Spatial coherence: A new method of quantifying myocardial electrical organization using multichannel epicardial electrograms

A new technique has been developed to quantify the organization of myocardial electrical activity. The spatial coherence technique employs the use of the magnitude-squared coherence (MSC) spectrum to analyze multichannel electrograms obtained during cardiac mapping. In this study, MSC values for all possible pairs of electrograms recorded from an epicardial plaque consisting of 112 electrodes were computed and systematically integrated to form a three-dimensional coherence surface, that is, a graphical representation of average coherence values versus the spatial orientation of one electrode to another. From this surface, two-dimensional graphs of average coherence versus electrode separation distance were derived, and the data were fitted to an exponentially decaying curve. Two novel parameters indicative of myocardial organization were then extracted from the curves, the coherence length parameter and the coherence plateau parameter. Higher values for these parameters were hypothesized to reflect greater levels of organization of the rhythm. A total of 164 mapping sessions were performed on nine dogs. Three-second data segments of normal sinus rhythm (NSR) and ventricular fibrillation (VF) were analyzed by using spatial coherence. Coherence lengths were found to be significantly longer (24.3 ± 13.4 vs 0.165 ± 0.053 mm, P = .0001) and coherence plateau values significantly higher (0.896 ± 0.104 vs 0.098 ± 0.008, P = .0001) for NSR than for VF. In addition, one instance of ventricular tachycardia, a rhythm more organized than VF, had coherence parameter values greater than those for VF but less than those for NSR (coherence length = 1.23 mm, coherence plateau value = 0.597). These results suggest that spatial coherence may be an effective means of quantifying the electrical organization of a particular cardiac rhythm. The advantages of this technique are (1) it extracts a single parameter from the vast amount of data that is generated in a typical mapping study, thus allowing easy characterization of different cardiac rhythms in terms of their level of organization, and (2) it does not require activation time detection, a process that is difficult when studying arrhythmias such as ventricular fibrillation.