A recent study comparing the performance of artificial intelligence (AI) to certified electrocardiogram (ECG) technicians in analysing ambulatory ECG recordings found that AI outperformed technicians, significantly reducing missed diagnoses while slightly increasing false alarms.
The group of international researchers suggested that using AI to analyse ECGs could facilitate direct-to-physician reporting, cutting costs, improving access to care and enhancing patient outcomes.
The AI model called DeepRhythmAI correctly diagnosed more cases than technicians in ECG analysis, demonstrating a 98.6% sensitivity in detecting critical arrhythmias compared to the 80.3% sensitivity achieved by technicians. However, the technology also increased the rate of false positive cases where arrhythmias were incorrectly identified (12 vs five events per 1,000 patient days).
The risk of a false-negative rate was significantly lower using DeepRhythmAI (3.2 vs 44.3 per 1,000 patients). However, the study showed that technicians were 14.1 times more likely to miss a diagnosis (95% CI = 10.4–19.0) than the AI technology.
AI promise for ambulatory ECG analyses
The study population was a random sample of 14,606 patients (mean age = 65.5 ± 10 years, 42.8% males) being monitored in the United States for clinical indications of heart disease such as palpitations, syncope, dizziness, and examination for atrial fibrillation. The patients were observed between 2016 and 2019 for a mean of 14 ± 10 days using an ambulatory ECG monitor and were referred by 1,079 physicians from 166 clinics.
Beat-by-beat annotation of 14,606 individual ambulatory ECG recordings was analysed by one of 167 certified ECG technicians. To compare the performance of the AI model and the technicians, a random sample of 5,235 rhythm events – of which 2,236 events were identified as critical arrhythmias – was selected for annotation by one of 17 cardiologist consensus panels.
The researchers concluded that the AI model showed excellent negative predictive value for critical arrhythmias, substantially reducing false-negative findings, but this came at a modest cost of increased false-positive findings.
With the development of ECG technology, large amounts of ECG data have become available, which need to be interpreted by human technicians. The DeepRhythmAI model can save technicians time and improve the rate of missed diagnoses, while facilitating direct-to-physician reporting. This can potentially reduce costs and improve access to care and outcomes in patients who need ambulatory ECG monitoring, the researchers said.
Reference
Johnson, L.S. et al. Artificial intelligence for direct-to-physician reporting of ambulatory electrocardiography. Nature Medicine. 2025, Feb. 10: DOI: doi.org/10.1038/s41591-025-03516-x.
