Abstract:The eikonal equation has become an indispensable tool for modeling cardiac electrical activation accurately and efficiently. In principle, by matching clinically recorded and eikonal-based electrocardiograms (ECGs), it is possible to build patient-specific models of cardiac electrophysiology in a purely non-invasive manner. Nonetheless, the fitting procedure remains a challenging task. The present study introduces a novel method, Geodesic-BP, to solve the inverse eikonal problem. Geodesic-BP is well-suited for GPU-accelerated machine learning frameworks, allowing us to optimize the parameters of the eikonal equation to reproduce a given ECG. We show that Geodesic-BP can reconstruct a simulated cardiac activation with high accuracy in a synthetic test case, even in the presence of modeling inaccuracies. Furthermore, we apply our algorithm to a publicly available dataset of a biventricular rabbit model, with promising results. Given the future shift towards personalized medicine, Geodesic-BP has the potential to help in future functionalizations of cardiac models meeting clinical time constraints while maintaining the physiological accuracy of state-of-the-art cardiac models.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to construct a digital twin model of cardiac electrophysiology through surface electrocardiogram (ECG). Specifically, the research objective is to inversely solve the initial conditions of cardiac electrophysiological activities, that is, the cardiac activation sequence, from non - invasive surface electrocardiogram recordings. This process involves solving the so - called "inverse eikonal problem", that is, inferring the internal electrical activity pattern of the heart from known electrocardiogram data. To achieve this goal, the authors propose a new method - Geodesic - BP (Geodesic Back - Propagation), which can efficiently optimize the parameters of the eikonal equation to reproduce a given electrocardiogram. ### Main contributions 1. **A new highly parallelized anisotropic eikonal solver**: Applicable to GPGPU architectures and machine - learning libraries, it can efficiently solve the problem of electrical activity propagation in cardiac electrophysiological models. 2. **A gradient calculation method based on back - propagation**: Associated with the calculation of geodesic paths, it can efficiently calculate gradients during the optimization process. 3. **Efficient algorithm implementation**: Demonstrates how to use the Geodesic - BP method to identify the cardiac activation sequence from non - invasive electrocardiogram recordings. 4. **Robustness verification**: Proves that the inverse - solving process is robust to model perturbations. ### Method overview 1. **Cardiac activation modeling**: - Use the anisotropic eikonal equation to describe the propagation of cardiac electrical activity. - The equation is as follows: \[ \begin{cases} \sqrt{\langle M(x)\nabla\phi(x),\nabla\phi(x)\rangle}=1, & \text{in }\Omega, \\ \phi(x_i) = t_i, & \text{for }(x_i,t_i)\in X_0, \end{cases} \] where \(\Omega\subset\mathbb{R}^d\) is a bounded domain, \(\phi:\Omega\rightarrow\mathbb{R}\) is the activation time map, and \(M:\Omega\rightarrow S_d\) is a symmetric positive - definite tensor field used to describe the orthogonal conduction velocity in the heart. 2. **Efficient solution of local optimization problems**: - Transform local optimization problems into convex optimization problems and solve them using the FISTA algorithm. - Improve the efficiency of the algorithm through parallel computing and pre - computing Lipschitz constants. 3. **Geodesic - BP**: - Use the back - propagation algorithm to calculate the gradient of the eikonal solution with respect to the initial conditions. - Achieve efficient gradient calculation through automatic differentiation techniques. 4. **Electrocardiogram calculation**: - Use the lead - field formula to convert the transmembrane potential of the heart into body - surface potential. - Improve computational efficiency by pre - computing discrete integral operators using the Gaussian quadrature method. ### Experimental results 1. **Simulation study**: - In a simplified 2D model, by optimizing the initial conditions, the electrocardiogram was successfully fitted, and the initial activation conditions could be relatively accurately restored. - The overall root - mean - square error (RMSE) was only 5.8 milliseconds. 2. **Application in actual 3D models**: - Using a publicly available rabbit heart model, by optimizing the initial conditions, the electrocardiogram was successfully matched. - The optimized model achieved a good matching effect on multiple leads, and the loss function converged after about 250 epochs. ### Discussion - **Efficiency**: The Geodesic - BP method exhibits extremely high efficiency on a single GPU and can complete complex optimization tasks in a short time. - **Robustness**: Even in the presence of perturbations in the forward model and lead - field, Geodesic - BP can still accurately restore the cardiac activation sequence. - **Applicability**

Digital twinning of cardiac electrophysiology models from the surface ECG: a geodesic backpropagation approach

Solving the Electrocardiography Inverse Problem by Using an Optimal Algorithm Based on the Total Least Squares Theory

Noninvasive Electroardiographic Imaging: Application Of Hybrid Methods For Solving The Electrocardiography Inverse Problem

Truncated total least squares: a new regularization method for the solution of ECG inverse problems.

Effect of Cardiac Motion on Solution of the Electrocardiography Inverse Problem.

Digital twinning of cardiac electrophysiology for congenital heart disease

Accurate and Efficient Cardiac Digital Twin from surface ECGs: Insights into Identifiability of Ventricular Conduction System

Physiology and Geometry-Informed Deep Learning Framework for 12-Lead ECG Cardiac Transmembrane Potential Recovery

Cardiac Digital Twin Pipeline for Virtual Therapy Evaluation

How Much Geometrical Detail Do We Need In Cardiac Electrophysiological Imaging? A Generic Heart-Torso Representation For Fast Subject-Specific Customization

A Framework for the generation of digital twins of cardiac electrophysiology from clinical 12-leads ECGs

Solving the Inverse Problem of Electrocardiography for Cardiac Digital Twins: A Survey

Digital Twinning of the Human Ventricular Activation Sequence to Clinical 12-lead ECGs and Magnetic Resonance Imaging Using Realistic Purkinje Networks for in Silico Clinical Trials

Cardiac Transmembrane Potential Imaging with GCN Based Iterative Soft Threshold Network

A Statistical Physiological-Model-Constrained Framework for Computational Imaging of Subject-Specific Volumetric Cardiac Electrophysiology Using Optical Imaging and MRI Data

[EM Algorithm for the Inverse Problem of Electrocardiography].

Integration of Different Cardiac Electrophysiological Models into a Single Simulation Pipeline

Probabilistic learning of the Purkinje network from the electrocardiogram

Physically motivated projection of the electrocardiogram

Personalized Topology-Informed 12-Lead ECG Electrode Localization from Incomplete Cardiac MRIs for Efficient Cardiac Digital Twins

A pipeline for developing digital cardiac twins integrating cardiovascular magnetic resonance and electrocardiographic imaging: results from the MyoFit46 study