Comparison Among Microvolt T-wave Alternans Detection Methods and the Effect of T-wave Delimitation Approaches

Microvolt T-wave Alternans (MTWA) is considered an early marker for subjects with high risk sudden cardiac death. Classical MTWA quantification (CM) requires accurate T-wave peak detection risk mark and approaches based on morphological T-wave analysis have been proved to be effective, such as Hilbert Transform approach (HT). This work proposes the hypothesis that accurate T-wave detection may lead to improvements on MTWA quantization. Three distinct methods were testes for T-wave identification: fixed-window approach, adjustable-window, and triangle area approach. These methods were applied on ECG signals from Physionet T-wave Alternans database. The results are compared using linear correlation and area above ROC curves. Also, the results were evaluated by positive predictive value (PPV) and the negative predictive value (NPV), sensitivity, specificity and accuracy. Results showed high dependency between HT and T-wave detection methods and the results variate as the detection method changes. The results indicate that CM, as expected, is not T-wave detection dependent and obtained statistical similar results for the three approaches.