3D Reconstructions from Animal Models in the ECGi Tikhonov Estimation at Sinus Rhythm

AngĂ©lica Drielly Quadros1, Gabrielli Ribeiro Cavagnolli1, Italo Sandoval Ramos de Oliveira2, Saleem Ullah3, Vinicius de Paula Silva4, Joao Salinet5
1UFABC, 2Samsung R&D Brazil, 3Center for Engineering, modelling and applied social sciences (CECS), Federal University of ABC (UFABC), 4Universidade Federal do ABC, 5HEartLab, Federal University of ABC


Abstract

Introduction: Studies in animal models have been crucial in deepening the understanding of arrhythmogenic mechanisms, particularly when paired with advanced imaging techniques like ECGi. The torso-tank model for rabbits showed an innovative approach to studying cardiac activity in a more con-trolled setting. In this study we investigate the feasibility of our 3D heart reconstruction geometry and recently developed torso-tank model at recon-structing epicardial electrograms in sinus rhythm.

Methodology: Hearts from New Zealand rabbits were isolated and perfused using a Langendorff system. Heart geometries were segmented and recon-structed based on 101 2D images acquired during a 360 degree-rotation. The 3D heart geometry was then compared with the obtained through MRI seg-mentation done on the Slicer 3D and similarity was evaluated using Dice coefficient. Epicardial electrical activity was reconstructed based on torso-tank signals from 60 electrodes, during sinus rhythm, solving the Boundary Element Method and Tikhonov regularization. Epicardial electrical activity was acquired simultaneously with torso-tank signals with 16-multielectrode arrays places on the atria and ventricle. The electrode position of the 3D reconstruction was extracted with its epicardium filtered signal and com-pared to the correspondent vertex in the estimated signal from a 4.2 second record with 8 activations from each area of the heart (right atrium, left atri-um and ventricle).

Results: Fifteen 3D heart reconstruction were performed including epicardial contact unipolar electrodes localization. The Dice coefficient between our 3D heart geometry reconstruction and through the MRI images was 0.74. Atrial and ventricle presented good signals similarity when compared with those measured directly on the epicardium.

Conclusion: 3D heart geometry presented a good similarity when compared with the obtained through the MRI images segmentation. Reconstructions signals presented a good shape similarity. We are studying the impact of different preprocessing and postprocessing techniques on the reconstruction similarities under different rhythms.