Applying nonlinear dynamics to ECG signal processing

This article presents a procedure used for describing dynamic electrical processes in the heart as the cause of the origin of the ECG signals. The Takens method of delays enables the authors to demonstrate the electrical activity of the heart in state space. It is difficult to say whether the state-space representation is more suitable for classification of heart disturbances than is the common time representation of the ECG signals. However, the analysis of planar vectorcardiographic loops in frontal, sagital, and horizontal planes, as often done by cardiologists, suggests that state space could be a promising approach. In the case of HRV data, it appears to be very promising. The authors suspect that some heart diseases could be better displayed in state space, but on the other hand, some diseases are rather difficult to recognize in that representation. From the authors' experiments with the correlation dimension of signals from their groups of patients (normal, bigeminy), it follows that the normal ECG signal has the smallest value of the correlation dimension (from 2.1 to 2.8, mean 2.46, STD 0.37). In the case of bigeminy, the correlation dimension increases with the increase of pathology (from 3.2 to 4.6, mean 3.81, STD 0.56). The opposite situation is valid for HRV data. The highest values of the correlation dimension are obtained with signals from healthy persons (15.3-7.6) and these values significantly decrease with pathology (3.7-4.9).