Left atrial segmentation from 3D respiratoryand ecg-gated magnetic resonance angiography

Abstract

Magnetic resonance angiography (MRA) scans provide excellent chamber and venous anatomy. However, they have traditionally been acquired in breath-hold and are not cardiac-gated. This has made it difficult to use them in conjunction with late gadolinium enhancement (LGE) scans for reconstructing fibrosis/scar on 3D left atrium (LA) anatomy. This work proposes an image processing algorithm for segmenting the LA from a novel MRA sequence which is both ECG-gated and respiratorygated allowing reliable 3D reconstructions with LGE. The algorithm implements image partitioning using discrete Morse theory on digital images. It is evaluated in the context of creating 3D reconstructions of scar/fibrosis with LGE.