Estimating Septum Rotation to Improve Accuracy of Cardiac Biventricular Bullseye Representation Using the UNISYS Algorithm

Nicolas Montagne1, Amael Mombereau2, Olivier Bernus3, Laura R Bear4
1IHU Liryc, 2Univ. Bordeaux, INSERM, CRCTB, U 1045, IHU Liryc, 3IHU Liryc, Inserm, CRCTB, U1045, Univ. Bordeaux, F-33000, 4IHU-LIRYC


Abstract

Background: Visualizing and comparing electrophysiological data obtained from the whole heart's surface can be challenging and has led to the development of tools such as the bullseye plot generated by the UNISYS algorithm. However, this representation is sensitive to the septum rotation towards the apex, complicating the localization of regions of interest. This study introduces a way of correcting septum rotation to facilitate the comparison and interpretability of bullseye plots. Methods: 4 male pig hearts underwent fluoroscopy to track their septum and calculate their rotation angle along the height of the ventricles. These hearts' shapes were transformed in bullseye plots by the UNISYS algorithm using different techniques: 3 new techniques based on septum rotation correction, and a reference technique. These 4 techniques have been evaluated with tracking performed by 2 different users, 3 times each, resulting in 96 bullseyes. Performance was assessed by measuring the distance between the septum and the middle of the bullseye to ensure they are properly centered, and reproducibility was evaluated on both septum tracking and final epicardial segmentation on the bullseye with percent agreement. Results: The 3 new techniques significantly improved septum centering compared to the reference technique (a deviation of 5-6% vs 18±4% of the total width of the bullseye respectively). This correction was most efficient near the apex (6-fold decrease in error compared to a 3-fold decrease near the base). Septal tracking was reproducible both intra- and inter-user (mean percent agreement > 0.9). All techniques were also reproducible for bullseye segmentation (mean percent agreement > 0.75), although the reference technique contained more outliers (6.7% with percent agreement <0.5 vs 0% with the other techniques). Conclusion: These new methods can improve left-right separation on bullseye plots, allowing for a better interpretability and identification of areas of interest on a simple whole heart's 2D representation.