Feasibility of a novel algorithm for automated reconstruction of the left atrial anatomy based on intracardiac echocardiography

Background Intracardiac echocardiography (ICE) is frequently used to guide electrophysiology procedures. The novel automated algorithm Cartosoundfam is a model-based algorithm which reconstructs a 3D anatomy of the left atrium (LA) based on a set of 2D intracardiac echocardiography (ICE) frames, without the need to manually annotate ultrasound (US) contours. Objective The aim of this study was to determine the feasibility of the Cartosoundfam module in routine clinical setting. Methods We included 16 patients undergoing LA mapping/catheter ablation. Two-dimensional US frames were acquired from the right atrium (RA) and the right ventricular outflow tract. The Cartosoundfam map was validated in two steps: (1) identification of anatomical structures (pulmonary veins [PV] and LA body and appendage) by alignment of the ablation catheter to the automated map; and (2) analysis of the automated lesion tags (Visitag) location in relation to the PV antrum of the Cartosoundfam map in nine patients with paroxysmal atrial fibrillation (AF) undergoing first time pulmonary vein isolation (PVI). Results Mean 2D US frames per patient were 29 +/- 6 and acquisition time was 16 +/- 4 min. All anatomical structures were correctly identified in all patients (step 1). In the step 2 validation, the median distance to the map was 2.0 (IQR: 2.4) mm and the majority of the Visitags were classified as satisfactory (69%) but all PV segments had some Visitags classified as unsatisfactory. Conclusion The automated ICE-based algorithm correctly identified the LA anatomical structures in all patients with a 69% anatomical accuracy of the Visitags alignments to the PV antrum segments.