Benchmark for Algorithms Segmenting the Left Atrium From 3D CT and MRI Datasets

Abstract

Knowledge of left atrial (LA) anatomy is important for atrial fibrillation ablation guidance, fibrosis quantification and biophysical modelling. Segmentation of the LA from Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) images is a complex problem. This manuscript presents a benchmark to evaluate algorithms that address LA segmentation. The datasets, ground truth and evaluation code have been made publicly available through the http://www.cardiacatlas.org website. This manuscript also reports the results of the Left Atrial Segmentation Challenge (LASC) carried out at the STACOM'13 workshop, in conjunction with MICCAI'13. Thirty CT and 30 MRI datasets were provided to participants for segmentation. Each participant segmented the LA including a short part of the LA appendage trunk and proximal sections of the pulmonary veins (PVs). We present results for nine algorithms for CT and eight algorithms for MRI. Results showed that methodologies combining statistical models with region growing approaches were the most appropriate to handle the proposed task. The ground truth and automatic segmentations were standardised to reduce the influence of inconsistently defined regions (e.g., mitral plane, PVs end points, LA appendage). This standardisation framework, which is a contribution of this work, can be used to label and further analyse anatomical regions of the LA. By performing the standardisation directly on the left atrial surface, we can process multiple input data, including meshes exported from different electroanatomical mapping systems.

A competitive strategy for atrial and aortic tract segmentation based on deformable models

“…Furthermore, the obtained result with the proposed methodology (with and without competition) was compared with the 5 best works assessed in the current benchmark (total of 9), namely a region growing formulation with rough contour initialization using an atlas-based approach (LTSI-VRG) (Sandoval et al, 2013), a probabilistic atlas approach (LUB-SRG) proposed by (Stender et al, 2013), a marginal space learning strategy with (SIE-MRG) (Tobon-Gomez et al, 2015) or without refinement (SIE-PMB) (Zheng et al, 2008) through graph-cuts, and a multi-atlas approach with global and local transforms for LA segmentation only (UCL-1C) (Zuluaga et al, 2013). Inter-observer variability is also available.…”

“…More details about the segmentation can be obtained in (Bourier et al, 2016). In order to identify the atrial body and similarly to the proposed in (Tobon-Gomez et al, 2015), multiple bounding boxes were generated around the pulmonary veins, vena cava and left and right atrial appendage (LAA and RAA). All bounding boxes were drawn around the ostia of each structure.…”

“…Only LA segmentation methods are allowed to be evaluated with the current benchmark. Further details about acquisition and ground truth generation are indicated in Table 1 and in (Tobon-Gomez et al, 2015).…”

“…Similarly to section 3.2, the current data was obtained from an available benchmark (Tobon-Gomez et al, 2015). Again, 30 datasets were acquired, 10 datasets are used as training, while the remaining 20 cases are used for the algorithm evaluation.…”

“…atrial bodies and aortic tract -Ao) through medical imaging has gained particular interest during the last decade. Several authors extracted/isolated multiple contours of the atrial anatomies using different segmentation strategies (Tobon-Gomez et al, 2015), proving the clinical relevance of a correct anatomical and functional assessment of each atrial region for global cardiac function quantification and even for risk stratification (Hoit, 2014;Melenovsky et al, 2014). Moreover, enhanced minimally invasive cardiac interventions were proposed, using electroanatomic mapping techniques (Rolf et al, 2014) or even superimposing pre-operative anatomical atrial surfaces (Bourier et al, 2016), extracted from highly detailed datasets, into intra-operative imaging (e.g.…”