For this study, researchers developed an AI model called WARN (Atrial Fiber Tension Warning) to help predict atrial fibrillation.

“WARN consists of a deep learning model that inputs short 30-second heart rate segments and outputs the probability of an imminent switch to AFib,” said Dr. Jorge Gonçalves, Professor at the Luxembourg Center for Systems Biomedicine at the University of Luxembourg. states.The authors of this study explained that Today’s medical news.

“The more likely this risk is, the more likely it is to progress to atrial fibrillation. This repeats every 15 seconds. If this probability exceeds a certain threshold, an alert is triggered. ,” Gonsalves said.

The WARN system was trained and tested on 24-hour recordings of ECG data collected using the following methods: holter device Worn by 350 people at Tongji Hospital in Wuhan, China.

“AI has shown amazing performance in looking for patterns in large datasets,” Gonsalves said.

“Such patterns are often very difficult for humans to identify. So we hypothesized whether there is some ‘hidden’ information in the heart rate over the few minutes that leads to the transition to AFib. has been verified. “Deep learning was able to find such clues in 280 records of Holter data from patients who progressed to atrial fibrillation,” he explained.

When testing the WARN model, Gonsalves and his team found that it could predict the transition from normal heart rhythm to atrial fibrillation with about 80% accuracy, with an average of 30 minutes of warning before onset.

“[These findings]are not surprising, because some patients may actually feel their heart beat a little or have an arrhythmia seconds to minutes beforehand.” Gonsalves explained.

“They can’t explain exactly what that sensation is, but if they can feel it, machines should be able to detect this ‘feeling’. And in fact, slight changes in heart rate dynamics can be seen before the onset of atrial fibrillation. ”

“On the other hand, it is surprising that we can detect AFib (within) 30 minutes from a model obtained from 280 patients,” he continued.

“Atrial fibrillation arises from a variety of conditions, and each patient’s disease is unique. Therefore, this model captures the ‘average’ changes in the dynamics of the 280 diseases associated with AFib. “It’s amazing that we got such high performance from a model that averaged the dynamics of so many patients,” said Gonsalves.

The study collected heart electrical activity data through a medical device, and Gonsalves said this same model could one day be used routinely. smart watch.

“Today, smartwatches can already do things like: tell me If you have AFib. “However, for many people, once atrial fibrillation begins, it can be difficult to return it to normal sinus rhythm,” he explained.

“Early warnings allow patients to take preventive measures, such as taking antiarrhythmic and anticoagulant medications early.”

“I envision a not-too-distant future in which simple wearable devices will monitor our cardiovascular systems 24/7, alerting us to subtle changes in its dynamics. These changes. can lead to certain illnesses, such as atrial fibrillation, and life-threatening events, such as a heart attack. In many cases, even a few minutes’ warning can lead to emergency medical care. and possibly prevent the onset of the disease and avoid related complications.”

— Dr. Jorge Gonçalves, lead study author

Next steps for future research include customizing the algorithm for individuals, Gonsalves said.

“The goal from the current model is to have individual patients wear a smartwatch for an extended period of time,” he elaborated.

“Each time we transition from sinus rhythm to AFib, the model learns specific features of that patient’s disease. Ultimately, we can significantly improve both the early warning window and each patient’s performance. .”

“This study showed that there is information at the moment leading to the onset of cardiovascular disease. This was a retrospective proof-of-concept study,” Gonsalves added.

“The next step is to create apps for different smartwatches and test them in prospective studies. I wouldn’t be surprised if related apps appear on Google Play or the Apple Store in the near future.”

MNT We also spoke with Paul Drury, MD, board-certified cardiologist and associate medical director of electrophysiology at MemorialCare Saddleback Medical Center in Laguna Hills, California, about the study.

“Typically, in addition to known triggers, excessive alcohol intake or stress, most atrial fibrillation episodes appear to be random,” Drury explained. “The discovery that there are ECG precursors that can reliably predict an atrial fibrillation episode is surprising.”

“To improve atrial fibrillation treatment, it is important to detect atrial fibrillation before it develops,” Drury continued.

“The ability to warn patients before the onset of atrial fibrillation allows us to treat it before it begins. One example is when a patient is given antiarrhythmic medication “as needed” to abort an episode. It is possible to take it. “This could help reduce emergency room visits and other morbidity that occur as a result of atrial fibrillation,” he said.

Drury said that while smart devices are currently very good at detecting atrial fibrillation, they cannot predict the condition.

“By having this technology available in these devices, patients with atrial fibrillation can be more proactive in their treatment of their atrial fibrillation,” he said. “This could include assisting in the detection of atrial fibrillation triggers so that impending atrial fibrillation episodes can be aggressively treated.”

“We would like to see if this AI technology can be implemented in smartwatches and other wearable technologies to track more patients with atrial fibrillation to see if they can accurately predict atrial fibrillation. Validation from such studies will be needed before recommending specific treatments for impending AF episodes.

— Dr. Paul Drury, Cardiologist