A Bayesian filtering application for T-wave alternans analysis

T-wave alternans (TWA) is the periodic fluctuation in temporal and spatial characteristics of T-wave morphology in every other heartbeat in electrocardiography (ECG) signal. Due to its microvolt amplitude and background noises, sophisticated signal processing techniques are required for its detection and estimation. This paper presents a three step TWA detection and estimation strategy which consists of filtering the ECG signal using a variant of Kalman filter (KF), segmenting the ST-T wave based on ECG phase and applying the spectral method to detect TWA and estimate its value. By virtue of the very nature of the strategy i.e., KF in time and SM in frequency domain, the technique provided significant advantages in terms of detection and estimation of TWA. Detection performance is compared in terms of probability of detection and estimation performance in terms of relative bias, standard deviation and mean square error.