Arrhythmias & electrophysiology
Artificial intelligence predicts undiagnosed atrial fibrillation in patients with embolic stroke of undetermined source using sinus rhythm electrocardiograms
Authors' conclusion
Our AI model demonstrated excellent diagnostic performance in predicting AF in patients with ESUS
Study context
- Population
- Patients with embolic stroke of undetermined source (ESUS) monitored with an insertable cardiac monitor
- Participants
- 352 patients · 4 tertiary hospitals
- Comparison
- Artificial intelligence (AI) model on sinus rhythm (SR) electrocardiogram (ECGs) for detection/prediction of atrial fibrillation (AF), against insertable cardiac monitor (ICM)-detected AF
- Outcome
- AF episodes lasting ≥1 hour detected by insertable cardiac monitor
- Follow-up
- 25.1-month follow-up
Methodological detailsmulticentre, 4 tertiary hospitals, Model developed using 737,815 SR ECGs from patients with and without AF
- Multicentre
- 4 tertiary hospitals
- Model developed using 737,815 SR ECGs from patients with and without AF
Performance
As reported in the abstract
Identification of AF ≥1 hour
| Strategy | AUC |
|---|---|
| AI algorithm | 0.806 |
| AI + clinical parameters | 0.880 |
AUC
AI algorithm0.806
AI + clinical parameters0.880
Further findings
- Identification of longer AF episodes (AI algorithm)
- 0.837 · AUC, AF ≥12 hours · AUC, AF ≥24 hours 0.879
Incidence of AF ≥1 hour on ICM · 25.1-month follow-up
- Patients with AF ≥1 hour
- 58 (14.4%)
AI-based AF risk score as ECG recording approached AF onset
- Risk score increased as ECG recording approached AF onset
- P for trend < 0.0001
Original publication
Artificial intelligence predicts undiagnosed atrial fibrillation in patients with embolic stroke of undetermined source using sinus rhythm electrocardiograms. Heart Rhythm. Published 15 March 2024. doi:10.
Summarised from the published abstract; every number is checked against it