Sensitivity of T-Wave Alternans Identification Algorithms to Residual Physiological Noise Affecting the ECG after Preprocessing.

To address the issue as to how and at what extent physiological noise that survives preprocessing affects TWA detection and quantification, a test was performed here on the fast-Fourier-transform spectral method (FFTSM), modified-moving-average method (MMAM), and adaptive-match-filter method (AMFM). These methods were applied to four synthetic ECG tracings respectively affected by no TWA, stationary TWA, and time-varying TWA. Absence and presence of physiological noise (from the MIT-BIH noise stress test database from the PhysioNet web site) were considered. Our results indicate that the FFTSM is robust to noise but has an intrinsic limitation in the precision of timevarying TWA quantification. Noise significantly affects TWA detection and quantification by the MMAM, while the AMFM offers a good compromise between robustness to noise and ability to identify both stationary and time varying TWA.