Multi-modal characterization of the left atrium by a fully automated integration of pre-procedural cardiac imaging and electro-anatomical mapping.

Background: The combination of information obtained from pre-procedural cardiac imaging and electro-anatomical mapping (EAM) can potentially help to locate new ablation targets. In this study we developed and evaluated a fully automated technique to align left atrial (LA) anatomies obtained from CT- and MRI-scans with LA anatomies obtained from EAM. Methods: Twenty-one patients scheduled for a pulmonary vein (PV) isolation with a pre-procedural MRI were enrolled. Additionally, a recent computed tomography (CT) scan was available in 12 patients. LA anatomies were segmented from MRI-scans using ADAS-AF (Galgo Medical, Barcelona) and from the CT-scans using Slicer3D. MRI and CT anatomies were aligned with the EAM anatomy using an iterative closest plane-to-plane algorithm. Initially, the algorithm included the PVs, LA appendage and mitral valve anulus as they are the most distinctive landmarks. Subsequently, the algorithm was applied again, excluding these structures, with only three iterative steps to refine the alignment of the true LA surface. The result of the alignments was quantified by the Euclidian distance between the aligned anatomies after excluding PVs, LA appendage and mitral anulus. Results: Our algorithm successfully aligned 20/21 MRI anatomies and 11/12 CT anatomies with the corresponding EAM anatomies. The average median residual distances were 1.9 ± 0.6 mm and 2.5 ± 0.8 mm for MRI and CT anatomies respectively. The average LA surface with a residual distance less than 5.00 mm was 89 ± 9% and 89 ± 10% for MRI and CT anatomies respectively. Conclusion: An iterative closest plane-to-plane algorithm is a reliable method to automatically align pre-procedural cardiac images with anatomies acquired during ablation procedures.