Abstract 2245: Process-Dependent Accuracy of 3D Image Integration For Atrial Fibrillation Ablation: Comparison of CartoMerge with 3D ICE Imaging
Background: Current technologies to fully “register” surrogate maps onto 3D CT (CartoMerge, CM) or 3D ICE images (CartoSound, CS) have emerged to guide ablation. Image guided ablation presupposes that surrogate images reflect actual real-time anatomy. This assumption has never been tested nor accuracy of CM and CS compared.
Methods: To evaluate the accuracy of the CM and CS images, 97 anatomic marking clips were pre-implanted in the atria/ventricles in 12 dogs. The target position accuracy of merged 3D CT or 3D ICE images was assessed by comparing location of clips established by real-time 2D ICE based-navigation (true clip site: gold standard) and apparent location on the 3D CT image or 3D ICE (test image).
Results: The CS position error of true clip site-vs-rendered 3D ICE clip image was 1.4±1.7 mm/1.4±1.7 mm for atria/ventricles. Since the CS image accuracy was fully guaranteed, the pre-acquired 3D CT images were merged with 3D ICE images. Using “landmark merge” to minimize distance of the landmark pairs (true clip site vs. 3D CT clip image), the CM position error was 2.9±1.1 mm/4.2±1.6 mm for atria/ventricles (average surface mismatch of 3D CT-vs-3D ICE images: 2.3±0.5 mm/2.8±0.6 mm for atria/ventricles). The “surface merge” (fit of 3D CT surface to surface contour of 3D ICE) minimized the surface mismatch to 1.7±0.4 mm/2.1±0.7 mm for atria/ventricles (p<0.0001 vs. landmark merge), but enlarged the position error at target clips (5.4±2.5 mm/6.6±3.1 mm, p<0.001 vs. landmark merge). Finally, after manual “adjusted merge” to minimize distance of each clip as landmark pairs and surface mismatch, position errors at each clip for atria/ventricles were 3.3±1.6 mm/4.8±2.0 mm with surface mismatches of 2.0±0.5 mm/2.7±0.7 mm, respectively.
Conclusion: Navigational mismatch of merged 3D CT images was significantly greater in ventricular than atrial chambers. Mismatch using 3D ICE images was reduced to 1.5 mm for both atrial and ventricular chambers. Despite the efforts of different merge processes, the overall CM position errors were larger than the CS errors, and are especially problematic for ventricular navigation with the concurrent use of “surface merge” algorithms. This information is critical for the effective employment of CT and ICE rendered volume imaging.