PGLM: Piecewise Generalized Linear Modeling of Ventricular Repolarization for Estimation of ECG T-wave Alternans

In ECG signals, T-wave alternans (TWA) are attributed to the variations in amplitude, phase, or shape of every alternating T- wave. TWA has been associated with lethal cardiac arrhythmia leading to sudden cardiac death (SCD). In cardiac research, TWA refers to microscopic TWA as its amplitude ranges in microvolts (ţV). Detection and estimation of TWA in embed noises are still challenging as require elaborated signal processing techniques. Cardiac signals are non-stationary and least-squares methods have an edge over spectral methods as signals’ nature is not compromised. This study presents a new piecewise generalized linear modeling (PGLM) for TWA estimation by exploiting Runge’s phenomenon associated with a higher-order polynomial. A comparative study of the proposed PGLM and least-squares polynomial modeling (LSPM) with the modified moving average method (MMAM) as a benchmark method is also presented. The LSPM requires a higher-order polynomial for TWA estimation and suffers signal oscillation at junction points due to Runge’s phenomenon. The proposed model outperforms MMAM for TWA estimation for a noise range of 10–40 dB with TWA amplitude variations 5–100 microvolts for stationary, non-stationary, and phase-changing TWA. The PGLM is computationally simpler and presents a powerful technique for TWA amplitude estimation for the clinical study of cardiac arrhythmia.