Electrocardiographic P-wave delineation based on adaptive slope Gaussian detection

Abstract

The study of the P-wave on the electrocardiogram is essential in the characterization of atrial conduction defects that may anticipate cardiac pathologies, such as atrial fibrillation. This evidence exhibits the need to develop reliable methods for accurate automatic delineation of P-waves. Many different strategies for delineating P-waves have been introduced. Nonetheless, they all share the same principle of smoothing aggressively the P-wave pattern to facilitate its delineation. However, that strategy may provoke morphological alterations in the P-wave under study that could lead to inaccurate delineation. Alternatively, the present work introduces a new delineation strategy grounded on the generation of a Gaussian model of the P-wave under study to assist its delineation and an adaptive slope threshold that takes into account the morphology of the preceding P-waves. The method was validated using the annotated QT database from Physionet. Delineation results provided a detection sensitivity of 100%, whereas the mean and standard deviation of the delineation error for the P-wave onset, peak and offset were 4.71 ± 9.59 ms, 2.82 ± 6.69 ms and 0.6 ± 9.79 ms, respectively. These results demonstrate that the proposed strategy provides accurate delineation of P-waves that outperforms others presented in the literature, in particular in terms of stability.