An Unsupervised Method to Detect the Left Atrial Appendages and Classify their Morphologies

Abstract

The left atrial appendage (LAA) is the site where the left atrial thrombi are most likely (90%) to develop. Despite the increasing interest that LAA has attracted over the last decade, the methods currently used to classify its morphology are mainly based on cardiologists’ judgment. Given the remarkable improvement of imaging techniques, we propose an unsupervised quantitative method that can overcome the limits of the current classification systems. The resulting classification system is objective and reproducible.