Fast screening algorithm for electrical alternans in ECG based on dynamical pattern recognition framework

In this paper, a new framework is presented to analyze the dynamical alternans pattern in ECG. As the beginning work, a brief screening algorithm is described to decrease the time consumption for the framework.The electrical alternans, especially the T wave alternans (TWA), is considered as the danger symptom to sudden cardiac death (SCD). The current methods identifying TWA are reviewed summarily. They are concluded into three types, i.e. time-domain methods, frequency-domain methods and statistical tests. All of the time-domain methods can be induced to dynamical pattern recognition framework. Based on the framework, the comparison between T waveforms is in fact to calculate the similarity of each other. Thus the similarity measures which are appropriate for time series are discussed. Because the alternans pattern concerns only the neighbor waves, it is feasible to eliminate the T waves without alternans by comparing consecutive three or four T waves. With a group of simple inequations, the fast screening algorithm can get rid of about half of the data from further calculation and comparison. For a time series with length N, the computation complexity of constructing a similarity matrix is N*N, and that of fast screening is about 2*N. Finally, the algorithm´s effect is illustrated by several records downloaded from open database Website.