心肺复苏自动化过程中的关键算法研究

发布时间：2018-08-01 09:48

【摘要】：心脏骤停(cardiac arrest，CA)，又称心源性猝死(sudden death)是指心脏的机械活动停止，同时左心室收缩不足或停止收缩。在美国每年大约有22.5万人死于院外(Out-of-hospital)心源性猝死，同时每年约有37～75万住院病人因心脏骤停实施心肺复苏术(cardiopulmonary resuscitation，CPR)。我国虽然还没有确切的心源性猝死的流行病学资料，但专家估计这个数字会达到每年600万人。由于心脏骤停常是冠心病的首发表现形式，有效的心肺复苏是抢救这类患者的唯一途径。心肺复苏是通过对心脏骤停的快速识别和积极抢救，人工重建或恢复自主呼吸与循环，避免发生心肺脑功能不全。快速采取基本生命支持(basic life support，BLS)是心肺复苏成功的关键。心脏骤停有两种不同的形式。一种是由于节律失常引起的心脏骤停(dysarhithmic cardiac arrest)，另一种是因为呼吸停止引起的心脏骤停(asphyxial or respiratory cardiac arrest)。前一种类型的心脏骤停患者大都出现心室颤动(ventricular fibrillation，VF)，而后一类心脏骤停则是由于溺死、药物过量或外伤引起，只有约5％～15％的患者出现心室颤动。对前一类型的患者强调早期除颤和即时的心肺复苏，而对后一种类型的患者则需要实施有效的胸外按压和人工通气治疗。目前心肺复苏有效的抢救成功率依然很低，只有对其不断改进才有可能降低死亡率，这主要是由于以下三个原因引起的。(1)心肺复苏开始得比较晚，包括胸外按压和人工通气。(2)胸外按压的效率较低以及频繁间断。(3)电击除颤不及时造成抢救时机的延误。心室颤动如不及时去除可在数分钟内转为心室静止(asystole)。为了增加早期除颤的机会，尽量缩短除颤时间，体外自动除颤仪(automatic external defibrillator, AED)应运而生。AED的最大特点是提高了电击除颤的自动化程度，是专为非医务人员和初级救生员设计使用的，其识别心室颤动的敏感性与特异性均超过94％。抢救人员只要发现患者意识丧失，无脉搏就可将AED置于患者的胸壁上并启动开关，AED感知心电信号，如能识别出室性心动过速(Ventricular Tachycardia，VT)，或心室颤动，就可自动除颤。应用AED后的研究显示，与单纯的基础生命支持相比可明显提高存活率。美国AHA和IAFC已要求每辆急救车和消防车均需配备AED。随着AED的推广和普及，可以期望更多的生命将会被挽救。由于心脏骤停患者绝大部分(60％-70％)发生在院前，而且在常温下心脏停搏5分钟后脑细胞即可发生不可逆损害，10分钟后脑细胞死亡。在此期间如果不实施心肺复苏术，则心脏的电活动就会逐渐消失，最后出现心室静止，心电图出现一条直线。心肺复苏和药物治疗可能会增加缺血心肌的血液和氧气循环，，从而将心室静止转变为心室颤动，而后方可电击除颤。因心室静止和心室颤动期间心脏停止收缩，因此无法检测到脉搏信号。如果心脏只存在心肌的电活动而没有相应的机械收缩，则称为心电机械分离(electromechanical dissociation，EMD)或无脉搏心电活动(pulseless electrical activity, PEA)。这种情况常会出现在药物治疗或心肺复苏但没有实施电击除颤，或由心室颤动转变为心室静止的过程中。目前心肺复苏有一个标准的操作指南(the international guidelines)，它根据对呼吸、脉搏以及心电节律的检测来确定相应的复苏措施。而目前使用的AED只能根据患者的心电波形做出相应的节律分析决定是否需要电击除颤，其它如呼吸检测、脉搏检查等均需要由目击者或抢救人员来判断。为实现心肺复苏的全自动化，目前仍需要解决以下一些问题： (1)胸外按压过程中的心电节律识别。心脏骤停患者的存活率随心室颤动持续时间的延长而迅速降低，平均每分钟下降约7％～10％。当施行电击除颤的时间延迟10～12分钟以上时，存活的可能性几乎为零。尽早应用基础的心肺复苏，并尽快实施电击除颤，可有效提高心脏骤停患者的存活率。但目前使用的体外自动除颤仪在实施电击除颤之前需要反复进行节律分析。否则如果将正常的非除颤心室搏动节律误判为除颤节律，并实施不必要的电击，那么将会对病人的心脏产生极大的损伤，并导致严重的后果。因此为确保心室颤动的正确识别，在节律分析期间，必须停止对病人的胸外按压和通气过程。这一过程大约需要12至20秒的时间。在这一过程中，电击除颤的成功率因为室颤时间的延长而大大降低，尤其是在院外病人的复苏期间。因此如果能有一种比较可靠的心室颤动节律识别算法，即使是在对病人实施胸外按压期间也能对心电波形进行可靠分析，那么病人存活的几率将会得到有效提高。 (2)胸外心脏按压的有效性监测分析。胸外心脏按压的质量也是成功复苏的关键，包括按压深度、按压频率和胸廓的回弹程度。尤其的恰当的按压深度，它是保持一定冠状动脉灌注压(coronary perfusion pressure，CPP)的关键。但是，研究表明许多心脏骤停患者在心肺复苏过程中没有得到有效的胸外按压，主要表现在按压频率较低、按压深度不足以及没有保持适当的循环血流。而在院外急救过程中，胸外按压由于没有得到有效的监测，整个过程就只能靠抢救者的感觉和视觉判断。随着复苏过程的进行，急救人员急需了解胸外按压的效果以及由此产生的病人心脏血流的变化，从而实施进一步的治疗，包括优化电击除颤或继续胸外按压治疗。 (3)实时呼吸及脉搏检测。在过去的20年里，心室颤动或室性心动过速在心脏骤停中出现的比例逐渐下降，已经低于50％，而无脉搏心电活动PEA及其它类型的心脏骤停所占比例并没有改变。心室颤动的下降通过心室静止增加得以补偿。