Standardized regionalization of the atria for 3D cardiac imaging, electroanatomical mapping and computational modeling. A multidisciplinary consensus of the PersonalizeAF consortium Open Access

3D imaging and high-resolution electroanatomical mapping have become an integral part of cardiac electrophysiology and the management of patients with arrhythmias. With further technological advances the significance of these modalities continues to grow. However, to perform regional quantitative analyses and intra- and inter-individual, as well as cross-modality comparisons, a universal definition of atrial regions and their boundaries is required. While for the left ventricle there is already an established standardized regionalization (AHA 18-segment model), there is no consensus for the regionalization of the atria.

Here we propose standardized left and right atrial segments based on anatomical, electrophysiological and clinical considerations, with precise definition of regional boundaries allowing for reproducible and automated regionalization.

In a multidisciplinary task force of the European PersonalizeAF consortium involving cardiologists and cardiac electrophysiologists, as well as specialists in cardiac imaging and computational modeling we developed a standardized regionalization dividing the left atrium into eight, and the right atrium into seven segments (15-segment bi-atrial model) (Fig. 1A).

As a proof-of-principle, two software algorithms for automatic regionalization of 3D atrial geometries based on the standardized 15-segment bi-atrial model were developed independently by different working groups of the PersonalizeAF consortium – one based on a commercially available software, the other being open-source.

The algorithm based on the commercial software obtains the 15-segment bi-atrial model by dividing the surface mesh in the defined regions along geodesics using the Fast Marching Method (Fig. 1C). For the open-source solution, a publicly available semi-automatic bi-atrial division pipeline was developed: In this algorithm, after standardizing the surface mesh by remeshing and clipping the pulmonary veins, the orifices are annotated automatically. Based on these anatomical landmarks, the boundaries of each region are inferred by calculating geodesics using Dijkstra’s algorithm.

Both algorithms were able to annotate the regions with high accuracy and very good agreement as indicated by interrater reliability testing (kappa >0.9), in geometries derived from 50 patients and 2 imaging modalities (CT and MRI), thus demonstrating the universal applicability and reproducibility of the standardized segments.