摘要
神经系统电刺激已经成为一种日益重要的神经科学机制探索工具和基本神经系统疾病的治疗手段,但基于期望响应的刺激信号获取需要大量的重复性实验.而且生理实验存在伦理性问题,基于纯软件仿真的电生理实验难以复现对应模拟量信号.利用硬件在环平台获取期望响应的最优刺激可解决上述问题,因此本文设计并实现了一个实时动态闭环电生理硬件在环平台,基于平台可以实现神经系统电生理,通过神经调控策略获取刺激信号,进行神经系统的刺激机制探索和基本神经调控手段的优化.本文设计的平台包括硬件回路和图形用户界面,其中硬件回路主要在数字信号处理器上构建,实现了人工神经系统和闭环控制器的片上集成;上位机图形用户界面的开发实现了人机交互,用户可根据需要进行不同模型、控制算法的闭环电生理实验.基于该平台,本文实现了皮层-基底核-丘脑回路神经网络的实时硬件计算,并获得了期望网络状态的刺激信号.结果表明,相比于I5-8400中央处理器的仿真模拟,该实时动态闭环电生理硬件在环平台可将神经元网络状态的高速计算提升近40倍.而且平台实现的迭代学习控制策略可进一步缓解传统比例积分闭环电生理实验中参数整定的难度,有效提高刺激信号的调控精度.
Neurological electrical stimulation has become an increasingly important tool for exploring neuroscience mechanisms and treating basic neurological diseases.However,further development of this tool is encumbered by the need for numerous repetitive experiments to acquire stimulus signals of the desired response.Moreover,ethical issues are a big concern in physiological experiments,and it is difficult to reproduce the corresponding voltage signals in electrophysiological experiments based on software simulation.However,the abovementioned problems can be resolved using a hardware-in-the-loop platform to obtain the desired optimal stimulation response.This study designed and implemented a real-time dynamic closed-loop electrophysiological hardware-in-the-loop platform to realize nervous system electrophysiology,acquire desired stimulation signals,and perform basic neural control strategy optimization.The platform features a hardware circuit mainly constructed around digital signal processing to enable the integration of an artificial neural system and a closed-loop controller.Its graphical user interface realizes human-computer interaction,allowing users to conduct closed-loop electrophysiological experiments involving different models and control algorithms.Moreover,this study realized the real-time hardware computation of the cortical-basal nucleusthalamic circuit neural network and obtained optimal stimuli that can regulate the network state according to expectations.The results show that the real-time dynamic closed-loop electrophysiological hardware can improve the highspeed calculation of the neural network state by approximately 40 times compared with the I5-8400 central processing unit.Finally,the iterative learning control strategy realized on the platform can further alleviate the difficulty of parameter setting in the traditional experiment on proportional integral closed-loop electrophysiological and effectively improve the regulation accuracy of stimulus signals.
作者
王江
刘维桐
常思远
梁家玮
刘晨
Wang Jiang;Liu Weitong;Chang Siyuan;Liang Jiawei;Liu Chen(School of Electrical and Information Engineering,Tianjin University,Tianjin 300072,China)
出处
《天津大学学报（自然科学与工程技术版）》
EI
CAS
CSCD
北大核心
2023年第7期735-744,共10页
Journal of Tianjin University：Science and Technology
基金
国家自然科学基金资助项目(62173241)
天津市自然科学基金资助项目(20JCQNJC01160).
关键词
闭环电生理
实时
硬件在环平台
神经计算模型
closed-loop electrophysiological
real-time
hardware-in-the-loop platform
neural computing model
