Signal processing subsystem validation for T-wave alternans estimation

Since alternans phenomena in the cardiac repolarization have been shown to be related to arrhythmogenesis, a number of sophisticated methods have been proposed to detect and estimate microvolt T-Wave Alternans (TWA). However, their robustness with respect to the inclusion and tuning of the processing stages has not always been analyzed and quantified in detail. We propose a procedure based on bootstrap techniques to study the effect of some relevant preprocessing stages in a TWA estimation system. A controlled data base was obtained by adding noise and TWA to control ECG signals. Several experiments were performed, each one to evaluate the influence of one characteristic of a processing stage in the whole TWA estimation system. For the analysis, different statistics (median, confidence interval width, and power) were obtained for the TWA amplitude estimation errors. It can be concluded that interactions among different preprocessing subsystems are complex, not always completely characterized, and small variations can affect significantly to the overall performance of the detection system.