ID: 4342839 AUTO-CONTOUR LA - A SEMANTIC SEGMENTATION APPROACH TO LEFT ATRIAL NAVIGATION IN INTRACARDIAC ECHOCARDIOGRAPHY

Intracardiac echocardiography (ICE) is critical in guiding interventional cardiology procedures targeting the left atrium (LA), such as ablation and device implantation. Current methods rely on manual segmentation, resulting in variability, prolonged procedures, and inaccuracies. We developed an automated semantic segmentation pipeline to produce left atrial anatomical contours using deep learning. Integrating catheter positional and orientation (P&O) data enables real-time projection of segmented LA anatomy into 3D space, significantly reducing manual interaction and operator variability.