【作者】 魏勤；

【导师】 刘泉；

【作者基本信息】 武汉理工大学 ， 通信与信息系统， 2012， 博士

【摘要】 麻醉是临床手术中不可或缺的关键环节,如何保证病人在手术过程中安全和无痛苦是麻醉工作的核心问题。在复杂多变的手术过程中,为了确保麻醉安全,麻醉医生必须长时间注意力高度集中地全面性地观察记录病患各种生理特征,并根据自身经验进行分析和判别病患术中的麻醉深度。然而,由医生的主观判断来评估麻醉深度,容易出现因为所获信息与个人经验的不足、身体的疲累与环境的干扰、潜在的因素和病患个体性差异而造成的误判。随着生物医学工程与现代信息处理技术的深入交互和发展,针对各种体征信号的测量和分析设备在很大程度上减轻了麻醉医生的工作负担。本文针对病患的各种生理体征信号在手术过程中麻醉和清醒状态下的差异性,将信号处理方法和传统医学方法相结合,提出了全身麻醉手术过程中的麻醉深度分析的新指标、新方法,并开发一个能够综合显示信息的实时术中全身麻醉深度分析系统,用于提高临床麻醉监测与评价的准确性,减少人为因素的诊断失误,促进麻醉监测与评价技术临床应用的发展,保障病患在手术过程中的麻醉安全和术后的快速良好恢复。本文的主要研究工作如下：(1)针对术中心电信号在手术过程中受到基线漂移、运动伪迹和工频噪声的干扰问题,通过对现有形态学滤波器数据进行改进以及以均方根误差作为选取形态学中结构元素长度的参数,提出了一种基于形态学的约束适应QRS波群滤波算法。同时提出了一种基于EMD和形态学的心电信号滤波算法。将EMD所得固有模式函数经形态学滤波并进行最小均方根误差对应长度的特征提取,该算法能有效地去除心电信号中基线漂移、运动伪迹和减弱工频干扰。(2)针对术中脑电信号在手术准备阶段受到眼动的干扰问题,提出了基于MEMD和样本熵的眼电干扰滤波算法。算法通过分析对低频眼电信号高度敏感的样本熵值,得出包含眼电干扰的脑电信号样本熵值均小于0.5。在此基础上,比较分析EMD、EEMD、CEEMD和MEMD在分解信号固有模式函数上的性能。(3)针对心率变异性在特定条件下麻醉深度诊断中失效的问题,提出了基于HHT的血流变异性麻醉深度评价指标。血流变异性是在检测手段和分析处理方面都优于心率变异性的一项生理变化参数。当心率变异性在受到特定麻醉药物作用以及手术电刀的影响下,血流变异性可作为术中病患麻醉深度和意识程度的评价指标。基于HHT的边际谱分析,通过比较血流变异性和心率变异性的副交感神经和交感神经频谱分布变化情况,经临床数据证明,血流变异性是一个能够在心率变异性受到干扰时替代其作为诊断病人的麻醉状态的生理指标。(4)针对脑电信号的非线性特点,提出了基于脑电信号分析的样本熵和多尺度熵作为麻醉深度评价指标。近似熵和样本熵均能通过实时分析脑电信号的复杂度来判断病患的麻醉深度和意识程度,但样本熵在性能和敏感度上要优于近似熵。脑电信号的多尺度熵是衡量在不同尺度上脑电信号的复杂度的指标,但是其在实时分析过程中受到信号采样频率和数据长度的限制。因此,提出了自适应多尺度重采样熵作为术中麻醉深度评价指标,它通过自适应地改变有限数量信号的采样率来有效地细化的信号分解尺度,从而建立脑电信号在不同尺度上复杂度分布与术中麻醉深度之间的关系。(5)针对麻醉过程中体征信号分析和麻醉深度评价指标研究,使用Borland C++Builder6.0开发了具有实时采集多个体征信号并用多种方法分析病患麻醉深度的系统平台。更多还原

