Atrial Fibrillation Signal Extraction Algorithm Based On Stacked Denoising Autoencoders

With the development of social population aging,the incidence of cardiovascular disease has been increasing.Atrial fibrillation(AF)is the most common type of arrhythmia in cardiovascular disease,which affects the life and health of the people,and is also the cause of major heart diseases such as heart failure and stroke.The ECG signals of patients with AF include signals that reflect the patient's atrial activity,which is the atrial fibrillation wave(f wave).Extracting f-waves from ECG signals is helpful for the study of electrophysiology of atrial fibrillation and has great significance in clinical medical research.According to the current clinical requirements for f-wave extraction accuracy,this article makes further research on f-wave extraction algorithms.The main research contents are as follows:(1)Proposes an algorithm for extracting AF signals based on a denoising autoencoder.ECG signals are divided into ventricular signals(QRST)and atrial signals(P wave).When atrial fibrillation occurs,the P wave is replaced by the f wave.Taking the ECG signals of patients with AF as input signals,a deep neural network is formed by stacking denoising autoencoders.This network can mine the deep information of the input signals and has strong learning ability.When training the network,noise is introduced into the input signal.It is expected that the noise interference can be removed by the deep neural network,the robustness of the network model is improved,and a more stable output can be obtained.Use the trained network to reconstruct the QRST wave from the input signal,and then eliminate the QRST wave from the atrial fibrillation ECG signal to obtain the f wave.Experiments on the extraction of simulated atrial fibrillation and the extraction of real atrial fibrillation show that the proposed algorithm can effectively and accurately extract atrial fibrillation signals.(2)Proposes atrial fibrillation signal extraction algorithm based on denoising autoencoders and independent component analysis.Independent component analysis uses the decoupling characteristics of the atrial and ventricular signals to separate atrial and ventricular signal components.The separated ventricular signal component is used as the input signal of the noise reduction autoencoder network,that is,part of the atrial signal component is eliminated from the input signal,so that the reconstructed QRST is more accurate,thereby eliminating the QRST in the AF signal to obtain f wave.Experiments on the extraction of simulated atrial fibrillation signals and the extraction of real atrial fibrillation signals prove that the algorithm Proposed in this paper can further improve the accuracy of extraction of atrial fibrillation signals.This work provides in-depth research on the extraction algorithm of AF signals,which provides a reliable basis for the subsequent verification of the role of AF treatment drugs and the AF assisted diagnosis.