这就要求第一目击者或急救人员快速判断病人在失去意识的情况下，是否具有呼吸、脉搏或者足够的血液循环。但这两项指标的检测对院外急救来说却非常困难。因为传统脉搏检测是通过感触病人颈动脉的搏动来实现的，呼吸检测则是通过贴近病人嘴部感觉呼吸气流和观察胸腔的变化来实现。这些适用于普通人群的方法，很难应用于心跳和呼吸极其微弱的冠心病患者。若不能准确地检测呼吸与脉搏，就不能正确区分由于节律失常和窒息引起的心脏停搏，并实施正确的复苏措施。针对以上心肺复苏过程中的问题，我们扩展了目前AED使用的心电采集及除颤电极功能，利用一对胸前除颤电极实现心电信号与胸阻抗信号的采集分析。通过对按压过程中心电信号的分析实现对胸外按压效果的监测，通过对胸阻抗信号的处理实现微弱呼吸与脉搏信号的检测，实现了心肺复苏的自动化实现方案，其中的主要算法包括： (1)不间断胸外按压过程中的心电节律识别算法。采用基于连续小波变换及形态一致性评估的分析方法实现对心电信号的节律识别。通过对心电信号中R波形态一致性的量化分析，可以区别规则性心电节律(organized rhythm)和不规则心电节律(disorganized rhythm)。对于规则性心电节律，通过连续小波变换中R波峰值出现的频率来估计心率的变化，以区别室性心动过速与正常节律。而对于不规则性心电信号，则通过幅度频率谱面积分析来区分心室颤动与心室静止。 (2)胸外按压有效性的监测与电击除颤的优化算法。早期对胸外按压有效性的监测通过对心室颤动信号的幅度分析来实现。此后动物与临床实验研究表明，心室颤动信号的频率与CPP呈相关性。但由于心电信号的幅度与频率均因病人的个体差异而对实验结果有较大的影响。本研究小组将心电信号的幅度与频率相结合，提出了一种基于幅度频率谱面积的分析方法，它定义为一定频带宽度下信号功率谱所包含的面积。我们期望这种用于对电击除颤优化分析的方法可扩展应用于对胸外按压有效性的实时监测中。 (3)实时呼吸及脉搏检测分析算法。利用心电检测／除颤用电极提取的胸阻抗信号由两部分组成：一是包含了频率较低但幅度较高的呼吸阻抗信号，二是频率略高但幅度较低的反映心脏机械活动的心阻抗信号。我们用心电信号作为参考，利用自适应滤波器将呼吸阻抗信号和心阻抗信号相分离。最后用峰值检测算法利用检测到的心阻抗信号幅度，判断检测脉搏信号的有无，并用呼吸阻抗信号的幅度估计潮气量的大小，从而确定呼吸的类别。为检验这些算法的有效性，我们对不同的算法进行了相应的实验设计与临床实验。对用AED记录的232例院外心脏骤停患者的心电信号分析结果表明，所提出的自动心电节律分析算法可以实现在胸外按压不间断情况下对心电节律的可靠分析。对除颤信号的检测敏感率为93％，特异性为89％。在一组由心律失常引起的心脏骤停动物模型中，幅度频率谱面积分析与CPP分析结果具有良好的相关性，并且对电击除颤结果的预测具有较高的准确性。在另一个由窒息引起的心脏骤停动物模型中，由心脏机械活动引起的心阻抗信号变化与脉搏电压具有良好的相关性，而由呼吸引起的呼吸阻抗信号的变化则与呼吸潮气量的变化正相关。临床研究结果表明，通过对脉搏及呼吸信号的无创检测，本研究提出的算法能够正确地区分因心律失常和窒息引起的不同类型的心脏骤停，从而及时地提示急救人员实施相应的复苏方案。而在不间断胸外按压情况下对心电节律的实时分析，则有效地避免了因节律分析造成的延误，提高患者的存活率，同时可以避免不必要或不成功的电击除颤，较好地解决了当前心肺复苏过程中存在的不足，实现了心肺复苏的全自动化。
[Abstract]:Cardiac arrest (CA), also known as sudden cardiac death (sudden death) refers to the mechanical activity of the heart, and the left ventricular systolic and contractile contraction. In the United States, about 225 thousand people die of sudden cardiac death outside the hospital (Out-of-hospital) in the United States each year, and about 37~75 million hospitalized patients undergo cardiopulmonary resuscitation every year because of cardiac arrest. (cardiopulmonary resuscitation, CPR). Although there is no exact epidemiological data on sudden cardiac death in China, experts estimate that this number will reach 6 million people a year.
Cardiac arrest is often the first manifestation of coronary heart disease. Effective cardiopulmonary resuscitation is the only way to rescue this type of patients. Cardiopulmonary resuscitation is a rapid identification and active rescue of cardiac arrest, artificial reconstruction or recovery of spontaneous breathing and circulation to avoid cardiopulmonary cerebral dysfunction. Basic life support (basic life) is adopted quickly. Support, BLS) is the key to the success of cardiopulmonary resuscitation.
