Semi-automatic left atrial appendage segmentation from 3D CCTA images

Left Atrial Appendage closure is a relatively novel percutaneous procedure for patients suffering from Atrial Fibrillation (AF). LAA closure has only recently been approved for clinical use and preparations for closure procedures are in most hospitals performed by manual segmentation of Coronary CT Angiography (CCTA) images. We present a novel semiautomatic LAA segmentation method using two seed points (starting and stopping points for segmentation) and manually selected threshold. The method performs the visualization of the segmented LAA and enables a doctor to select 3 points which define a plane in 3D space (the ostium) where the LAA will be occluded. Performance of the proposed method is evaluated on 8 CCTA datasets manually segmented by a cardiovascular surgeon. Proposed method achieves an average DICE coefficient of 91.7% on evaluation datasets.