Computer detection of non-stationary T wave alternans using a new correlation method

Detection of microvolt T wave alternans (TWA) is a non-invasive method to identify patients at risk for sudden cardiac death. ECGs show that visible TWA is often nonstationary. Thus, we developed a new correlation method (CM) for TWA detection, and we tested CM's ability to detect non-stationary TWA in comparison with accepted spectral method (SM). In a simulation study CM and SM were used to evaluate stationary and non-stationary TWA of different amplitude. Other simulated conditions included: background noise, poor synchronization and windowing of the T waves; and amplitude respiration modulation of the T wave. In our comparison of CM and SM, we found that only CM was able to detect non-stationary TWA. CM was more robust to a poor synchronization and windowing of T waves, but affected more by high amplitude modulation than SM. Both CM and SM detected TWA in the presence of background noise.