【Abstract】 Anaesthsia is an indispensable key factor in the clinical surgery, and the way how to ensure the safty of the patients during the surgeries without pain is the core issue in the anaesthetic works. In the process of complex operation, in order to protect the safty of patients, the anesthesiologist must observe the various recordings of physiological signals detected by patients undergoing surgeries. And at the same time, the jugement of depth of anaeshesia (DOA) is provided by the anesthesiologists’own experiences, which is prone to be mistake when the received information and doctor’s experiences are not enough, and the influence of accumulated fatigure, environmental interferency, or ever the potential factors and the difference between patients all would make the decision mistake. However, following the deep interaction of biomedical engineering and modern information processing technology, the measurement and analysis device for kinds of physiological signals are helpful to decrease the workload of anesthesiologists.In this thesis, based on the difference of patient’s physiological signs in state of anaesthesia and consciousness, the features of electrocardiogram (ECG) and electroencephalogram (EEG) are processed respectively, and some new index and methods for analyzing the DOA are presented for monitoring the state of patients during the surgeries with the combination of signal processing and traditional medicine. And a system for analyzing the DOA in general anaesthesia at real time based on multiple physiological signals and multiple methods is developed and tested in clinical surgeries. The purpose of it is to enhance the accuracy of monitoring DOA in clinical surgeries, to reduce the diagnostic errors caused by human factors, to promote the clinical application to DOA monitoring and improve the technical development, to ensure the safety of patients during surgeries and rapid and good recovery after surgeries.The main contents of this thesis are as follows:(1) In order to overcome the interference of baseline drift, movement artifacts and power line interference in the ECG signals, algorithms for detecting QRS waves and filtering are proposed in this thesis. Based on the improvement on the current morphological filter in the problem of missing data in computation, a conditionally adaptive QRS waves dectection algorithm based on the improved morphological filter is presented, which utilizes the root mean square error (RMSE) as the parameter to detect the length of QRS waves in the ECG signals. This algorithm can provide the QRS waves from the ECG signals disturbed by baseline drift. And then a filtering algorithm for power line noise based on the empirical mode decomposition (EMD) and improved morphological filter is able to filter the base line drift, movement artifacts and decrease the power line interference effectively, according to the minimum of RMSE in each intrinsic mode function (IMF) decomposed by EMD.(2) In order to resolve the disturbance of electrooculography (EOG) mixed in the EEG signals, a filtering algorithm based on multiple empirical mode decomposition (MEMD) and sample entropy is presented. With respect to the high sensitivity of sample entropy value to the EOG with low frequency, the sample entropy is less than0.5if EEG signals involve the EOG. After comparison of the performance in decompose the intrinsic mode from original signals among the EMD, ensemble EMD, complementary ensemble EMD and MEMD, the MEMD with the sample entropy as the threshold is capable of reconstructing the EEG and EOG from original EEG signals.(3) According to heart rate variability (HRV) is affected in some particular conditions, a new biomarker blood flow variability (BFV) is presented as an estimate for monitoring the DOA based on the analysis of Hilbert-Huang transform (HHT). BFV is a better biomarker than HRV in aspects of means of detection and analytical processing. And its distribution of power spectrum in specific frequency band can reflect the function of human cardiovascular system and autonomic system indirectly. While the HRV is disable to monitoring the DOA, which causes by the influence of anesthetic drugs and diathermy effect, the ratio of sum of sympathetic division and sum of parasympathetic division in spectrum of BFV is a index for evaluate the state of patients and level of their consciousness. Proven by the clinical experiment, BFV is a good biomarker to replace the HRV in monitoring the DOA while useless of HRV. (4) Based on the entropy theory in nonlinear analysis, the index for monitoring DOA based on the analysis of EEG through approximate entropy(ApEn), sample entropy(SampEn) and multiscale entropy(MSE) is proposed. According to the characteristics of ApEn, SampEn and MSE, there are differences in analysis of EEG through them. In analysis of EEG at real time during the surgeries, SampEn is more sensitive and effective than ApEn in monitoring DOA, even though both of two methods are adaptive to analyze the complexity of EEG signals. Analysis of EEG by MSE can reveal the complexity of EEG in different scales, however, the limitation of number of data and sampling rate is the main problem we meet in monitoring DOA through MSE at realtime. Thus, the multiscale entropy based on the adaptive resampling is presented to estimate the complexity of EEG in more detailed small scales through changing the sampling rate in limited data, so as to establish the relationship between distribution of complexity of EEG in different scales and the DOA.(5) With the analysis of physiological signals of patients during surgeries and study on the estimate of DOA, a system for analyzing the DOA in general anaesthesia at real time based on multiple physiological signals and multiple methods is developed based on the Borland C++Builder6.0and tested in clinical surgeries.更多还原