Robust and real-time detection of normal QRS using a novel processing method based on the topology of electrocardiogram reconstructed in phase space

Designing automatic and real-time detection systems regarding cardiac diseases first requires an accurate detection of normal QRS complex of the heart's electrical activity (electrocardiogram). This research was conducted with the aim of achieving a topological pattern of normal QRS complex that is able to model the qualitative behavior specific to normal QRS complex from real-time electrocardiogram (ECG) time series. We believe that the geometric structures or topology of the points reconstructed out of ECG signal in phase space include richer information for classification of nonlinear signals and in addition to quantity, it provides the quality of trajectory or orbit of time series in the form of distance and angle in successive points as information of the dynamics governing the system. However, there is shortage of analytical tools for extraction of such information. In this research, a new processing method based on trajectory or orbit tracking reconstructed out of ECG signals in phase space was presented based on geographical directions. Eventually, a unique and special behavioral pattern of normal QRS was extracted and presented. In order to preserve the geometrical structures and the order of points reconstructed from time series in phase space, considering the chaotic nature of ECG signal, which is nonlinear and non-stationary, no noise removal and filtering have been done on the selected ECG. Among other advantages of the proposed algorithm are no application of windowing and segmentation onto the time series. Accordingly, in response to independence of the length of the time series and removing the error related to estimation of data length, one can dispense with preprocessing and it can lead to decreased computational load. Next, using databanks of Fantasia-Database and MIT-BIH Normal-Sinus-Rhythm-Database, the efficiency of the proposed processing method for robust detection of normal QRS was evaluated. The criteria of the positive predictive value, accuracy, sensitivity of 100%, and detection error rate of 0% were calculated.

Keywords:

Normal QRS , Detection , Topology , Phase Space , Real-Time , Robust