AI-assisted ECG screening for atrial fibrillation in older adults: comparison of short-term and one-week monitoring
O Rantula, J A Lipponen, J Halonen, H Jantti, T T Rissanen, M P Tarvainen, N S Naukkarinen, E S Seljola, O E Santala, J Sedha, L Karavirta, L Palmberg, T Rantanen, T J Martikainen, J E K HartikainenAbstract
Background
Atrial fibrillation (AF) is the most common sustained arrhythmia and a major risk factor for stroke. In patients at high risk of stroke, even a single documented AF episode prompts evaluation for anticoagulation therapy. Conventional ECG methods are poorly suited for prolonged rhythm monitoring, whereas wearable devices enable long-term ECG recording in everyday life. However, extended monitoring generates large datasets that are time-consuming and resource-intensive if reviewed manually.
Purpose
We evaluated the clinical usefulness of artificial intelligence (AI) assisted AF detection from brief single-lead ECG recordings obtained during a short physical challenge (six-minute walk test, 6MWT) and during extended seven-day home ECG monitoring. Specifically, we assessed whether AF detected during 6MWT could substitute for AF detection during prolonged monitoring.
Methods
A total of 461 adults aged 75–85 years underwent both single-lead ECG recording during a 6MWT and seven days of continuous ECG monitoring using a medical-grade single-lead device (Faros 180, Bittium Corporation, Oulu, Finland). Participants were living independently and extended ECG monitoring was performed at home during everyday activities. An AI-based arrhythmia analysis algorithm (Awario, Heart2Save, Kuopio, Finland) was used for automatic AF screening of the ECG recordings. All flagged episodes were subsequently reviewed and confirmed by a physician. For extended monitoring, AF was defined as physician-verified AF with a burden of ≥6 minutes per day during monitoring. During the 6MWT recording, any physician-verified AF episode was considered a positive finding.
Results
A total of 461 participants completed both 6MWT ECG recording and extended home ECG monitoring. During 6MWT, new AF was detected in 5 individuals (1.08%). AI-based pre-screening identified potential AF in 87 participants during extended home monitoring. After physician verification, 16 cases were excluded due to artefacts or insufficient AF duration, resulting in 71 confirmed AF cases. Of the 62 participants with previously diagnosed AF, AF was detected during monitoring in 45 individuals. Thus, 26 cases represented newly diagnosed AF (5.64%). All AF cases detected during 6MWT were also identified during extended home monitoring. Overall, extended home ECG monitoring detected over five-fold more new AF cases compared with brief recordings (26 vs. 5).
Conclusion
AI-assisted extended single-lead ECG monitoring was more effective than brief ECG recordings obtained during a 6MWT in detecting new AF in older adults. While short recordings may identify a minority of AF cases, prolonged monitoring improves clinically relevant AF detection and characterization. AI-supported pre-screening may facilitate scalable AF detection by reducing clinician workload, with ongoing algorithm development aimed at improving detection accuracy and minimizing artefact-related findings.