Digital heart twins predict dangerous rhythm risk without invasive tests

Ventricular tachycardia, a dangerously fast heart rhythm, is a leading cause of sudden cardiac death. Assessing who is at risk remains difficult: current clinical approaches are described by the researchers as either inaccurate, invasive or expensive. A new study in Medical & Biological Engineering & Computing proposes a non-invasive computational alternative built on patient-specific digital models of the heart.

The team, led by Olivier Crabbé with collaborators including Pascal Haigron and Karim El Houari, presents a tool called a pacing outcome predictor, or POP. It is designed to quickly identify the pacing conditions needed to induce ventricular tachycardia in three-dimensional personalized cardiac computer models. In other words, instead of testing a patient in the electrophysiology laboratory, the algorithm searches the digital twin for the stimulation settings that would trigger the arrhythmia.

Building such models faces three major limitations, according to the authors: the computational cost of running detailed electrophysiology simulations, inaccurate geometry of the arrhythmogenic substrate such as scar tissue, and unknown patient-specific tissue electrophysiology properties. The POP addresses the cost problem directly, and the researchers use it to test ventricular tachycardia induction across multiple model versions of the same patient, covering variations in scar geometry and in tissue electrophysiological properties.

An added layer of model reduction further accelerates the calculations. The results from all these model variants are then combined into a personalized ventricular tachycardia risk metric based on in-silico inducibility across the patient’s set of models. Rather than relying on a single simulation, the metric reflects how consistently the arrhythmia can be provoked across a range of plausible representations of that patient’s heart.

The approach was evaluated retrospectively in a cohort of 10 patients with repaired Tetralogy of Fallot. In this small cohort, the metric showed complete separation between patients with and without clinically inducible ventricular tachycardia, correctly classifying all 10 of 10 patients. The authors note this suggests potential for non-invasive ventricular tachycardia risk assessment in this population.

Because the study is retrospective and involves only 10 patients, the findings should be read as an early proof of concept rather than a validated clinical tool. The work remains a modeling study; no patient outcomes were altered by the predictor.

By combining faster prediction algorithms, uncertainty over scar geometry and tissue properties, and a risk metric that aggregates across model variants, the researchers aim to make in-silico arrhythmia testing practical enough for clinical use.

Subject of Research: Non-invasive prediction of ventricular tachycardia inducibility using personalized cardiac digital twins

Article Title: Predicting in-silico ventricular tachycardia induction on fuzzy cardiac digital twins

Article References: Crabbé, O., Bessière, F., L’Eplattenier, P., Grivot, C., Ditac, G., Rochette, M., Martins, R., Haigron, P., & El Houari, K. (2026). Predicting in-silico ventricular tachycardia induction on fuzzy cardiac digital twins. Medical & Biological Engineering & Computing. https://doi.org/10.1007/s11517-026-03699-6

Image Credits: AI Generated

DOI: 10.1007/s11517-026-03699-6

Keywords: ventricular tachycardia, cardiac digital twins, electrophysiology, sudden cardiac death, Tetralogy of Fallot, risk stratification, computer modeling, scar geometry, model reduction, pacing outcome predictor, Predicting, in-silico