Research On Key Technologies Of Mobile Smart Electrocardiography Monitoring System

Cardiovascular disease(CVD)has been harming people's health seriously,as there are 290 million patients with CVD in China.Hence,the prevention and treatment of CVD is urgent.The detection and diagnosis of cardiac disease,which is a type of disease that is common in CVD and has a high mortality rate,is challenging due to its sudden and transient symptoms.Traditional electrocardiogram(ECG)monitoring systems have problems in both real-time monitoring and continuous monitoring scenarios,making them not suitable for daily ECG monitoring.Therefore,we designed a mobile smart ECG monitoring system in this thesis.We studied the key problems in the hardware and software implementation,signal processing,disease analysis,and clinical application,to realize the everyday ECG signal acquisition and processing based on mobile smart ECG monitoring system.We first designed and implemented the mobile ECG monitoring prototype system to achieve real-time processing,display,and analysis of ECG signals.The system included a mobile smart ECG monitoring device,an application,and a management software,to solve the everyday application problems of traditional static and dynamic ECG monitoring methods,as well as lay the foundation for the research on key technologies of mobile smart ECG monitoring system.Meanwhile,we proposed a lossy-to-lossless ECG compression framework and analyzed its actual average compression performance with communication power optimization level.We also implemented this framework with a three-stage pipeline hardware structure,to make the mobile smart ECG monitoring system compatible with the limited communication power consumption and storage resources of mobile platforms.Furthermore,aiming at the common lead inversion problem in mobile smart ECG monitoring,an automatic cardiac polarity correction algorithm was proposed to improve the performance of ECG signal analysis.Secondly,we proposed a real-time electromyogram(EMG)noise suppression algorithmbased on Kalman filter to denoise the significant EMG interference mainly faced by mobile ECG monitoring.In order to solve the problem of morphological distortion of QRS waves caused by traditional EMG denoising,we first preprocessed the ECG signal to remove the baseline wandering,powerline noise,and high frequency interference.The QRS waves were then extracted and retained,while EMG noise in the remaining regions was suppressed by a one-dimensional Kalman filter,so that we could enhance the performance of EMG denoising,as well as preserve the morphology of QRS waves;resulting in improving the detection accuracy of QRS waves and P waves.Moreover,a smart heart beat classification method based on XGBoost was proposed to improve the analysis of arrhythmia using single-lead ECG signals.A total five waveform morphological features regarding the RR interval and ECG amplitude were designed,combining with wavelet decomposition features.This 56-dimensional feature vector was extracted in every cardiac cycle of ECG for XGBoost classification training.The trained XGBoost classifier then automatically divided the heart beats into five categories,i.e.normal,supraventricular ectopic,ventricular ectopic,fusion,and unknown.While achieving better classification performance than traditional machine learning methods,the algorithm retained the advantages of training cost and model size compared to deep learning methods with similar performance.Finally,since the standard 12-lead ECG signal is commonly used in the diagnosis of cardiac abnormalities in clinical practice,we proposed a piecewise linear synthesis method for standard 12-lead ECG signals based on adaptive region segmentation,in order to improve the clinical value of the reduced-lead ECG signals collected by the mobile smart ECG monitoring system.With this algorithm,the ECGs were segmented by cardiac cycles.Each cycle was further divided into four regions according to different cardiac electrical activity stages.A personalized linear regression algorithm is then applied to these regions respectively for improved ECG synthesis.In addition,we introduce a metric,namely the ST-level critical denivelation ratio,to evaluate the impact of lead reconstruction methods on key sensitive parameters of ECG.After concluding the works of this thesis,we planned our future research on flexible electrode based wearable ECG monitoring device,motion artifact removal in daily ECG signal monitoring,automatic ECG diagnosis using deep learning methods,as well as lead signal reconstruction optimization and non-standard lead system.