Combined method for detection of atrial late potentials

Abstract

The work is devoted to improvement of methods for noninvasive identification of low-amplitude components of electrocardiogram (ECG) - atrial late potentials (ALP) which are markers of potentially dangerous heart rhythm disorders. A combined method for ALP detection based on wavelet analysis, decomposition in the basis of eigenvectors and classification by neural network is proposed. As the result of the numerical experiment the dimension of ALP feature vector was minimized that made it possible to distinguish between 2 classes "norm - no ALP" and "pathology - ALP are present" with minimum error classification.