Development of an Algorithm for Clustering Cardiac ECG Signals with Post-Correction for Long-Term ECG Monitoring

Abstract

Introduction. The most common method for diagnosing cardiovascular diseases is the method of ECG monitoring. In order to facilitate the analysis of the obtained monitorograms, special software solutions for automated ECG processing are required. One possible approach is the use of algorithms for automated ECG processing. Such algorithms perform  clustering of cardiac signals by dividing the ECG into complexes of similar cardiac signals. The most representative complexes obtained by statistical averaging are subject to further analysis.Aim. Development of an algorithm for automated ECG processing,  which performs clustering of cardiac signals by dividing the ECG into complexes of similar cardiac signals.Materials and methods. Experimental testing of the developed software was carried out using patient records provided by the Pavlov First State Medical University of St  Petersburg. The software module was implemented in the MatLab environment.Results. An algorithm for clustering cardiac signals with post-correction for the tasks of long-term ECG monitoring and a software module on its basis were proposed.Conclusion.  The presence of a small number of reference cardiac signal complexes, obtained through ECG processing using the proposed algorithm, allows physicians to optimize the process of ECG analysis. The as- obtained information serves as a basis for assessing dynamic changes in the shape and other parameters of cardiac signals for both a particular patient and groups of patients. The paper considers the effect of synchronization errors of clustered cardiac signals on the shape of the averaged cardiac complex. The classical solution to the deconvolution problem leads to significant errors in finding an estimate of the true form of a cardiac signal complex. On the basis of analytical calculations, expressions were obtained for the correction of clustered cardiac signals. Such correction was shown to reduce clusterization errors associated with desynchronization, which creates a basis for investigating the fine structure of ECG signals.