The impact of ventricular shape variations on inverse electrocardiography: A feasibility study

Abstract

Inverse electrocardiography (IECG) estimates cardiac electrical dynamics from body surface electrocardiographic data. As a common practice, all existing IECG problems are solved on anatomically-detailed heart and torso models derived from tomographic images of individual subjects. This practice constitutes a major obstacle to clinical translation of IECG methods, imposing high demands on the quality and processing of medical images. Because anatomical modeling is always associated with variations due to different factors such as image quality and segmentation methods, we design a novel and systematic approach to statistically quantify the impact of ventricular shape variations on the diagnostic accuracy of IECG methods. We propose a novel use of statistical shape modeling to account for the variations in subject-specific anatomical modeling, and from it to generate ventricular models with controlled variations, whose relation to the variations of IECG outputs are then statistically assessed. In this study, we test the feasibility of the proposed approach considering two existing IECG methods for epicardial potential reconstruction and transmural action potential imaging. Both phantom and real-data experiments report statistical equivalency of IECG diagnostic accuracy on ventricular models with local variations. This study demonstrates the feasibility of the proposed approach to be generalized to establish the proper level of anatomical details needed in ventricular modeling, which has the potential to change the common practice and facilitate the clinical translation of IECG research.