For patients living with ventricular tachycardia, the most dangerous part of the disease is often hidden inside electrically damaged heart tissue. Ventricular tachycardia, or VT, begins when abnormal electrical signals race through the lower chambers of the heart, sometimes forming a self-sustaining circuit that overwhelms the normal heartbeat. Catheter ablation is widely used to interrupt these circuits, but identifying the precise tissue responsible can be difficult, particularly in hearts scarred by a previous heart attack or other disease. A new study led by A. M. Qadri, U. Rohrer and F. O. Campos, published in Nature Cardiovascular Research, reports that computational models can offer more useful guidance when they acknowledge what they do not know. By incorporating uncertainty into virtual simulations of VT, the researchers show how personalized heart models could become more reliable tools for planning ablation procedures.
Modern cardiac modeling attempts to recreate a patient’s heart in software. Medical images such as magnetic resonance imaging or computed tomography can be used to reconstruct the shape of the ventricles and identify regions of scar tissue. Electrical properties are then assigned to different areas of the digital heart, allowing researchers to simulate how an impulse travels through healthy muscle, damaged tissue and the narrow channels that may sustain an arrhythmia. These models can generate maps of possible VT circuits and suggest where radiofrequency energy or another ablation technology should be delivered. The promise is significant: instead of relying only on measurements collected during an invasive procedure, clinicians could enter the operating room with a patient-specific forecast of the most important targets.
The difficulty is that no computational heart model is a perfect replica of a living patient. Medical images have limited resolution, especially when they are used to distinguish thin strands of surviving muscle within a scar. The exact electrical behavior of diseased tissue is also difficult to measure directly. Researchers must estimate values such as conduction speed, electrical excitability and the connection between cells. Small changes in these assumptions can alter the simulated path of an electrical wave. A model may therefore produce a convincing map of one possible circuit while overlooking another equally plausible route. Treating a single simulation as definitive can create a false sense of precision, potentially directing ablation toward tissue that is associated with the arrhythmia but is not essential to maintaining it.
Qadri and colleagues addressed this problem by treating the model’s inputs as uncertain rather than fixed. Instead of building one virtual heart with one set of electrical properties, the approach evaluates a family of models representing the range of conditions that could be consistent with the available patient data. This type of analysis can involve systematically varying tissue characteristics, anatomical interpretations and electrophysiological parameters, then observing whether the predicted arrhythmia behaves consistently. The result is not simply a map of where VT might occur, but a map of confidence: areas repeatedly implicated across simulations are more robust candidates, while predictions that appear only under a narrow set of assumptions are recognized as fragile.
That distinction could transform the way clinicians interpret computational predictions. A conventional model might identify a long list of locations where an electrical wave could circulate. An uncertainty-aware model can rank those locations according to how stable the prediction remains when the underlying assumptions change. If a critical isthmus—a narrow pathway that allows a re-entry wave to persist—appears in many plausible simulations, it may represent a strong ablation target. If the apparent pathway shifts dramatically from one model to another, the result can alert the clinical team that additional mapping or imaging is needed. In this way, uncertainty is not treated as a weakness to hide, but as information that can improve decision-making.
The technical challenge is particularly important because VT is a re-entry phenomenon. In a healthy ventricle, an electrical impulse spreads rapidly and then disappears as the tissue temporarily becomes unable to respond. In a scarred heart, islands of surviving cells can be surrounded by electrically inactive tissue. Signals may be forced through slow, narrow channels and return to regions that have recovered their excitability, creating a circular wave that repeats itself. Computer models solve mathematical descriptions of this process across a three-dimensional representation of the ventricle. They calculate how voltage changes across the tissue over time and how local conduction interacts with scar architecture. Because the behavior is highly sensitive to microscopic structure, uncertainty in anatomy and tissue properties can have consequences at the scale of a clinical treatment.
The study’s central message is that the best ablation strategy may not be the one suggested by a single “most likely” simulation. Instead, it may be the strategy that remains effective across many plausible versions of the patient’s heart. This is a form of robust planning. Researchers can test whether a proposed lesion interrupts the predicted electrical circuit under different parameter settings, anatomical interpretations or activation patterns. A target that consistently disrupts VT across the ensemble of models may be more dependable than one that succeeds only in an idealized scenario. Such an approach could also help reduce unnecessary ablation, since it focuses attention on tissue with the strongest and most reproducible relationship to the arrhythmia.
The findings arrive as cardiac electrophysiology increasingly moves toward personalized, image-based treatment. During an ablation procedure, clinicians already combine electroanatomical mapping, imaging and physiological testing to locate abnormal tissue. Computational models could add a forward-looking layer, allowing teams to explore possible interventions before applying them to the patient. Yet for this technology to influence routine care, it must be transparent about its limitations. A probability map or confidence score is more clinically meaningful than an apparently exact prediction that conceals the assumptions behind it. The researchers’ uncertainty-aware framework therefore aligns with a broader movement in biomedical artificial intelligence and simulation: models should communicate not only their answer, but also how dependable that answer is.
The work does not mean that software can replace invasive mapping or clinical judgment. VT circuits can be dynamic, and the electrical behavior recorded during a procedure may differ from what was inferred from images obtained earlier. Drug effects, changes in heart rhythm and the stimulation protocols used in the laboratory can all influence what is observed. Instead, the computational approach is best understood as a decision-support system that helps clinicians prioritize hypotheses and identify where uncertainty is greatest. Its greatest value may come when the model and the patient’s real-time electrical data disagree, prompting a closer investigation rather than an automatic commitment to either source.
If validated in broader patient populations and integrated into clinical workflows, uncertainty-aware modeling could make personalized VT ablation more efficient and more precise. It could help electrophysiologists distinguish essential parts of an arrhythmia circuit from neighboring scar that is merely associated with it, potentially reducing procedure time and limiting damage to functional myocardium. The approach also offers a framework for future digital-heart technologies, including simulations of other rhythm disorders and individualized testing of treatment strategies. The study’s most important lesson is straightforward but powerful: in medicine, acknowledging uncertainty can produce better predictions than pretending it does not exist. For patients facing a life-threatening rhythm, that change in attitude could turn virtual hearts into more trustworthy guides for restoring a safer one.
Subject of Research: Computational modeling of ventricular tachycardia and uncertainty-aware guidance for catheter ablation
Article Title: Accounting for uncertainty in computational models of ventricular tachycardia improves ablation guidance
Article References: Qadri, A.M., Rohrer, U., Campos, F.O. et al. Accounting for uncertainty in computational models of ventricular tachycardia improves ablation guidance. Nat Cardiovasc Res (2026). https://doi.org/10.1038/s44161-026-00856-w
Image Credits: AI Generated
DOI: https://doi.org/10.1038/s44161-026-00856-w
Keywords: ventricular tachycardia, cardiac ablation, computational cardiology, virtual heart models, uncertainty quantification, electrophysiology, arrhythmia, personalized medicine, cardiovascular research
Tags: advances in virtual heart modeling techniqueschallenges in locating arrhythmogenic tissuecomputational modeling in cardiac arrhythmiaimpact of model uncertainty on cardiac interventionimproving ablation success with computational toolspersonalized heart models for VT treatmentrole of medical imaging in VT ablation planningscar tissue identification in heart imagingsimulation-based electrical activity mapping in heartuncertainty incorporation in cardiac simulationsventricular tachycardia ablation guidancevirtual simulations for ventricular tachycardia
