Effect of noise on frequency and time domain methods of T wave alternans

The purpose of this study is to objectively evaluate the defection accuracy of T Wave Alternans (TWA) methods with a simulation platform and different simulated noise sources including muscle noise, baseline wunder, beat alignment error and possible physiological TWA shift modulated by RR interval variations. We designed several new methods for time-domain TWA evaluation including a new clustering analysis and a TWA time-series analysis. We demonstrated that the alignment error which can be caused by beat detection error and/or by beat-to-beat physiological variations can be one of the most severe noise sources for microvolt TWA. We also demonstrated that a new preprocessing method based on principal component analysis (PCA) can compensate the TWA alignment error and improve detection accuracy by more than 50% for both frequency and time domain methods.