Cardiac arrest has two different forms. One is dysarhithmic cardiac arrest caused by arrhythmia and the other is cardiac arrest caused by respiratory arrest (asphyxial or respiratory cardiac arrest). The former type of cardiac arrest is mostly ventricular fibrillation (ventricular fibrillation, VF). The latter type of cardiac arrest is caused by drowning, drug overdose or trauma, only about 5% to 15% of the patients have ventricular fibrillation. Early defibrillation and immediate cardiopulmonary resuscitation are emphasized in the previous type of patients, while effective external compression and artificial ventilation are required for the latter type of patients.
At present, the effective rescue success rate of cardiopulmonary resuscitation is still very low. It is possible to reduce the death rate only by continuous improvement. This is mainly due to the following three reasons. (1) cardiopulmonary resuscitation begins relatively late, including external chest compression and artificial ventilation. (2) the efficiency of chest compressions is low and frequent interruption. (3) defibrillation is not made in time. The delay of the time of rescue.
Ventricular fibrillation, if not removed in time, can turn to asystole in a few minutes. In order to increase the opportunity for early defibrillation, the defibrillation time is shortened as far as possible. In vitro automatic defibrillator (automatic external defibrillator, AED) is the largest characteristic of.AED to improve the degree of automation of electric shock defibrillation, which is specially for non medical personnel and The primary lifesaver designed to use the sensitivity and specificity of identifying ventricular fibrillation more than 94%. rescuers can put AED on the chest wall of the patient and start the switch without pulse, and AED to perceive the ECG signals, such as identifying ventricular tachycardia (Ventricular Tachycardia, VT), or ventricular fibrillation, Automatic defibrillation. Research after the application of AED shows that the survival rate can be significantly increased compared with the simple basic life support. AHA and IAFC in the United States have required each emergency vehicle and fire engine to be equipped with AED. with the promotion and popularization of AED, and more life will be expected to be saved.
