T-wave alternans search over 24 hour holter ECG recordings based on singular value decomposition

This paper proposes an efficient method to search for T-wave alternans (TWA) over 24 hour Holter ECG recordings. After appropriate pre-processing to remove baseline drift and artifact, data are segmented to 2 minute successive time intervals. For each beat in the segment, singular value decomposition is applied to derive orthogonal characteristic signals. Then two prominent orthogonal signals are used for the TWA search. A pair of alternans indices is defined for each beat as the orthogonal waveform distance between the target beat and the adjacent two beats. When alternans presents, the first index will be larger than the second index. The periodogram of the sequence of alternans indices in each segment yields a useful alternans measure named Alternans Ratio (AR). To show the effectiveness of the measure, the method is applied to 25 control and 24 data from patients with various cardio vascular disorders. AR distribution showed prominent differences among subject groups. It has been demonstrated that the measure AR is not only useful to detect the presence of TWA but the AR distribution can be used for the stratification of the TWA risk.