| Objective: Cardiac arrest is a fatal disease that seriously threatens the health of human beings.Without effective treatment,most of them will die immediately.According to the survey,newly reported cases in China of cardiac arrest exceed 500 thousand in each year,ranking first in the world.The number will increase as the population ages and the incidence of coronary heart disease increases in recent years.Cardiopulmonary resuscitation,which includes artificial ventilation,chest compression and defibrillation,is the most effective rescue measure for cardiac arrest.It can help the patients to establish temporary blood perfusion and breathing,to restore the heart mechanical activity,eventually to return to spontaneous circulation.Over the past few decades,the development and promotion of cardiopulmonary resuscitation have saved tens of thousands of lives worldwide.It fully affirmed the significance of the work in this field.Early defibrillation is one of the key links of cardiopulmonary resuscitation.Some reports showed that the implementation of defibrillation within 3-5 minutes can improve the survival rate to 50-70%.In order to shorten the time from onset of cardiac arrest to defibrillation,more and more public occasions are equipped with automatic external defibrillators which can automatically analyze ECG signals of patients,and prompt the rescuer to implement the defibrillation or automatically implement it at the right time.However,chest compressions will introduce cardiopulmonary resuscitation artifacts into the ECG,seriously interfere with the rhythm analysis of automated external defibrillators.Traditionally,reliable analysis can only be implemented with current automated external defibrillators when all the rescuer left the patients.According to the survey,the average discontinued interval of chest compressions will be more than 15 s and interruptions of chest compressions were averaging 24%-57% of the total arrest time.These interruption intervals reduced the probability of restoration of spontaneous circulation.Some animal experiments have shown that the likelihood of successful resuscitation is reduced by as much as 50% by a 20-s interruption of chest compressions.The interruption of chest compressions is considered to be a major factor that leads to poor outcomes in cardiac arrest patients.In summary,to design a reliable algorithm used for ECG rhythm detection with uninterrupted chest compressions can minimize the interruption time of chest compressions,and eventually improve the survival rate of patients suffered from cardiac arrest.Methods and Contents: In this paper,a reliable ECG analysis algorithm was designed based on adaptive cardiopulmonary resuscitation artifacts filter and antiinterference ECG rhythm detection algorithm.The main work outlined as follows:(1)Research on the algorithm of adaptive cardiopulmonary resuscitation artifacts removal.Firstly,the spectral characteristics of all kinds of typical ECG rhythm signals and CPR artifact signals were analyzed,and the difference in spectral distribution was compared.On this basis,the reference signal reflecting fundamental frequency of the cardiopulmonary resuscitation artifacts was obtained using the improved noise assisted multi-channel EMD.These reference signals were used to construct multi harmonic LMS filter model,which can further establish the precise model of cardiopulmonary resuscitation artifacts.The key parameters of this filter were optimized with the index of SNR gain.Finally,the real ECG rhythms were restored from the corrupted signals without any additional reference signal channels.(2)The construction and optimization of the robust BP neural network for detection of life-threatening arrhythmias.Based on a review of related published documents,21 metrics were extracted from the ECG signals using 10 different algorithms.Each one of these metrics respectively characteristics each aspect of the ECG signals,such as morphology,gaussianity,spectra,variability,complexity,and so on.The performance of these metrics in ECG rhythms classification was evaluated.A BP neural network for life-threatening arrhythmias detection was constructed based on these metrics and optimized by genetic algorithm.(3)Simulation experiments.Corrupted ECG signals at different SNR levels were constructed to simulate the surface ECG during uninterrupted chest compressions combined with the openly available ECG database and cardiopulmonary resuscitation artifacts.The detection performance of the BP neural network proposed in this paper was evaluated on this database.Under the same conditions,the filter performance of the adaptive artifacts suppression algorithm designed in this paper was evaluated and compared with other existing algorithms.(4)Evaluation by animal experiments.The animal model of VF was established by alternating current delivered to the ventricle of landrace pigs.Real artifacts corrupted ECG signals were obtaind by standard cardiopulmonary resuscitation and annotated according to other physiological signals such as arterial pressure and EtCO2.Based on this database,performance of the detection algorithm of life-threatening arrhythmias during uninterrupted chest compressions was verified and compared with other existing algorithms.The innovation of this paper was that a reliable ECG analysis algorithm with uninterrupted chest compressions was designed only using surface ECG based on noise assisted multi-channel EMD,multi harmonic LMS filter,genetic algorithm and BP neural network,without needing any reference channels.Results:(1)The detection sensitivity,specificity and accuracy of the BP neural network in the database without artifacts was respectively 99.36%,100% and 99.11%,superior to the traditional feature classification method.With the increasing degree of artifacts interference,the detection specificity of the traditional feature classification method decreased rapidly.In a lower degree of artifacts interference(SNR=-3),the specificity has decreased to 16.36%,while the specificity of the BP neural network still remained at 82.66%,indicating that the BP neural network proposed in this paper has a better anti-interference performance.(2)The cardiopulmonary resuscitation artifacts in all kinds of ECG signals can be successfully suppressed with the adaptive filter algorithm based on EMD and multi harmonic LMS.The detection specificity has been improved significantly.Even in the condition of high degree of artifacts interference(SNR=-12),the detection specificity was 86.55% and the area under the ROC curve was 0.9733,better than other existing filter algorithm.This advantage is more obvious when the artifacts has more harmonic components.(3)In animal experiments,the ECG signals filtered by the adaptive filtering algorithm designed in this paper had a good performance in rhythm detection.The sensitivity and specificity were increased to 95.78% and 91.58%,and the area under the ROC curve was 0.9805,significantly better than the other algorithms.Conclusion: An algorithm for cardiopulmonary resuscitation artifacts suppression and ECG rhythm detection algorithm during uninterrupted chest compressions was proposed in this paper.Compared with other existing algorithms,this algorithm has shown better performance whether in simulation experiments and animal experiments.It would enable compressions to continue during reliable ECG analysis.The interruption of chest compressions can be avoid to the maximum extent by using it to replace the traditional method.This study provided software support for optimization of the traditional defibrillation monitor.With the continuous improvement of the algorithm,it will play an important role in the first aid of cardiac arrest. |
PDF Full Text Request