Impact of anatomical variations in ventricular shape on non-invasive electrocardiographic imaging

Non-invasive electrocardiographic imaging (ECGI) combines body surface electrocardiographic information with anatomical data to estimate electrical dynamics of the heart on the surface or transmurally across the depth of the myocardium. As a standard practice, ECGI techniques rely on anatomically detailed heart and torso models derived from high quality tomographic data. This practice not only imposes high demands on the image quality and processing, it is also associated with some degree of variations and uncertainty due to image quality, inter/intra-individual variability in segmentation and the differences in segmentation techniques. Though the importance of global anatomical parameters on the accuracy of ECGI solutions are established, the role of local variations in anatomical details remains unknown. In this study, we address this problem by designing an approach to statistically analyze the impact of local variations in ventricular models on the diagnostic accuracy of ECGI methods. It is achieved by developing a set of ventricular models with locally different anatomical details from trained statistical shape models, followed by performing a statistical hypothesis test of equivalence on the ECGI outputs. Two of the existing ECGI methods on epicardial and transmural potential imaging are used in our phantom and real-data experiments. Both experiments for two ECGI methods report statistically equivalency of ECGI diagnostic accuracy except for practically irrelevant differences on ventricular models with local variations. This finding is important in changing the standard practice and facilitating the clinical translation of ECGI research.