Researchers have used artificial intelligence (AI) to evaluate patients’ electrocardiograms (ECGs) in a targeted strategy to screen for atrial fibrillation (AF). In the digitally-enabled, decentralised study, AI identified new cases of AF that would not have come to clinical attention during routine care.
Earlier research had already developed an AI algorithm to identify patients with a high likelihood of previously unknown AF. The algorithm for detecting AF in normal sinus rhythm from an ECG is licensed to Anumana, an AI-driven health technology company, by nference and Mayo Clinic.
“We believe that atrial fibrillation screening has great potential, but currently the yield is too low and the cost is too high to make widespread screening a reality,” says Peter Noseworthy (Mayo Clinic, Rochester, USA), lead author of the study which was published in The Lancet. “This study demonstrates that an AI-ECG algorithm can help target screening to patients who are most likely to benefit.”
The study enrolled 1,003 patients for continuous monitoring and used another 1,003 patients from usual care as real-world controls. The findings showed that AI can indeed identify a subgroup of high-risk patients who would benefit more from further intensive heart monitoring to detect AF, supporting an AI-guided targeted screening strategy.
The AI algorithm can identify patients who, even though they are in normal rhythm on the day of the ECG, may have an increased risk of undetected episodes of AF at other times. Such patients can then undergo additional monitoring to confirm the diagnosis.
“Traditional screening programs select patients based on age (65 or older) or the presence of conditions such as high blood pressure. These approaches make sense because advanced age is one of the most important risk factors for atrial fibrillation. However, it is not feasible to repeatedly conduct intensive heart monitoring in more than 50 million older adults across the country,” says Xiaoxi Yao, the senior author of the study.
“The study shows that an AI algorithm can select a subgroup of older adults who might benefit more from intensive monitoring. If this new strategy is broadly implemented, it could reduce undiagnosed AF, and prevent stroke and death in millions of patients across the globe,” says Yao.
The next step in this research will be a multicentre hybrid trial focused on the effectiveness of implementing the AI-ECG workflow in diverse clinical settings and patient populations.
“We hope that this approach will be particularly valuable in resource-constrained environments where the rate of undetected AF may be particularly high, and resources to detect it may be limited. However, more work is needed to overcome barriers to implementation, and further studies must evaluate targeted screening strategies in these environments,” says Noseworthy.
“Now that we have demonstrated that AI-driven AF screening is possible, we will also need to show that patients with screen-detected atrial fibrillation benefit from treatment to prevent stroke,” says Noseworthy, “Our ultimate goal is to prevent strokes. I believe the current study has brought us one step closer.”
