Qian Tao1, Esra Gucuk Ipek2, Rahil Shahzad1, Floris F. Berendsen1, Saman Nazarian2, and Rob J. van der Geest1
1Department of Radiology, Leiden University Medical Center, Leiden, Netherlands, 2Department of Cardiology, The Johns Hopkins University School of Medicine, Baltimore, MD, United States
The extent and distribution of left
atrial (LA) scar, visualized by LGE MR, can provide important information for
treatment of atrial fibrillation (AF) patients. However, in current practice,
to extract such information requires substantial manual effort and expertise.
In this study, a fully automatic method was developed to segment LA and PV’s in
LGE-MRI, combining robust multi-atlas segmentation and flexible level-set based
segmentation optimization. The method demonstrated comparable accuracy to
manual segmentation, with improved 3D continuity. The method enables automated
generation of patient-specific LA and PV geometry models, and potentially objective
LA scar assessment for individual AF patients.