基于不同化学突触神经元网络抗扰特性比较  

Comparative study on the anti-interference characteristics of different chemical synaptic neuron networks

作　　者：宋英健 逯迈 李响 SONG Yingjian;LU Mai;LI Xiang(Key Laboratory of Optoelectronic Technology and Intelligent Control(Ministry of Education),Lanzhou Jiaotong University,Lanzhou 730070,China)

机构地区：[1]兰州交通大学光电技术与智能控制教育部重点实验室,甘肃兰州730070

出　　处：《中国医学物理学杂志》2024年第11期1392-1403,共12页Chinese Journal of Medical Physics

基　　金：国家自然科学基金(51867014);甘肃省教育厅项目(2024CXPT-11)。

摘　　要：目的:基于5类不同的化学突触模型,研究基于Hodgkin-Huxley神经元模型的神经元网络抗扰特性。方法:采用Hodgkin-Huxley神经元模型和化学突触,利用数值模拟的方法构建不同拓扑结构的生物神经元网络,通过构建不同化学突触神经元网络对比研究了在叠加一定高斯白噪声的正弦波信号刺激下神经元抗扰的特异性,分析不同刺激信号下神经元网络的信号传递模式。采用相关系数法计算网络中神经元10在两种刺激信息下第0~50 ms的相关系数,进行定量分析。结果:神经元网络的抗扰特性与化学突触模型具有相关性,基于不同化学突触模型搭建的神经网络抗扰波形不同。第1~5类化学突触搭建的网络在两种刺激信号下,神经元10前50 ms相关系数分别为0.141 2、0.914 5、0.996 5、0.567 3、0.881 6。由此可见,第2、第3类化学突触相比于其他几类化学突触有更良好的抗干扰能力;第2、第3类化学突触模型最为复杂;第4、第5类化学突触延时反应最快。结论:第2、第3类化学突触模型最适合研究神经网络抗扰特性,此研究为构建神经元网络过程中选择化学突触提供了一定程度的启示。Objective Based on 5 different chemical synaptic models,the anti-interference characteristics of neural network based on Hodgkin-Huxley neuron model are compared and studied.Methods The Hodgkin-Huxley neuron model and chemical synapses were used to construct biological neuron networks with different topologies by numerical simulation.By constructing different chemical synaptic neuron networks,the specificity of neuron anti-interference under the stimulation of sine wave signal superimposed with certain Gaussian white noise was studied,and the signal transmission mode of neuron networks under different stimulation signals was analyzed.The correlation coefficient method was used to calculate the correlation coefficient of the 0-50 ms of neuron 10 in the network under two kinds of stimulation signals,and the quantitative analysis was carried out.Results The anti-interference characteristics of the neural network were related to the chemical synaptic model,and the anti-interference waveforms of neural networks constructed based on different chemical synaptic models were different.Under the two stimulation signals,the correlation coefficients of the first 50 ms of neuron 10 in the network constructed by the 5 types of chemical synapses were 0.141 2,0.914 5,0.996 5,0.567 3 and 0.881 6,respectively.The types 2 and 3 chemical synapses exhibited better anti-interference ability than the other types;the types 2 and 3 chemical synapse model were the most complex;and the types 4 and 5 chemical synapses had the fastest delayed response.Conclusion The types 2 and 3 chemical synapse models are most suitable for studying the anti-interference characteristics of neural networks.This study provides some insights into the selection of chemical synapses in the construction of neuron networks.

关 键 词：Hodgkin-Huxley模型 神经元网络 数值模拟 化学突触 抗扰特性 

分 类 号：R318[医药卫生—生物医学工程] Q811.2[医药卫生—基础医学]