Specific neural coding of fMRI spiking neural network based on time coding

Abstract

Specific neural coding is the key to achieving advanced cognitive function in a bio-brain, which can form an identifying coding pattern for external stimulation. The performance of specific neural coding depends extremely on brain-like models. However, the bio-interpretability of the topology of a brain-like model is still insufficient. The purpose of this paper is to investigate a more biological interpretative brain-like model verified by the performance of specific neural coding. In this study, we used the topology constrained by human brain functional magnetic resonance imaging (fMRI) to construct a new spiking neural network (SNN) as a brain-like model called fMRI-SNN. In the fMRI-SNN, the nodes are Izhikevich neuron models, and the edges are synaptic plasticity models with time-delay. Then, we investigated the specific neural coding of fMRI-SNN, and discussed its mechanism. Our results indicated that: (i) fMRI-SNN has obvious specific neural coding based on time coding for different external stimulations. Furthermore, our discussion on relevance analysis implies that the intrinsic element of specific neural coding is synaptic plasticity. (ii) The specific neural coding of fMRI-SNN outperforms that of scale-free SNN and small-world SNN. Furthermore, our discussion on dynamic topological characteristics implies that the network topology is an element that impacts the performance level of the specific neural coding. (iii) Taking a speech recognition task as a case study, the performance of fMRI-SNN outperforms that of scale-free SNN and small-world SNN in terms of speech recognition accuracy.