Aim. Cardiac resynchronization therapy (CRT) is a key treatment method for patients with heart failure, however, 30-40% of patients do not demonstrate clinical improvement and remain non-responders. The successful implementation of CRT depends significantly on the placement of the electrodes particularly the left ventricular (LV) lead which is limited by the individual anatomy of the coronary sinus (CS) veins. Although CS anatomy varies widely among individuals comprehensive studies on its geometric variability are limited with most research based on autopsied human hearts. This study aims to analyse variations in CS anatomy.
Methods. A cohort of 107 patients underwent cardiac computed tomography with contrast enhancement using a proprietary developed protocol followed by manual segmentation of the heart and CS. To standardize 3D data based on CS voxels, projections were made onto the LV epicardial surface. A universal coordinate system defined on the LV was then used to assign "base-apex" (position along the long axis of the LV) and "circle" (circumferential rotation around the long axis of the LV) coordinates to each CS point. The Wasserstein distance function (Wp) was employed to quantify differences between CS geometries. We estimated pairwise distances between two CS projections for all possible pairs from the dataset, and distances characterizing the deviation of each CS from the entire dataset were identified. Distributions of these distances were then constructed for each metric.
Results. The distributions followed a log-normal pattern. For pairwise distances, the median and variance of the Wp distributions were 0.43±0.08. For distances between individual CS and the whole dataset: 0.46±0.08.
Conclusion. Our numerical results align well with the empirical knowledge of the high variability in CS anatomy. Furthermore, the proposed approach can be utilized for numerically estimating the similarity of datasets across a broad patient population.