Atrial fibrillation is one of the most common cardiac arrhythmias worldwide. It affects one in ten people over the age of 80, and a quarter of all strokes are associated with it. However, conventional treatment methods face major challenges. Symptoms often recur after initial treatment, and it is not possible to pinpoint exactly where the fibrillation originates. However, each subsequent clinical treatment places a considerable strain on the heart. The research group “Computational Cardiac Modeling” led by Dr. Axel Loewe at the KIT Institute for Biomedical Engineering (IBT)  is working on a solution: modeling the human heart as a system of mathematical equations. In this approach, the organ is individually translated into mathematical equations and computer code. The goal: a controlled research environment in which individual parameters can be specifically modified. “We are trying to make the biologically, physiologically, and medically complex system that is the heart tangible – to clearly describe its functions from a biological perspective and translate them into technical representations,” explains Loewe. This precision offers a decisive advantage in cardiovascular disease research over traditional methods such as animal experiments or in vitro models, where many influencing factors are difficult to control.

From data measurement to real-time prediction

The study, in which Loewe and his team are participating, has been underway since spring 2025 in two phases, in collaboration with the Karlsruhe Municipal Hospital and the University Hospital of Frankfurt. It is currently in the first phase, in which 30 patients from the group affected by relapses are being measured with particular care in order to generate individually personalized computer models: Both the precise anatomy of the atrium and the electrical activity in the heart are recorded using electrodes on the catheters. This data forms the basis for the personalized computer models. Nearly 20 people have already been enrolled, and initial clinical results were presented in April 2026 at the European Heart Rhythm Congress in Paris. However, the path to obtaining high-quality study data is more challenging than expected: technical difficulties or unforeseen medical or personal complications can result in planned participants ultimately being unable to be included in the study. Once all patients in the first phase have been enrolled and individually represented in the model, the second phase of the study will begin. Here, the simulation is intended to run in real time while the patient lies on the operating table. This places high demands on the speed and logistics of the simulation system developed by the research group at KIT: “During the procedure, we have only a very limited window of time to retrieve the data from the clinical system, personalize the model, and run the simulation,” explains Loewe. The entire data pipeline must be further optimized for this purpose, on the one hand by accelerating the algorithms and on the other by using specialized accelerator hardware - one of the central technical challenges of the coming months within the study, which will continue until 2028.

A tool with far-reaching implications beyond the study

The simulation environment has potential far beyond the current study - and in several directions. For one thing, the computer model not only allows for testing far more potential trigger points than is clinically possible, but it also enables the investigation of the fundamental mechanisms underlying cardiac arrhythmias: Under what conditions do they arise, and what would need to be changed to prevent them? In addition, the personalized models open up the possibility of virtually testing various treatment options - such as different ablation strategies or medications - in advance. And last but not least, the environment is suitable for developing new medical devices: catheter shapes, electrode spacing, or signal quality can be simulated before a device is even built. For medical technology companies, the simulation environment is therefore particularly valuable: it allows for comprehensive preliminary studies on the computer that would otherwise only be possible with animal or in vitro models. Currently, for example, a collaboration is underway with the Freiburg-based company Stockert, which manufactures generators for ablations. Training AI systems on simulated heart data is also a growing field of application: where clinical data is scarce and expensive, large, cleanly annotated datasets can be generated within the model and used to train AI models - thus saving real patient data for the final evaluation.
 

On the Path to Clinical Approval

The KIT computer model is not yet permitted to directly influence surgical procedures on humans, as the methodology has not yet been approved as a medical device. However, a systematic review under real-world conditions in the second phase of the study would be a decisive milestone on this path: “With prospective validation, a future study could then actually provide model-based recommendations on where ablations would be particularly promising and where they would not.” In the three-year study, data collection in the clinics and modeling at KIT have been underway for about a year. The next steps are clearly defined: complete the first phase, further optimize the computational speed of the simulation pipeline, and finally move on to the second phase – with the long-term goal of firmly establishing computer models in everyday clinical practice. If the results are promising, the next step would be a follow-up project specifically aimed at obtaining regulatory approval as a medical device. Companies interested in helping shape this path and actively supporting the approval process are cordially invited to contact the team. The potential is vast: If atrial fibrillation can be triggered from multiple points, strategically placed lesions could immediately neutralize all of these triggers – a finding that could fundamentally transform ablation practice in cardiology.