Research On Atrial Fibrillation Signal Extraction Algorithm Based On Combining ICA And ESN

With the aging of society,the morbidity and mortality of cardiovascular diseases have been increasing,and now it has surpassed neoplastic diseases to become the number one disease.Atrial fibrillation(AF),as the most common cardiovascular disease in clinic,affects about 5% of the population over 50 years old.It is also an important cause of severe heart disease such as stroke and heart failure.In recent years,because the digital features of atrial fibrillation signals have shown important clinical value in the determination of arrhythmia drug effects,atrial fibrillation-assisted diagnosis,and location of fibrillation.Extracting AF signal become an important method to help clinicians further analyze atrial fibrillation.In order to meet the clinical requirements for real-time and accurate extraction of atrial fibrillation signals in dynamic ECG signals,this paper makes a further research on atrial fibrillation signal extraction algorithm.The main contents of this paper are as follows:(1)In this paper,a novel AF signal extraction algorithm based on genetic echo state network is proposed.In order to solve the blindness of artificial random selection of reservoir parameters in the process of extracting AF signal based on echo state network(ESN),a genetic algorithm has been introduced to optimize the reservoir parameters to generate an optimal network reservoir for extracting atrial fibrillation signals.At the same time,in order to accelerate the evolution,an adaptive genetic operator has been designed to meet the real-time requirements of the echo state network to extract the AF signal while ensuring the search range.The experimental results show that the relative root mean square error(RRMSE)of the proposed algorithm in the simulated AF signal extraction experiment reaches 0.22.And the spectral concentration(SC)of the proposed algorithm in the real AF signal extraction experiment reaches 42%.It shows that the proposed algorithm can accurately and effectively extract the AF signal while ensuring real-time extraction of AF signals.(2)In this paper,a novel AF signal extraction algorithm based on combining ESN and independent component analysis(ICA)is proposed.Based on the decoupling characteristics of atrial signals and ventricular signals of ECG signals,independent component analysis hasbeen used to separate ventricular components and non-ventricular components.The ventricular component have been selected as the input reference signal of the echo state network,and the QRST wave has been estimated by the nonlinear mapping ability of the network.Therefore,it is more accurate to eliminate the QRST in the mixed signal to obtain the AF signal.The experimental results show that the RRMSE of the proposed algorithm in the simulated AF signal extraction experiment reaches 0.14,and the SC in the real AF signal experiment reaches 64.86%.It shows that the proposed algorithm can further improve the extraction accuracy of AF signal under the premise of ensuring local extraction of AF signal.This paper makes an in-depth study of the extraction algorithm of AF signal,and provides a reliable basis for the verification of the therapeutic effect of atrial fibrillation and the auxiliary diagnosis of AF.