Computational model predicts the hemodynamic response of patients following atrial fibrillation

Atrial fibrillation (AFib) is a cardiac disorder in which the chambers of the heart beat rapidly and irregularly. It's the most common type of arrhythmia and the leading cardiac cause of stroke. 

While several treatments-ranging from medication to surgery-exist, the search continues for improved options to address AFib, which the National Institutes of Health (NIH) forecasts will affect up to 12 million people in the U.S. by 2050. 

One emerging strategy includes electrical stimulation, known as neurostimulation, which researchers believe could potentially suppress, treat, or even reverse the disorder.

At the experimental level, neurostimulation has been found to help not only AFib, but heart failure with reduced ejection fraction and hypertension. However, when it was applied in large clinical studies, the results were underwhelming. Part of the problem is dosage. Researchers didn't have a way to know how much stimulation to give. They were flying blind without real-time feedback on how the body was responding."

Oluwasanmi Adeodu, postdoctoral researcher in Lehigh University's Department of Chemical and Biomolecular Engineering

In a paper published in the latest issue (October 29, 2025) of PLOS ONE, Adeodu and his team present a model that features the human cardiovascular system, the processing centers in the brain that control the heart, and the information pathway linking the heart to the brain. The model was designed to predict the hemodynamic response of patients following an onset of AFib and, ultimately, can help answer questions about where and at what levels neurostimulation should be applied, thereby moving the approach closer to becoming an accepted form of personalized treatment for AFib. 

"This was a translational exercise," says Adeodu, who was lead researcher on the paper, whose co-authors include Mayuresh Kothare, the R. L. McCann Professor of Chemical and Biomolecular Engineering and Associate Dean for Research in Lehigh's P.C. Rossin College of Engineering and Applied Science. "As engineers, we took what clinicians know about AFib-and all the physiological changes that occur in patients-and turned those facts into math. Our goal was to answer the question: Can our model match known effects of AFib on easy-to-measure hemodynamic quantities such as blood pressure and heart rate? If the answer was yes, we could use it to explore new connections."

The answer was yes. The team validated their model against clinical data and found that its predictions of heart rate, stroke volume, and blood pressure matched what doctors see in real patients. One especially interesting result: the model flagged a part of the atrioventricular node-a structure in the heart-as a strong candidate for stimulation. That same area is already a target for ablation therapy, suggesting the model is on the right track.

Now, Adeodu says, the stage is set for further study. Researchers can use the Lehigh team's model as a tool to explore where-and how much-neurostimulation can alter the elevated and irregular heart rate, or improve the compromised baroreflex sensitivity associated with AFib. They can leverage the model's predictive power to test the effects of stimulating different parts of the cardiovascular system, without relying on animal or human subjects. Once they optimize their approach using the model, they can test those strategies in patients. As clinicians use the model, their feedback will help further refine it.

Kothare (corresponding author) sees this paper as the successful result of a collaborative effort by co-authors Dr. Raj Vadigepalli, formerly of Thomas Jefferson University, now at the University of New Mexico; Michelle Gee, a chemical engineering graduate student at the University of Delaware; and Dr. Babak Mahmoudi of Emory University. The project was the outcome of a $2.2 million NIH grant through the Stimulation of Peripheral Activity to Relieve Condition (SPARC) program that was focused on developing software and modeling tools for optimizing the delivery of neurostimulation signals to peripheral nerves to treat conditions such as cardiac arrhythmia, hypertension, and stomach and bladder disorders. Kothare and Mahmoudi co-led the project, which concluded in 2023–2024. 

"The main advantage of this model is that it is computationally tractable, unlike more complex three-dimensional cardiac models that require high performance computing infrastructure to solve the resulting equations," says Kothare. "This low computational cost makes it particularly attractive for rapid testing and even real-time use, allowing bidirectional flow of data and information between the patient and the virtual representation of the cardiac system of the patient, in the true framework of a 'digital twin.'" 

Ultimately, says Adeodu, the long-term goal is an automated, wearable device that continuously monitors physiological feedback and delivers the appropriate stimulation to counter AFib. For now, translating a serious condition into math and lines of code has brought clinicians closer to truly personalized cardiac care.

"Once you have a good model, it opens up a whole new world of insights and connections," says Adeodu. "It's classic proof that when people come together from different fields, amazing things can happen."