An efficient end-to-end computational framework for the generation of ECG calibrated volumetric models of human atrial electrophysiology

TL;DR

This work delivers an automated, end-to-end pipeline for generating high-fidelity volumetric biatrial models with fiber architecture and a universal coordinate framework, embedded within a torso conductor to enable realistic ECG predictions. It introduces a flexible, cable-based representation of inter-atrial conduction pathways, a rule-based labeling system for anatomical structures, and a scalable two-stage workflow (anatomical twinning and functional modeling) to create digital twins and virtual cohorts. The framework supports real-time or near-real-time forward modeling with both full bidomain and lead-field approaches, enabling parameter sweeps and P-wave–driven calibration under clinically realistic constraints. Demonstrated on 50 patient datasets, the method achieves automated mesh generation with high quality, efficient computation, and controllable inter-atrial conduction effects, advancing precision cardiology toward scalable digital twins and personalized therapy planning.

Abstract

Computational models of atrial electrophysiology (EP) are increasingly utilized for applications such as the development of advanced mapping systems, personalized clinical therapy planning, and the generation of virtual cohorts and digital twins. These models have the potential to establish robust causal links between simulated in silico behaviors and observed human atrial EP, enabling safer, cost-effective, and comprehensive exploration of atrial dynamics. However, current state-of-the-art approaches lack the fidelity and scalability required for regulatory-grade applications, particularly in creating high-quality virtual cohorts or patient-specific digital twins. Challenges include anatomically accurate model generation, calibration to sparse and uncertain clinical data, and computational efficiency within a streamlined workflow. This study addresses these limitations by introducing novel methodologies integrated into an automated end-to-end workflow for generating high-fidelity digital twin snapshots and virtual cohorts of atrial EP. These innovations include: (i) automated multi-scale generation of volumetric biatrial models with detailed anatomical structures and fiber architecture; (ii) a robust method for defining space-varying atrial parameter fields; (iii) a parametric approach for modeling inter-atrial conduction pathways; and (iv) an efficient forward EP model for high-fidelity electrocardiogram computation. We evaluated this workflow on a cohort of 50 atrial fibrillation patients, producing high-quality meshes suitable for reaction-eikonal and reaction-diffusion models and demonstrating the ability to simulate atrial ECGs under parametrically controlled conditions. These advancements represent a critical step toward scalable, precise, and clinically applicable digital twin models and virtual cohorts, enabling enhanced patient-specific predictions and therapeutic planning.

Figures (14)

• Figure 1: Schematic outline of the end-to-end framework for the generation of -calibrated volumetric models of patient-specific human atria. After image acquisition, the workflow comprises nine steps: automatic multilabel segmentation, automatic label augmentation on the blood pools, extrusion of the volumetric bilayer walls, automatic selection of atrial orifices, automatic annotation of anatomical structures and fiber generation, generation of , registration and generation of a torso volume conductor, generation of s, and cardiac electrophysiological simulation and P-wave generation. We moreover provide the paper section index where each step is detailed.
• Figure 2: (a) Point-wise curvature on the blood pool surface mesh. The figure also highlights the correlation between the surface region of high curvature and anatomical structures to mark. (b) Minimal set of labels selected based on surface curvature, including , and on the , and , and on the . (c) Instance of solution and isosurfaces on the blood pool volumetric mesh. (d) Vein ostiae and the discardable tissue labeled on the blood pool volumetric mesh.
• Figure 3: Left: Mesh quality computed on the biatrial model. Right: Boundaries on the and 15 problems employed to determine atrial anatomical structures and tissues known to have different properties.
• Figure 4: (a) Identification and labeling of the atrial orifices. Different labels are used for endocardial and epicardial tissue layers. (b) Anatomical structures identified on the biatrial anatomy, including , , s, , and part of the . (c) Generated rule-based atrial fiber architecture. (d) Final biatrial volumetric mesh, augmented with labeled anatomical structures.
• Figure 5: (a) Distribution of the coordinate $\alpha$, representing the $$-to- coordinate for the $$, and the lateral-to-septal coordinate for the . (b) Distribution of the coordinate $\beta$, representing the lateral-to-septal coordinate for the , and the posterior-to-anterior coordinate for the . (c) Distribution of the coordinate $\gamma$, representing the endocardial-to-epicardial coordinate for both atria. (d) Projection of the labels on the space generated by the before (left) and after (right) the solution of the linear elasticity problem.