Development and validation of a fully automatic algorithm to align 3D MRI and electro-anatomical mapping anatomies of the left atrium. (19th May 2022)

• Record Type: Journal Article
• Authors: Hermans, BJM; Bijvoet, GP; Holtackers, RJ; Mihl, C; Luermans, JGLM; Vernooy, K; Linz, D; Chaldoupi, SM; Schotten, U
• Is Part Of: Europace. Volume 24:Supplement 1(2022)
• Publication Date: 2022-05-19
• Subjects: Arrhythmia -- Treatment -- Periodicals; Cardiac pacing -- Periodicals; Catheter ablation -- Periodicals; Heart -- Physiology -- Periodicals; Electrophysiology -- Periodicals; 617.4120645
• DOI: 10.1093/europace/euac053.248
• Languages: English
• ISSNs: 1099-5129

Abstract

Abstract: Funding Acknowledgements: Type of funding sources: None. Background: The role of pre-procedural cardiac imaging in the guidance and planning of ablation procedures is becoming increasingly important. Emerging non-invasive techniques such as late gadolinium enhancement magnetic resonance imaging (LGE MRI) and electrocardiographic imaging (ECGi) can potentially help to locate ablation targets prior to the ablation procedure. To be able to integrate LGE MRI and ECGi information into targeted ablation procedures, a reliable alignment between cardiac imaging and electro-anatomical mapping (EAM) is required. Purpose: To develop and evaluate a fully automatic technique to align pre-procedural MRI anatomies with EAM anatomies of the left atrium (LA). Methods: Twenty-one patients scheduled for a (re-do) pulmonary vein (PV) isolation with a 3D pre-procedural LGE MRI were enrolled in this study. LA anatomy was segmented from the MRI dataset using ADAS-AF. During the ablation procedure LA anatomy was recorded with an HD-grid (Ensite) or Pentaray catheter (CARTO). The MRI segmentation and EAM were performed by different cardiologists blinded for each other's results. Anatomies of both MRI and EAM were aligned using an iterative closest point-to-plane algorithm in custom-made software in Matlab 2021a. With this algorithm, the distance between MRI anatomy voxels (=points) and the surface of the EAM anatomy (=plane) is minimized by translating and rotating the MRI anatomy until the total residual distance is minimized. The result of the alignment is quantified by calculating the Euclidian distance between the aligned anatomies after excluding PVs and the mitral anulus. Results: The algorithm was successfully applied in 18/21 patients (n=11 CARTO, n=7 Ensite). In the remaining 3 patients, the algorithm could not align the anatomies because of a large difference in LA volume or PV anatomy between the two techniques. In the analysed patients, the average distance between anatomies was 2.7±0.77mm. The top of Figure 1 shows the alignment of the anatomies with the smallest (patient A) and the largest (patient B) residual distances as well as the distances between these anatomies for both patients (right) with purple ≤2.5mm and red ≥5.0mm. The distributions of distances (bottom left) show that, after alignment most of the MRI anatomy is closer than 5mm from the EAM anatomy in every patient. On average, 87.6±10.4% of the atrial surfaces showed distances below 5.0mm between the two anatomies and 55.1±13.2% of the surfaces was within 2.5mm from each other. Results did not differ between Ensite and CARTO anatomies. Conclusion: LA anatomy obtained from 3D LGE MRI can automatically and reliably be aligned with LA anatomy recorded during an ablation procedure with an EAM system using an iterative closest point-to-plane algorithm.