Because most of the patients with cardiac arrest (60%-70%) occur before the hospital and the brain cells can have irreversible damage after 5 minutes of cardiac arrest at normal temperature, the brain cells die after 10 minutes. During this period, the electrical activity of the heart will gradually disappear if the cardiopulmonary resuscitation is not implemented, and the ventricular rest is finally appeared, and a cardiogram appears in the heart. Cardiopulmonary resuscitation and drug therapy may increase the circulation of blood and oxygen in the ischemic myocardium, and then turn the rest of the ventricle into ventricular fibrillation, and then electric shock defibrillation. The heart stops contraction during the rest of the ventricle and ventricular fibrillation, so the pulse signal can not be detected. If the heart only exists electrical activity of the heart, there is no corresponding The mechanical contraction is called electromechanical dissociation (EMD) or pulseless electrical activity (PEA). This often occurs in the process of drug treatment or cardiopulmonary resuscitation, without electric shock defibrillation, or ventricular fibrillation to the rest of the ventricle.
Currently, cardiopulmonary resuscitation has a standard operation guide (the international guidelines) to determine corresponding recovery measures based on detection of respiratory, pulse, and electrocardiographic rhythms. The current use of AED can only make corresponding rhythmic analysis based on the electrocardiogram of the patient to determine whether electric shock defibrillation is required, other such as breathing detection, pulse. Cardiopulmonary resuscitation (CPR) should be fully automated. The following problems still need to be solved:
(1) cardiopulmonary rhythm identification in the process of chest compression. The survival rate of patients with cardiac arrest rapidly decreases with the duration of ventricular fibrillation. The average survival rate is about 7% to 10%. per minute. The possibility of survival is almost zero when the time of electric shock defibrillation is delayed more than 10~12 minutes. Electric defibrillation can effectively improve the survival rate of patients with cardiac arrest. But the current automatic defibrillator needs repeated rhythms before defibrillation is implemented. Otherwise, if the normal non defibrillating ventricular beat rhythm is misjudged as defibrillation rhythm and the unnecessary electric shock is implemented, it will produce the heart of the patient. In order to ensure the correct identification of ventricular fibrillation, it is necessary to stop the external compression and ventilation of the patients during the rhythm analysis. This process takes about 12 to 20 seconds. In this process, the success rate of defibrillation is greatly reduced because of the prolongation of the time of ventricular fibrillation. During the recovery of patients outside the hospital, a reliable algorithm for ventricular fibrillation rhythm identification, even if a reliable analysis of the ECG waveform is carried out during external compression of the patient's chest, then the probability of patient survival will be effectively improved.
(2) monitoring and analysis of the effectiveness of the chest compression. The quality of the external cardiac compression is also the key to successful recovery, including the compression depth, the compression frequency and the rebound degree of the chest. The critical depth of compression is the key to maintaining a certain coronary perfusion pressure, CPP. However, the study shows many hearts In the process of cardiopulmonary resuscitation, patients with CPR did not get effective chest compressions, mainly in lower pressing frequency, less pressing depth and no proper circulation blood flow. In the process of first aid, the external pressure of the chest was not effectively monitored, and the whole process could only depend on the sense and visual judgment of the rescuers. With the process of resuscitation, emergency personnel urgently need to understand the effects of chest compressions and the resulting changes in the blood flow of the patient's heart, so as to carry out further treatment, including the optimization of electric shock defibrillation or the continuation of chest compressions.
(3) real time breathing and pulse detection. In the past 20 years, the proportion of ventricular fibrillation or ventricular tachycardia in cardiac arrest has declined gradually, already less than 50%, and the proportion of PEA without pulse electrocardiography and other types of cardiac arrest has not changed. It is required that first witnesses or first aid personnel quickly judge whether the patient has breathing, pulse, or sufficient blood circulation in the absence of consciousness. But the detection of these two indicators is very difficult for the first aid, because the traditional pulse detection is achieved by the pulsation of the patient's neck movement. It is close to the patient's mouth feel breathing air flow and observation of the changes in the chest. These are difficult to apply to common people. It is difficult to apply to patients with very weak heartbeat and breathing. If you can't accurately detect breathing and pulse, you can't correctly distinguish the cardiac arrest caused by arrhythmia and asphyxia, and carry out the correct recovery. Sue measures.
