The morphological classification of heartbeats as dominant and non-dominant in ECG signals

Abstract

Surface electrocardiography (ECG) is the art of analyzing the heart's electrical activity by applying electrodes to certain positions on the body and measuring potentials at the body surface resulting from this electrical activity. Usually, significant clinical information can be obtained from analysis of the dominant beat morphology. In this respect, identification of the dominant beats and their averaging can be very helpful, allowing clinicians to carry out the measurement of amplitudes and intervals on a beat much cleaner from noise than a generic beat selected from the entire ECG recording. In this paper a standard clustering algorithm for the morphological grouping of heartbeats has been analyzed based on K-means, different signal representations, distance metrics and validity indices. The algorithm has been tested on all the records of the MIT-BIH Arrhythmia Database (MIT-BIH AD) obtaining satisfying performances in terms of averaged dominant beat estimation, but the results have not been fully satisfactory in terms of sensitivity and specificity. In order to improve the clustering accuracy, an ad hoc algorithm based on a two-phase decision tree, which integrates additional specific knowledge related to the ECG domain, has been implemented. Similarity features extracted from every beat have been used in the decision trees for the identification of different morphological classes of ECG beats. The results, in terms of dominant beat discrimination, have been evaluated on all annotated beats of the MIT-BIH AD with sensitivity = 99.05%, specificity = 93.94%, positive predictive value = 99.32% and negative predictive value = 91.69%. Further tests have shown a very slight decrement of the performances on all detected beats of the same database using an already published QRS detector, demonstrating the validity of the algorithm in real unsupervised clustering situations where annotated beat positions are not available but beats are detected with a high-performance beat detector.

Unsupervised ECG Analysis: A Review

“…Most ECG clustering studies detect the peak of the QRS complex, called the R peak, across the signal and consider the interval between two consecutive R peaks, i.e., the whole cardiac cycle, as the segmentation unit. There are very few studies that have considered other characteristic points of the ECG signal for segmentation [33]- [35]. Given that the abnormal morphologies of PR, ST, and TP segments of an ECG cycle can indicate common cardiac disorders, such as myocardial ischemia, hypokalemia, and atrial fibrillation [32],…”

“…[8], [33], [35], [74], [78]- [87] Fuzzy c-means Handles overlapped clusters. Handles noise and outlier.…”

PcHD: Personalized classification of heartbeat types using a decision tree

“…ECG traces of heartbeats measure the two main activities of the heart: ventricular, characterized by the QRS complex and T waves; and atrial, characterized by P waves [20]. Analysis of the QRS complex remains the simplest noninvasive method of diagnosing a variety of heart diseases.…”

“…De Chazal et al [6,7], Rodriguez et al [9], Llamedo and Martinez [10], De Oliveira et al [12], Yeap et al [13], and Zhang et al [14] used waveform features to detect the QRS complex, while Osowski et al [8] used higher-order statistics and Hermite coefficient features that can be effective for modeling cumulative indicators, but complicated signals need to be approximated by a linear combination of these functions. Chiarugi et al [20] and Almeida et al [21] proposed techniques for detecting P wave, characteristics of atrial activity after the detection of the QRS complex.…”