Denoising of ECG by Statistical Adaptative Thresholding and Detection of T-Wave Alternans Using Principal Component Analysis

One of the most important tasks in the delineation of the ECG is the detection of the spikes in noisy recordings and to estimate TWA in the ECG on a single-lead basis. The wavelet transform is a very appropriate pre-filtering step prior to the thresholding of coefficients in order to locate spikes. However, regardless of the great performances, it depends on the setting of empirical parameters. Thus to avoid the dependence of empirical parameters, an algorithm has been developed to automatically estimate optimal parameters from the data for signal filtering and denoising of ECG is performed by PCA and adaptative thresholding. From the obtained denoised signal TWA alternans has been detected and estimated by spectral method.