Research On Prediction Of Sudden Cardiac Death Based On ECG Signal

Sudden cardiac death refers to the sudden and unexpected death caused by cardiac arrhythmia and hemodynamic mutation,which is directly reflected in the occurrence of malignant arrhythmias such as ventricular tachycardia.With the popularization of wearable devices and telemedicine,the acquisition and analysis of ECG signals are more convenient,and the research on arrhythmia identification and sudden cardiac death prediction based on ECG signals is of great significance.However,due to the suddenness and high risk of malignant arrhythmias,and the ECG signals collected by wearable devices are usually single-lead and have low signal-to-noise ratio,therefore,under the condition of singlelead ECG signal with low signal-to-noise ratio,the research on arrhythmia identification and sudden cardiac death prediction method is still full of challenges.In this paper,an arrhythmia recognition algorithm is proposed based on single-lead ECG signals,and the accuracy is 96.43%.A sudden cardiac death prediction algorithm is proposed based on low signal-to-noise ratio single-lead ECG signal,the average prediction accuracy rate reached 93.22%within 30 minutes before the occurrence of SCD,and a comprehensive index SCDI to measure the risk of SCD is also proposed.This paper proposes an algorithm for arrhythmia identification suitable for single-lead ECG signals.The algorithm extracts the RR interval features,high-order statistical features,wavelet features and IMF component features from the ECG signal from multiple dimensions in the time-frequency domain,and classifies the beat types into five categories according to AAMI’s recommendations with an accuracy rate of 96.43%.Compared with other studies based on multi-lead or deep learning methods,this method achieves excellent performance with a lower amount of data processing and algorithm complexity.Second,this paper proposes a sudden cardiac death prediction algorithm suitable for single-lead low SNR ECG signals in a wearable environment.This method adds electrode motion artifact noise to the ECG signal with a signal-to-noise ratio of 12dB to simulate the low signal-tonoise ratio signal quality collected in wearable devices.In addition,this method introduces the malignant heart failure database as a supplement to the normal ECG signal category,so as to expand the control range for comparison with SCD high-risk signals.The algorithm extracts the 12dimensional features of ventricular late potential,T wave alternation and corrected QT interval from the signal,and the average prediction accuracy rate reaches 93.22%within 30 minutes before the occurrence of sudden cardiac death.When only using the normal sinus rhythm database as a control,the average accuracy rate reached 95.43%within 30 minutes,achieving excellent performance.In addition,this paper proposes an independent risk index SCDI to measure the risk of sudden cardiac death.The method combines the twodimensional features that contribute greatly in the algorithm to give the calculation formula of SCDI.This index has a good distinguishing effect on high-risk signals of sudden cardiac death and health signals,and can provide auxiliary references in clinical trials to a certain extent.