Using a new time-independent average method for non-invasive cardiac potential imaging of endocardial pacing with imprecise thorax geometry

Cardiac electrical imaging from body surface potentials is a technology with great potential for pre-procedure planning in the context of ventricular ablation, based on body surface measurements of arrhythmic beats. Three clinically desirable properties of such an imaging system are the ability to localize endocardial as well as epicardial initiation sites, the ability to use fewer body surface leads than typical in a body surface mapping system, and the ability to maintain accuracy while limiting dependence on extensive anatomical imaging. At the same time,in the setting of premature ventricular beats, it is typically easy to obtain measurements of multiple beats with the same initiation site. Since sensitivity to measurement noise makes increased signal SNR desirable, multiple beats offer the possibility of improved accuracy. Here we compare standard ensemble averaging of the body surface measurements, reconstruction of individual beats followed by averaging of the results, and an averaging method developed by our group that is less sensitive to timing and propagation velocity variability from beat to beat, in the context of our recently reported method for endocardial pacing site localization using limited torso imaging. For a set of pacing site on the RV and LV endocardia of 3 subjects, we recorded multiple trials of 120 lead ECG and carried out inverse reconstructions using all three averaging methods. The reconstructed heart potentials were then used to estimate the pacing sites that were then validated against recorded pacing locations. The solutions from our time insensitive averaging method show improvement in localization accuracy over the solutions obtained with ensemble averaging, although the average of single-beat reconstructions results in better precision for most pacing sites.