Method for left atrial appendage segmentation using heart CT images

Abstract

In this study, we propose a method to automatically extract the details of the left atrial appendage region from heart CT images in order to facilitate the preoperative planning of Left Atrial Appendage Occlusion. Generally, it is difficult to automatically classify the left atrial appendage region in a heart CT image because the heart is a very complicated organ. Therefore, in addition to the segmentation method using fully convolutional neural networks, we performed an automatic extraction of only the left atrial appendage region using mini-batch and adversarial training. This method was applied to heart CT images made with a contrast medium. With this method, it becomes possible to automatically obtain information necessary for preoperative planning support of left atrial appendage closure from heart CT images.