In order to solve the problems in the process of cardiopulmonary resuscitation, we extend the function of ECG acquisition and defibrillation electrode used by AED, and use a pair of predefibrillation electrodes to collect and analyze the ECG signal and thoracic impedance signal. Through the analysis of the central electrical signal of the pressing process, we can monitor the effect of chest compression, through the impedance signal to the chest. Processing realizes the detection of weak respiratory and pulse signals, and realizes the automatic implementation of CPR. The main algorithms include:
(1) ECG rhythm recognition algorithm in uninterrupted chest compressions. The rhythmic recognition of ECG signals is realized by continuous wavelet transform and morphological consistency evaluation. The regular ECG rhythm (organized rhythm) and irregular ECG joint can be distinguished by quantitative analysis of the conformance of R wave in ECG signals. Law (disorganized rhythm). For regular ECG rhythm, the change of heart rate is estimated by the frequency of the peak of peak of R wave in continuous wavelet transform in order to distinguish ventricular tachycardia and normal rhythm. For irregular ECG, the area of ventricular fibrillation and ventricular quiescence by amplitude frequency spectrum area analysis.
(2) an optimal algorithm for monitoring the effectiveness of compressions and defibrillation. Early monitoring of the effectiveness of chest compressions was achieved by the amplitude analysis of ventricular fibrillation signals. Animal and clinical studies have shown that the frequency of ventricular fibrillation signals is related to CPP, but the amplitude and frequency of ECG signals are due to the patient's The study group combines the amplitude and frequency of the ECG signal and proposes an analytical method based on the amplitude frequency spectrum area. It is defined as the area contained in the signal power spectrum under a certain band width. We expect that this method can be extended to the optimization analysis of electric shock defibrillation. It is applied to real-time monitoring of chest compression effectiveness.
(3) real-time breathing and pulse detection analysis algorithm. The impedance signals extracted from electrocardiogram detection / defibrillation electrodes are composed of two parts: one is a respiratory impedance signal with a lower frequency but a higher amplitude, and two is a cardiac impedance signal with a slightly higher frequency but a lower amplitude that reflects the mechanical activity of the heart. Our attentively electrical signal is used as a reference. An adaptive filter is used to separate the respiratory impedance signal and the impedance signal of the heart. Finally, the peak detection algorithm is used to determine whether the pulse signal is detected by the detected amplitude of the impedance signal, and the magnitude of the tidal volume is estimated with the amplitude of the respiratory impedance signal, thus the category of the respiration is determined.
In order to test the effectiveness of these algorithms, we have carried out the corresponding experimental design and clinical experiments on different algorithms. The results of electrocardiogram analysis of 232 patients with cardiac arrest by AED show that the proposed automatic ECG rhythm analysis algorithm can be reliable to the reliability of ECG rhythm under uninterrupted chest compressions. Analysis. The sensitivity of defibrillation signal detection was 93%, and the specificity was 89%. in a group of cardiac arrest animal models caused by arrhythmia. The amplitude frequency spectrum area analysis was well correlated with the results of CPP analysis, and had high accuracy for the prediction of electric shock defibrillation results. In another, the heart sudden caused by asphyxia. In the animal model, the change of the cardiac impedance signal caused by cardiac mechanical activity is closely related to the pulse voltage, and the change of respiratory impedance signals caused by respiration is positively related to the change of the volume of respiratory tide.
The results of the clinical study show that the proposed algorithm can correctly distinguish the different types of cardiac arrest caused by arrhythmia and asphyxia by the noninvasive detection of pulse and respiratory signals, and prompt the emergency personnel to implement the corresponding resuscitation scheme in time. The analysis can effectively avoid the delay caused by the rhythm analysis, improve the survival rate of the patients, and avoid the unnecessary or unsuccessful electric shock defibrillation, and better solve the shortcomings of the current cardiopulmonary resuscitation.
【学位授予单位】：南方医科大学
【学位级别】：博士
【学位授予年份】：2007
【分类号】：R459.7;R311

【参考文献】