A new artificial intelligence (AI) tool could revolutionize the detection of atrial fibrillation, identifying patients at risk before they even show symptoms. This groundbreaking system examines millions of anonymized health records to flag indicators associated with atrial fibrillation (AF), a heart condition resulting from irregular heartbeats, which often leads to increased stroke risk.
Currently, around 1.6 million people are diagnosed with AF across the UK, but the British Heart Foundation (BHF) estimates there are many thousands of individuals who remain undiagnosed. AF can manifest through symptoms such as palpitations, dizziness, shortness of breath, and fatigue, yet many people experience no obvious signs of the condition, leaving them unaware of their heightened stroke risk.
The newly-developed tool, named FIND-AF, has been crafted by scientists and clinicians from the University of Leeds and Leeds Teaching Hospitals NHS Trust, backed by funding from the BHF. Utilizing machine learning, the tool evaluates medical records to assess the likelihood of developing AF based on factors including age, sex, ethnicity, and pre-existing health conditions such as heart failure, high blood pressure, and diabetes.
The algorithm was trained on 2.1 million anonymized health records and validated against another 10 million. Patients identified as high-risk receive portable electrocardiogram (ECG) devices to monitor their heart rhythm, which they use twice daily for four weeks or whenever they feel symptoms. The data collected is transmitted to the trial team for analysis, and if irregular heart rhythms are detected, the patient's General Practitioner (GP) is notified to discuss treatment options.
John Pengelly, a 74-year-old retired Army captain from Bradford, West Yorkshire, participated in the study having initially received a letter inviting him to contribute. Despite not displaying any noticeable symptoms, the AI system flagged him as high-risk, leading to his diagnosis of AF. “If it benefits somebody then great, I want to help,” Pengelly expressed, sharing his experience of the trial.
"They sent me a little digital monitor, and a few times daily, I had to put my thumbs on it for about two minutes to take readings," he recounted. “After a few weeks, I sent the kit back—it was really straightforward. I didn’t have symptoms before, but I was found to have AF. Now I take medication every day to reduce my stroke risk.” This emphasis on early detection showcases the tool's potential to save lives.
Atrial fibrillation disrupts the heart's normal rhythm, causing the upper chambers (atria) to contract irregularly, which can reduce the efficiency of the heart. While the causes of AF are not fully understood, it primarily affects older adults and individuals with chronic conditions. This irregular beating often results from abnormal electrical impulses overpowering the heart's natural pacemakers.
Chris Gale, professor of cardiovascular medicine at the University of Leeds, spoke candidly about the challenges of undiagnosed AF: “All too often, the first sign of someone living with AF is when they suffer from a stroke. This can be devastating, changing lives instantly and resulting in significant costs for healthcare services.”
There are various classifications of AF: paroxysmal (episodic, often resolving without treatment), persistent (lasting longer than seven days), permanent (always present), and longstanding (experienced for over one year). Notably, AF is implicated as a contributing factor for nearly 20,000 strokes annually within the UK, illustrating its public health relevance.
Dr. Sonya Babu-Narayan, associate medical director at the BHF, underlined the urgency for increased detection: “We have effective treatments available for individuals with AF at high stroke risk. Currently, some patients are missing out simply because they are unaware they carry this hidden health threat.” With this new AI tool, data harvested from GP records can provide insights leading to timely interventions.
The research not only aims to catch AF early but also aspires to pave the way for larger-scale trials across the UK healthcare system. "The use of AI and big data is set to be game-changing for our healthcare services, making identification and testing of conditions like AF easier and more efficient," Dr. Ramesh Nadarajah remarked from Leeds Teaching Hospitals NHS Trust.
Health Secretary Wes Streeting has highlighted the transformative potential of AI within healthcare, emphasizing its role in advancing earlier diagnosis and intervention methods.
Currently, NHS figures indicate significant progress: five years ago, ambitions were established to raise the percentage of AF patients prescribed blood-thinning medications to 90%—a goal now exceeded, with 92% reporting they have been prescribed treatment. This proactive step is credited with saving countless lives through stroke prevention.
The emergence of the FIND-AF tool marks a hopeful advance toward reducing undiagnosed AF cases and, by extension, could significantly curb incidences of preventable strokes across the UK, underscoring the ripple effect of early detection and treatment.
