Research paper
Odor pattern recognition of a novel bio-inspired olfactory neural network based on kernel clustering

https://doi.org/10.1016/j.cnsns.2022.106274Get rights and content

Highlights

  • We proposed a novel bio-inspired olfactory neural network model based on a traditional olfactory bulb model composed of mitral cells, granule cells, and periglomerular cells.

  • The results indicated that inhibitory synaptic plasticity could balance excitatory and inhibitory synaptic currents in the olfactory cortex.

  • And our model showed different firing patterns with different odor stimulations, and recognition of pure odors and mixed odors could be realized.

Abstract

Olfactory system is an important component in sensory nervous system. When an olfactory receptor receives odor stimulation, it transfers chemical signals into electrical signals, and delivers to the olfactory bulb, where integrates and codes the olfactory information, further to the cerebral olfactory cortex to generate olfaction. Establishment of a novel bio-inspired olfactory neural network for the study of olfactory information processing is helpful to understand how olfactory system can effectively differentiate different types and concentrations of odor. Based on a traditional olfactory bulb model composed of mitral cells, granule cells, and periglomerular cells, olfactory cortex was presented to establish a novel bio-inspired olfactory neural network model. Meanwhile, inhibitory synaptic plasticity was considered when network was receiving stimulation. The results of simulation indicated that inhibitory synaptic plasticity could balance excitatory and inhibitory synaptic currents in the olfactory cortex with specific firing patterns under odor stimulation. Olfactory cortex showed different firing patterns with different odor stimulations, and also presented similar firing patterns and different firing strengths for the same type of odor at different concentrations. Meanwhile, based on hierarchical clustering and fuzzy clustering, recognition of pure odors and mixed odors could be realized.

Introduction

In neuroscience, there exists various computing models concerning the study of olfaction, aiming to explore potential mechanism of odor information processing during olfaction formation. However, due to the complexity of information exchange and transformation between different olfaction levels [1], the mechanism has been still remained elusive [2], [3]. Olfactory system has been considered as a nervous system to investigate memory formation, target recognition, and information storage, mainly including 3 levels: olfactory epithelium (a specialized epithelial tissue inside the nasal cavity that is involved in smell), olfactory bulb (a neural structure of the vertebrate forebrain involved in olfaction), and olfactory cortex (area of the cerebral cortex that processes information about odors and receives nerve signals from the olfactory bulbs) [4]. Lateral inhibition inside olfactory bulb (i.e., inhibitory effect caused by an interconnection between mitral cells (MC) with inhibitory granule cells (GC) and periglomerular cells (PG)) is considered as an important mechanism of olfactory formation. It is mediated by circuits that involve reciprocal dendrodendritic connections between mitral and granule cells [5], [6], [7], [8]. However, the majority of olfactory models are limited in a single component of olfactory system in a low-level [9], [10], [11], [12], [13], such as olfactory receptor in olfactory epithelium, as well as neural response of olfactory bulb and its subdivision (olfactory glomeruli). Additionally, for the task of odor pattern recognition, coupling effect between the olfactory bulb and olfactory cortex is of great importance. Thus, introduction of olfactory cortex in the model is not only extremely necessary, but also essential to explore neural mechanism of olfactory system. Besides, the majority of current models often abstractly describe components of olfactory system, without biological basis or details, leading to incomplete computing model of olfactory system.

