Principal component analysis based method for detection and evaluation of ECG T-wave alternans

The method for detection and evaluation of T-wave alternance in ECG was elaborated for monitoring of the status of the patients in the Intensive Care Unit of Cardiology Clinics. 24 h ECG recordings registered mostly in patients after myocardial infarction were used for elaboration of the method. Data preprocessing included ECG structural analysis, respiration and/or other factors caused baseline wander removal and T-wave duration correction using modified Bazettpsilas formula. The arrays of samples representing T-wave of one cardio cycle in all ECG leads were concatenated into one array. Such arrays from all cardiocycles formed two dimensional array of samples, representing all samples of T-waves of all cardio cycles in the recording. Principal component analysis method was used to reduce the dimensionality of the set of samples, concentrating representation of interrelated variables into principal components. T-wave alternance usually was represented by one or mostly by few principal components. Amplitude of specific alternance of the coefficients of these principal components used to represent T-wave of every cardio cycle gives the quantitative estimate of the phenomena. Cluster analysis of these coefficients could reveal the variety of T-wave morphology during T-wave alternance. The method was tested on the ECG recordings from CinC/Physionet Challenge 2008 database.