润色论文网提供毕业论文写作、职称论文写作、职称论文发表及期刊论文发表服务！QQ：542793091

摘要：全麻意识状态的监测是临床手术中的重要环节。但全麻机制尚未完全清楚，意识状态的监测在不同全麻药物中尚未建立统一的监测指标。利用高密度脑电(EEG)信号的功能连接分析，研究全麻意识状态具有重要意义。

基于EGI公司256导高密度静息态脑电数据，本文分别采用相干性、稀疏表达以及偏最小二乘等功能连接计算方法，分析了三组全麻药物(丙泊酚、氯胺酮、七氟烷)作用下的脑功能连接模式变化情况。主要内容包括：

本文采集了24位肺占位切除手术患者(每组麻醉药各8位患者)在清醒状态和麻醉状态下各5分钟的静息态EEG数据。首先，预处理三组全麻高密度脑电数据，包括滤波、坏导替换、伪迹去除等步骤；为保留全脑EEG信号特征并提高运算效率，选取70导脑电数据，并使用功率谱分析获取频域特征。然后，采用稀疏表达算法、偏最小二乘算法和相干分析算法分别计算连接矩阵；定性分析直观的连接热度图；构建连接网络，分析连接网络图；计算平均聚类系数(ACC)、平均最短路径长度(ASPL)等复杂网络分析指标，定量分析不同意识状态下连接性变化情况。

研究结果表明：1、稀疏表达与偏最小二乘分析中无需考虑容积导体效应消除的问题，而相干性分析受容积导体干扰较大。2、稀疏表达算法和偏最小二乘算法构建连接网络图时，连接值可自适应选取；而相干分析算法要通过遍历搜索选择阈值。3、稀疏表达分析与偏最小二乘分析在全脑网络分析中的ASPL指标能有效区分清醒状态与麻醉状态(P<0.05)，而相干分析在全脑网络中难以获得一致的ASPL(P>0.05)。4、稀疏表达在不同麻醉组的全脑网络分析中，取得一致性稳定的结论：麻醉状态下的ASPL相比清醒状态下增加了约1.5，其中，丙泊酚组增加1.612、氯胺酮组增加1.368、七氟烷组增加1.767。

基于稀疏表达计算的功能连接分析在区分麻醉诱导的意识丧失和清醒状态方面具有更好的性能。研究结果为临床全麻意识状态监测奠定了一定理论基础。

The monitoring of general anesthesia is animportant part of clinical surgery. However, the general anesthesia mechanismis not fully understood, and the monitoring of consciousness has not yetestablished a unified monitoring indicator for different general anesthetics.It is of great significance to study the state of general anesthesia by usingthe functional connection analysis of high-density brain Electroencephalogram(EEG) signals.

Based on EGI's 256-channel high-densityresting-state EEG data. The changes of brain functional connectivity patternsunder three groups of general anesthetics (propofol, ketamine, sevoflurane)were analyzed by using coherence, sparse expression and partial least squaresfunctional connectivity methods. The main contents include:

The EEG data of awake state and anesthesiadata, 5 minutes each, were obtained for each of the 24 lung-occupied resectionpatients (8 patients in each group of anesthetics). Firstly, three groups ofhigh-density EEG data were preprocessed, including filtering, bad conductorreplacement and artifact removal. In order to preserve EEG signalcharacteristics and improve operation efficiency, 70 EEG data were selected andfrequency domain characteristics were obtained by power spectrum analysis.Then, sparse expression algorithm, partial least squares algorithm andcoherence analysis algorithm are used to calculate connection matrix,qualitative analysis of intuitive connection heat map, construction ofconnection network, analysis of connection network graph, calculation ofaverage clustering coefficient (ACC), average shortest path length (ASPL) andother complex network analysis indicators, quantitative analysis ofconnectivity changes in different states of consciousness.

The results show that: 1. The problem ofelimination of volume conductor effect is not considered in sparse expressionand partial least squares analysis, and coherence analysis is greatlyinterfered by volume conductor. 2. When the sparse expression algorithm and thepartial least squares algorithm are used to construct the connection networkgraph, the connection values can be adaptively selected; and the coherentanalysis algorithm selects the threshold by traversing the search. 3. Sparseexpression analysis and partial least squares analysis the ASPL index in wholebrain network analysis can effectively distinguish between awake state andanesthesia state (P<0.05), while coherent analysis is difficult to obtainconsistent ASPL in whole brain network (P> 0.05). 4. Sparse expression inthe whole brain network analysis of different anesthesia groups was consistent:the ASPL under anesthesia increased by about 1.5 compared with the awake state,of which, the propofol group increased by 1.612, the ketamine group increasedby 1.368, and the sevoflurane group increased by 1.767.

Functional connectivity analysis based onsparse expression calculations has better performance in distinguishing betweenanesthesia-induced loss of consciousness and awake state. The research resultslaid a theoretical foundation for the monitoring of clinical anesthesia consciousness.

关键词：全麻静息脑电；功能连接分析；偏最小二乘法；稀疏表达计算；功能网络分析

resting EEG of general anesthesia;functional connectivity analysis; partial least squares method; sparseexpression calculation; functional network analysi