In the present study, in order to explore the recognition effect of the olfactory system on odor pattern, with the help of a traditional olfactory bulb model, a novel bio-inspired olfactory nervous system model was established, including the olfactory bulb composed of MC, GC, and PG, and the olfactory cortex composed of pyramidal cells (Pyr), feedforward cells (FF), and feedback cells (FB) [3], [9], [14], [15]. The amount of neurons was equally scaled based on anatomy [4], [16]. Moreover, it has been indicated that inhibitory spike-timing-dependent plasticity (iSTDP) has a decorrelation effect on the formation of odor expression that is similar with lateral inhibition [17], and synaptic connections between neurons may vary with environmental stimulation [18], [19], [20], [21]. Thus, iSTDP was also taken in the present study into account to explore its effect on the balance of synaptic currents in olfactory network, in addition to its influence on learning of odor stimulation. Furthermore, the purpose of the current study was to establish a complete olfactory system model to realize recognition in different types of odor stimulation without being influenced by changes in odor concentration. Traditional methods often evaluate recognition capability of the olfactory system on odor stimulation via calculating the correlation between neural activity of neurons [22], [23], which can only reveal the similarity between two odors, while cannot classify various odors into different types. Based on the kernel function of neuronal spike trains, similarity and distance of network firing patterns under different odor stimulations were assessed, and pure odor was recognized via hierarchical clustering, while the proportion of odor component in the mixed odor was recognized via fuzzy clustering. The presented model can provide a theoretical basis for higher-level olfactory network, which is helpful to establish a further perfect model of functional olfactory system model to explore potential mechanism of olfaction formation and principle of olfactory coding.

Section snippets

Network topology

The model includes olfactory bulb and olfactory cortex. The olfactory bulb contains MC, PG, and GC. The olfactory cortex contains Pyr, FF, and FB [3], [9], [14], [15]. The network connectivity inside the olfactory bulb is shown in Fig. 1.

In the same olfactory glomeruli (GLO), there is an interconnection between MC and PG. The MC cause an inhibitory effect on the MC in the same GLO via exciting PG (i.e., cyclic inhibition). Between different GLO, MC have also an interconnection with GC in the

Learning effects of iSTDP on the olfactory system

The olfactory system is closely associated with learning and memory capabilities, where synaptic plasticity plays a pivotal role. In addition, investigating synaptic plasticity is helpful to understand the change of firing pattern in the olfactory system under odor stimulation.

In the present study, 3000 ms persistent stimulated-odor was forced on the model. Fig. 3 shows an example of odor input. In this figure, the odor input begins at 500 ms and ends at 2500 ms (two cycles) and the

Conclusions

The olfactory system is an important component in the sensory nervous system. After olfactory receptor senses odor stimulation, it converts chemical signal into electrical signal and transmits to the olfactory bulb. Corresponding spatial-time coding is formed due to the interaction between various neurons in the olfactory bulb. The processed olfactory information is eventually transmitted into the olfactory cortex, generating different olfactory senses [2], [3].

Establishment of a novel

CRediT authorship contribution statement

Xuying Xu: Conceptualization, Writing – review & editing, Supervision. Zhenyu Zhu: Conceptualization, Software, Writing – original draft. Yihong Wang: Methodology, Software, Writing – reviewing, Supervision, Funding acquisition. Rubin Wang: Supervision, Project administration, Funding acquisition. Wanzeng Kong: Investigation. Jianhai Zhang: Investigation.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This study was funded by the National Natural Science Foundation of China (Funding numbers: 11472104, 11872180, 12172132, 11802095, 12072113, 61633010, 61473110) and the Natural Science Foundation of Shanghai, China (No. 19zr1473100).

References (32)

  • WillhiteD.C. et al.

    Viral tracing identifies distributed columnar organization in the olfactory bulb

    Proc Natl Acad Sci

    (2006)
  • HopfieldJ.J.

    Odor space and olfactory processing: Collective algorithms and neural implementation

    Proc Natl Acad Sci

    (1999)
  • LiG. et al.

    A two-layer biophysical model of cholinergic neuromodulation in olfactory bulb

    J Neurosci

    (2013)
  • DavideP. et al.

    Understanding odor information segregation in the olfactory bulb by means of mitral and tufted cells

    PLoS One

    (2014)
  • DavisonA.P. et al.

    Dendrodendritic inhibition and simulated odor responses in a detailed olfactory bulb network model

    J Neurophysiol

    (2003)
  • RubinD.B.

    Dynamical mechanisms of odor processing in olfactory bulb mitral cells

    J Neurophysiol

    (2006)
  • Cited by (0)

    1

    Both authors contribute equally to this work.

    View full text