Skip to Main Content
In this work, a new method for ectopics removal from ECG recordings is proposed. The method distinguish between normal and ectopic beats through a forward/backward level windowing strategy. Next, by using an adaptive correlation index, the 15 most similar ectopic beats to the one under cancellation are clustered, thus serving to get their eigenvector matrix by singular value decomposition. Finally, the highest variance eigenvector is taken as the template for ectopic beat cancellation. In order to evaluate the proposed method performance, an index called reduction ectopic rate (RER) was defined. The proposed technique has been compared with previously published average QRST cancellation techniques. Twenty 5 hour-length segments extracted from Holter ECG recordings of 20 different AF patients, with a high percentage of ectopic beats, were used in the study. Results provided maximum RER values of 5.76 and minimum of 1.67 being, in average, of 3.57 ± 0.25. In contrast, when ectopic beats were considered as normal beats, maximum RER was 1.49 and minimum 0.88, being in average 1.11 ± 0.25. To sum up, the proposed method is able to improve notably the AA extraction from Holter ECG recordings by reducing notably the residua provoked by the presence of ectopic beats.