space and decreased ambient GABA concentration. Acting on extrasynaptic GABA receptors, the decrease in ambient GABA concentration reduced inhibitory current in a tonic manner. This depolarized principal cells below their firing threshold during the ongoing spontaneous time period and accelerated their reaction speed to a sensory stimulus. During the stimulus presentation period, the Nsen inhibited the DMN and caused DMN deactivation. The DMN deactivation made Nsen Ia cells cease firing, thereby stopping the glial membrane hyperpolarization, quitting the GABA import, returning to the basal ambient GABA level, and thus enhancing global inhibition. Notably, the stimulus-relevant P cell firing could be maintained when GABAergic gliotransmission via Ia-glia signaling worked, decreasing ambient GABA concentration around the stimulus-relevant P cells. This enabled the Nsen to reliably detect the stimulus. We suggest that intrinsic default model network activity may accelerate the reaction speed of the sensory network by modulating its ongoing-spontaneous activity in a subthreshold manner. Ambient GABA contributes to achieve an optimal ongoing spontaneous subthreshold neuronal state, in which GABAergic gliotransmission triggered by the intrinsic default model network activity may play an important role.

Local intracortical circuitry not only for feature binding but also for rapid neuronal responses

Neurons of primary sensory cortices are known to have specific responsiveness to elemental features. To express more complex sensory attributes that are embedded in objects or events, the brain must integrate them. This is referred to as feature binding and is reflected in correlated neuronal activity. We investigated how local intracortical circuitry modulates ongoing-spontaneous neuronal activity, which would have a great impact on the processing of subsequent combinatorial input, namely, on the correlating (binding) of relevant features. We simulated a functional, minimal neural network model of primary visual cortex, in which lateral excitatory connections were made in a diffusive manner between cell assemblies that function as orientation columns. A pair of bars oriented at specific angles, expressing a visual corner, was applied to the network. The local intracortical circuitry contributed not only to inducing correlated neuronal activation and thus to binding the paired features but also to making membrane potentials oscillate at firing-subthreshold during an ongoing-spontaneous time period. This led to accelerating the reaction speed of principal cells to the input. If the lateral excitatory connections were selectively (instead of “diffusively”) made, hyperpolarization in ongoing membrane potential occurred and thus the reaction speed was decelerated. We suggest that the local intracortical circuitry with diffusive connections between cell assemblies might endow the network with an ongoing subthreshold neuronal state, by which it can send the information about combinations of elemental features rapidly to higher cortical stages for their full and precise analyses.

Intermittent stimuli increase alternation of ambiguous figures

Using the repetitive appearance-disappearance stimuli of the Necker cube, we studied how perceptual alternations between two interpretations differed depending on whether the stimulus was shown intermittently or continuously. We found alternations were faster when stimuli were presented intermittently than when presented continuously...

Reduction of trial-to-trial perceptual variability by intracortical tonic inhibition

Variability is a prominent characteristic of cognitive brain function. For instance, different trials of presentation of the same stimulus yield higher variability in its perception: subjects sometimes fail in perceiving the same stimulus. Perceptual variability could be attributable to ongoing-spontaneous fluctuation in neuronal activity prior to sensory stimulation. Simulating a cortical neural network model, we investigated the underlying neuronal mechanism of perceptual variability in relation to variability in ongoing-spontaneous neuronal activity. In the network model, populations of principal cells (cell assemblies) encode information about sensory features. Each cell assembly is sensitive to one particular feature stimulus. Transporters on GABAergic interneurons regulate ambient GABA concentration in a neuronal activity-dependent manner. Ambient GABA molecules activate extrasynaptic GABA receptors on principal cells and interneurons, and provide them with tonic inhibitory currents. We controlled the variability of ongoing-spontaneous neuronal activity by manipulating the basal level of ambient GABA and assessed the perceptual performance of the network: detection of a feature stimulus. In an erroneous response, stimulus-irrelevant but not stimulus-relevant principal cells were activated, generating trains of action potentials. Perceptual variability, reflected in error rate in detecting the same stimulus that was presented repeatedly to the network, was increased as the variability in ongoing-spontaneous membrane potential among cell assemblies increased. Frequent, transient membrane depolarization below firing threshold was the major cause of the increased neuronal variability, for which a decrease in basal ambient GABA concentration was responsible. We suggest that ambient GABA in the brain may have a role in reducing the variability in ongoing-spontaneous neuronal activity, leading to a decrease in perceptual variability and therefore to reliable sensory perception.

GABA diffusion across neuronal columns for efficient sensory tuning

Synaptic (phasic) lateral inhibition between neuronal columns mediated by GABAergic interneurons is, in general, essential for primary sensory cortices to respond selectively to elemental features. We propose here a neural network model with a nonsynaptic (tonic) lateral inhibitory mechanism. While firing, intrasynaptic GABA molecules spill over into extracellular space and accumulate in neuronal columns. Through accumulation in and diffusion across these columns, a level of ambient (extracellular) GABA changes in a neuronal activity-dependent manner. Ambient GABA molecules act on extrasynaptic receptors and provide neurons with tonic inhibitory currents. We examined whether and how the diffusion of GABA molecules across neuronal columns affects tuning performance of the network to a feature stimulus: selective responsiveness. The GABA diffusion led to reducing ambient GABA in the stimulus-relevant column while augmenting ambient GABA in stimulus-irrelevant columns, thereby improving the tuning performance. The GABA diffusion was effective especially when provided with a broader sensory input. Interestingly, this diffusion-based, nonsynaptic (tonic) lateral inhibitory scheme worked well together with the conventional, synaptic (phasic) lateral inhibitory scheme, enhancing the sensory tuning. We suggest that the nonsynaptic lateral inhibition, mediated through GABA diffusion across neuronal columns, may be beneficial for the cortex to tune to sensory features.