Research On Diagnosis Of Sudden Cardiac Death Based On ECG Signals And TWA

Sudden Cardiac Death is seriously harmful to the human health. Each year there are about 12 million patients die from SD, which 75% are SCD patients. If people have this symptom will die within one hour. But timely implement the Early Electric Defibrillation is the best way to save life. So the effective way to deal with SCD is not salvage but prediction and prevention before it. Electrocardiogram(ECG) is a reflection of the electrical activity of the Heart, and its characteristic parameters has extremely high value for myocardial ischemia and sudden death diagnosis. TWA is a non-stationary variability of ECG phenomenon, which has become a non-invasive and independence predictor of ventricular arrhythmias and SCD.The thesis introduced a modern non-stationary signal processing method to study characteristic parameters of ECG and TWA. Then realizing SCD diagnosis is based on pattern recognition technology. Accomplished works are as follows:The Improvement of EMD: A novel boundary processing method was proposed to decrease the boundary distortion of EMD by means of signal extending. Then according to the characteristic of human's ECG signals, we selected the optimal stopping criteria for sifting.Research on ECG Signals Pretreatment:The improved EMD was used to decompose the ECG signal into stationary Intrinsic Mode Functions (IMFs) and residual components. In order to highlight the characteristic parameters of ECG, the two IMFs of low frequencies were reconstructed after being de-noised with threshold method. ECG features location: The studies show QRS waves,P wave and T wave were highlighted by analyzing the frequencies of IMFs and residual components. We adopted the plane geometry method accompanied with a series of estimate method to orient the ECG features.Detection of TWA:Systematic analysis and TWA detection can be studied in the dissertation. The first of all,we will discuss the characteristics and detection difficult of TWA then compare the detection methods of TWA,the principles of various methods will be analyzed in detail.Because of simulation results,these methods are simulated and the strengths and weaknesses of variety detection methods are illustrated. In order to detect TWA ,A method which combines the spectral analysis and correlation method is designed. Using singular value decomposition to calculate the relevant index,which can effective avoid the interference of noise and improve the accuracy of detection. SCD diagnosis:We proposed a new method of diagnosing SCD on the basis ofSVM. Firstly, we have introduced statistical learning theory and SVM theory. Then the SVM classifier was built to diagnose the SCD. At the same time,by comparing with the traditional neural network(BP),the results showed that the SVM had better diagnosis ratio and generalization ability.