Sensor-enabled 3D electroanatomical model optimization in in-vivo porcine and benchtop models

Abstract Background/Introduction Multielectrode mapping catheters enhance the speed and accuracy of 3-dimensional data collection. However, with variable catheter configurations, data acquisition and efficiency may differ. We sought to compare the HD Grid X mapping catheter to a multi linear 8-spline catheter (spline catheter) with respect to chamber modeling speed and accuracy in in-vivo and bench setups. Methods In-vivo: In 6 pigs, 4 physicians used each catheter in a randomized order to create unedited models of the left atrium and ventricle (LA and LV) during coronary sinus atrial pacing. A sensor-enabled ablation catheter marked five locations per chamber while applying 10-15g of force. These markers, which denoted the ‘true’ location of the endocardial surface, were compared to the nearest surface location defined by each mapping catheter. Bench: In a tank setup designed to emulate an LA to pulmonary vein (PV) transition, chamber models were created during robotic manipulation of each mapping catheter. Resulting models were compared to a "gold-standard" model generated using robotic manipulation of sensor-enabled ablation catheters. Additionally, a static shape analysis was performed with each catheter held by a fixture in a deformed (HD Grid X catheter-bent 90o degrees; spline catheter-3 splines bunched) and undeformed state. Images of the physical catheters were captured to determine the true catheter shape and were compared to the renderings from the respective electroanatomical mapping systems (EnSite X v3.1 and Carto v7.2). Results In-Vivo: The mean error between the ablation catheter generated points and the models for HD Grid X trended lower than for the spline catheter by 11.7% (p=0.63; 2.57mm, 2.91mm) in the LA. In the LV, the mean error for HD Grid X was 34.4% (p=0.04; 2.04 mm, 3.67mm) less than for the spline catheter. In the LA, time for model creation for HD Grid X trended higher than for the spline catheter by 14.0% (p=0.34; 524s, 451s). However, in the LV, modeling time for HD Grid X was 22.9% (p=0.05; 368s, 478s) less than for the spline catheter. Bench: Compared to the ablation catheter created models, HD Grid X had a median bloat of 1.70mm (5.91mm max) while the spline catheter had 2.53mm (11.45mm max). The RMS error (mm) for the static shape analysis of electrode positions in an undeformed state was 0.60 for HD Grid X and 2.26 for the spline catheter. In a deformed state, the error was 0.82 for HD Grid X and 2.07 for the spline catheter. Conclusion In this translational model, the HD Grid X catheter created faster and more accurate unedited LV model geometry than the spline catheter, with no statistically significant difference noted in the LA. Bench data further supports the enhanced geometry accuracy seen in-vivo with HD Grid X, which may result in superior anatomical delineation, and faster mapping data acquisition times.