Research Of T Wave Alternans Analysis Algorithm Based On Particle Filter

SCD (Sudden Cardiac Death) has become the second largest cause of death after cancer. A large number of clinical studies suggest that there is a patent relationship between TWA (T Wave Alternans) and ventricular arrhythmias, which is an efficient non-invasive index to predict malignant ventricular arrhythmias and SCD. Since TWA in the ECG are of microvolt amplitude, non-stationary and non-linearity and non-gaussianity, it’s quite difficult to obtain index with diagnostic value. Therefore, the accuracy of TWA analysis has become the key issue to predict SCD.Many signal processing methods have been proposed to analysis TWA in single lead ECG But these methods have poor sensitivity and low accurracy to non-stationary and non-Gaussian TWA. For TWA analysis in multi-lead ECG, there are few methods exept for some single lead-based methods. And the shortcoming of these methods is low robustness. To improve the accuracy of quantification of single lead TWA and the robustness analysis of multi-lead TWA, a TWA analysis algorithm based on particle filter is proposed in this paper.(1) For non-stationary, non-linear and non-Gaussian TWA in single lead ECG, a non-linear state space model with Laplace noise model was proposed first to describe the TWA analysis system. Then, particle filters is applied to estimate TWA amplitude of single lead ECG after preprocessing. Specially, general PF and Gaussian Sum PF are implemented to analysis single lead TWA.(2) For TWA analysis in multilead ECG, PF method combined with joint analysis information is proposed to detect and estimate multilead TWA. According to a multilead TWA analysis scheme, after preprocessing, a linear transformation using principal component analysis (PCA) is processed on multilead ST-T complex. Then PF-based method is applied to each lead for TWA detection. To obtain TWA with clinical value, TWA amplitude is extimated using PF after signal reconstruction.From the simulation results, our proposed single lead TWA analysis method based on PF is demonstrated the most actual and robust for TWA quatification. Also, non-stationary TWA can be tracked actually and in real-time. For our multi-lead TWA analysis method based on PCA and PF, it has the best performance on detection and estimation of TWA and has good robustness and high anti-noise ability.