Multiscale-Wavelet Regularization Method for the Inverse Problem of Electrocardiography

Abstract

Purpose: The objective of the inverse problem in electrocardiography (ECG) is noninvasively to recover regional information about intra-cardiac electrical events for diagnosis by reconstructing the heart surface potential distribution from body surface potentials. However, ECG inverse problem is ill-posed, new regularization methods are needed to obtain a more accurate solution. Methods: In this paper, we propose a new approach which utilize different regularization methods in wavelet domain to solve the ECG inverse problem. Due to the different distribution of ECG signal in different wavelet coefficients which effect on heart surface potential reconstruction, our method applies different regularization methods for different wavelet coefficients. Conclusion: We use the weighted underestimate indicator(WUI) and the weighted overestimate indicator(WOI) of lesion to evaluate our methods on the basis of public simulation dataset. By comparing results of the reconstruction of heart surface potentials of our new approach with the other two traditional approaches, Tikhnonv regularization method in time domain and Tikhnonv regularization method in wavelet domain, it is proved that our approach can improve the accuracy of the solution of ECG inverse problem.