DIVAID: Automatic Division of Bi-Atrial Geometries into Clinically Important Regions

Recent technological advances in cardiac imaging, e.g.cardiac computed tomography (CT), magnetic resonance imaging (MRI) and electroanatomical mapping (EAM) allow detailed analyses of cardiac morphology and function.However, to perform regional quantitative intraand inter-individual, as well as cross-modality comparisons, a standard for the regionalization of the heart is required.We developed a 15-region bi-atrial model based on anatomical, electrophysiological and clinical considerations.Moreover, we devised a pipeline (DIVAID) to automatically divide any bi-atrial geometry according to our proposed definitions.The automatic division results were compared to blinded manual expert regionalizations in 20 bi-atrial geometries from multiple acquisition modalities.The mean regional overlap (Dice score) between both division results across all geometries for both atria combined was 0.92.The mean Euclidean distance between regional boundaries derived from either manual or automatic division in both atria was 1.70 mm.Given the high concordance between both division results, this algorithm enables standardized, consistent, reproducible and operatorindependent regional quantitative comparisons of the atria between multiple modalities, patients and centers, which may facilitate cardiac research towards personalized treatment approaches.