Left atrial wall segmentation from CT for radiofrequency catheter ablation planning
Document Type
Conference Proceeding
Publication Date
1-1-2016
Journal
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume
9401
First Page
71
Last Page
78
URL with Digital Object Identifier
10.1007/978-3-319-31808-0_9
Abstract
Atrial fibrillation is the most common cardiac arrhythmia and a major cause of ischemic stroke. It is believed that measurements of the thickness of a patient’s left atrial wall can improve understanding of the patient’s disease state, as well as assist in treatment planning for radiofrequency catheter ablation. Left atrial wall thickness can be measured and visualized from segmented contrast-enhanced cardiac CT images, but segmentation itself is challenging. Here we present a pipeline for segmenting the left atrial wall, using a hierarchical constraint structure in order to distinguish between the atrial wall and other muscular structures. Using this approach, the left atrial wall was successfully differentiated from adjacent structures such as the aortic wall. The method was compared to manual segmentation on ten clinical CT images of patients undergoing radiofrequency catheter ablation for atrial fibrillation. Similarity between the methods, by Dice coefficient, was found to be 0.79, and the rMSE of the epicardial segmentation was found to be 0.86 mm. A roadmap to automation for clinical translation is also presented.