Empirical mode decomposition for improved least square T-wave alternans estimation

More than 80% of emergency patients succumb to cardiac related including Sudden Cardiac Deaths (SCD) in Pakistan whereas the global deaths due to SCD are 30-40% of total deaths. T-wave alternans refer to variation in amplitude, phase or wave shape in alternating T waves in ECG signals and has been linked with the SCD and ventricular arrhythmia. TWA detection is still a challenge as in many cases amplitude of TWA is well below indispensable process noises. Empirical mode decomposition (EMD) is a empirical signal processing technique employed as noise removal tool for nonlinear and non-stationary TWA signals. Recently very matured and developed least square techniques have been employed for the detection of TWA, which were previously not explored as considered limited practicability. Least square works in time domain and better for the non-stationary TWA signals that gives the technique an edge over the other techniques that assume stationary nature of the signals. Least square with higher order polynomial curve fitting has been recently explored as a promising TWA estimation technique for risk stratification of SCD. In this paper we are presenting a novel EMD based least square T-wave alternans estimation scheme which outperforms estimation in 10-40 dB Gaussian noise. The performance comparison of the technique with bench mark modified moving average method (MMAM) for TWA estimation confirms the enhanced performance for the various practical scenarios of TWA.