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

TL;DR

This study tackles the problem of inferring subject-specific ventricular activation from noninvasive surface ECGs, revealing that multiple activation maps can produce the same 12-lead QRS but that physiological priors on Purkinje–muscle junctions and a gradient-based Geodesic-BP calibration enable accurate, clinically credible digital twins. The authors couple an anisotropic eikonal forward model with lead-field ECG computation, pseudo-bidomain torso simulations, and PMJ-based constraints to identify initiation sites $X_0$ that reproduce observed signals, while also exploring an ensemble of solutions to quantify uncertainty. Key findings show that, in unrestricted settings, activation maps may be non-unique despite excellent ECG fits; imposing subendocardial constraints and increasing observation density substantially reduce tau variability and improve endocardial activation fidelity, though exact uniqueness remains elusive. The work demonstrates that noninvasive, high-fidelity cardiac twins can be calibrated efficiently and credibly, with implications for precision cardiology, while highlighting areas for improvement, such as explicit modeling of the Purkinje network and Bayesian uncertainty quantification.$\,$

Abstract

Digital twins for cardiac electrophysiology are an enabling technology for precision cardiology. Current forward models are advanced enough to simulate the cardiac electric activity under different pathophysiological conditions and accurately replicate clinical signals like torso electrocardiograms (ECGs). In this work, we address the challenge of matching subject-specific QRS complexes using anatomically accurate, physiologically grounded cardiac digital twins. By fitting the initial conditions of a cardiac propagation model, our non-invasive method predicts activation patterns during sinus rhythm. For the first time, we demonstrate that distinct activation maps can generate identical surface ECGs. To address this non-uniqueness, we introduce a physiological prior based on the distribution of Purkinje-muscle junctions. Additionally, we develop a digital twin ensemble for probabilistic inference of cardiac activation. Our approach marks a significant advancement in the calibration of cardiac digital twins and enhances their credibility for clinical application.

Figures (11)

• Figure 1: Geodesic-BP: fast definition of a from the surface . Geodesic-BP grandits_digital_2023 optimizes the distribution of until the mismatch between recorded and simulated is minimized. Physiological constraints for the are automatically imposed during the optimization process.
• Figure 2: Endocardial surface restrictions limits the domain of possible locations for . The are restricted to lie inside the subdomain $S$ which is spanned by a band, equidistant to the endocardium $S_e$.
• Figure 3: The in silico solution generated from the physiologically-detailed . The map $\tau$ is shown together with the in blue (top). The bottom row shows the corresponding, calculated 12-lead .
• Figure 4: Electrode locations of the 12-lead (left) compared to configurations comprising $32$, $64$ and $128$ electrodes from an anterior (top) and posterior (bottom) view. For each electrode the uni-polar lead field was computed using the as reference. Electrodes for computing the are shown in red.
• Figure 5: Identifiability problems of inverse methods. Left panel: Using Geodesic-BP, the identification of optimal locations and timings to fit a given () can be achieved with high fidelity. The distribution obtained from 20 optimization runs with different initializations forms a tight envelope around the ECG to be reconstructed (left, $\mathcal{V}_\mu \pm \mathcal{V}_\sigma$). Right panel: The two samples with the largest difference in are shown that reveal significant variability in the solution space which is indicative of limited identifiability in the general unrestricted case. We highlight a few of the major differences in (red dashed ellipses). This prompts for quantification of this variability and for a-priori constraints to reduce the non-uniqueness of the solution space.