Deep Learning Formulation of ECGI for Data-Driven Integration of Spatiotemporal Correlations and Imaging Information

The challenge of non-invasive Electrocardiographic Imaging (ECGI) is to re-create the electrical activity of the heart using body surface potentials. Specifically, there are numerical difficulties due to the ill-posed nature of the problem. We propose a novel method based on Conditional Variational Autoencoders using Deep generative Neural Networks to overcome this challenge. By conditioning the electrical activity on heart shape and electrical potentials, our model is able to generate activation maps with good accuracy on simulated data (mean square error, MSE = 0.095). This method differs from other formulations because it naturally takes into account spatio-temporal correlations as well as the imaging substrate through convolutions and conditioning. We believe these features can help improving ECGI results.

[1]  Felipe Alonso Atienza,et al.  Inverse Problem of Electrocardiography: Estimating the Location of Cardiac Ischemia in a 3D Realistic Geometry , 2015, FIMH.

[2]  Maxime Sermesant,et al.  Fast Personalized Electrophysiological Models from CT Images for Ventricular Tachycardia Ablation Planning , 2018 .

[3]  Hervé Delingette,et al.  Noninvasive Personalization of a Cardiac Electrophysiology Model From Body Surface Potential Mapping , 2017, IEEE Transactions on Biomedical Engineering.

[4]  Max Welling,et al.  Auto-Encoding Variational Bayes , 2013, ICLR.

[5]  Kawal S. Rhode,et al.  A Fast-Marching Approach to Cardiac Electrophysiology Simulation for XMR Interventional Imaging , 2005, MICCAI.

[6]  Max Welling,et al.  Semi-supervised Learning with Deep Generative Models , 2014, NIPS.

[7]  Linwei Wang,et al.  Generative Modeling and Inverse Imaging of Cardiac Transmembrane Potential , 2018, MICCAI.

[8]  Thomas Pock,et al.  Learning a variational network for reconstruction of accelerated MRI data , 2017, Magnetic resonance in medicine.

[9]  Yves Coudière,et al.  Improving the Spatial Solution of Electrocardiographic Imaging: A New Regularization Parameter Choice Technique for the Tikhonov Method , 2017, FIMH.

[10]  Christopher Burgess,et al.  beta-VAE: Learning Basic Visual Concepts with a Constrained Variational Framework , 2016, ICLR 2016.

[11]  Y. Rudy,et al.  Electrocardiographic Imaging: II. Effect of Torso Inhomogeneities on Noninvasive Reconstruction of Epicardial Potentials, Electrograms, and Isochrones , 2001, Journal of cardiovascular electrophysiology.

[12]  Maxime Sermesant,et al.  Fast personalized electrophysiological models from computed tomography images for ventricular tachycardia ablation planning. , 2018, Europace : European pacing, arrhythmias, and cardiac electrophysiology : journal of the working groups on cardiac pacing, arrhythmias, and cardiac cellular electrophysiology of the European Society of Cardiology.

[13]  Y. Coudière,et al.  Inverse Problem of Electrocardiography: estimating the location of cardiac isquemia in a 3D geometry , 2017 .

[14]  Hervé Delingette,et al.  Transfer Learning From Simulations on a Reference Anatomy for ECGI in Personalized Cardiac Resynchronization Therapy , 2019, IEEE Transactions on Biomedical Engineering.

[15]  Laura Bear,et al.  Effect of the torso conductivity heterogeneities on the ECGI inverse problem solution , 2015, 2015 Computing in Cardiology Conference (CinC).

[16]  Nicholas Ayache,et al.  Non-Invasive Personalisation of a Cardiac Electrophysiology Model from Body Surface Potential Mapping. , 2016, IEEE transactions on bio-medical engineering.