A recent study published in Nature Medicine found that an artificial intelligence model outperformed human technicians in analyzing long-term electrocardiogram (ECG) recordings. Researchers tested the AI, called DeepRhythmAI, against 17 panels of expert physicians in a large international study.
Long-term ECGs record every heartbeat, but reviewing them for abnormalities is time-consuming and resource-intensive. The current study included over 200,000 days of ECG data from 14,606 individual patients who had recorded an average of 14 days each. The AI algorithm was able to identify severe arrhythmias with high accuracy, missing only 0.3% of cases compared to 4.4% for human analysis.
The study aimed to determine the potential benefits of using AI to replace human technicians in analyzing long-term ECG recordings. With a shortage of trained staff worldwide, this approach could address the bottleneck in healthcare and provide faster diagnostics at a lower cost.
Researchers designed the study with specific characteristics in mind, including near-perfect sensitivity to identify potentially serious arrhythmias and minimizing false positives. The AI model exceeded expectations, ruling out severe arrhythmia with 99.9% confidence in a 14-day ECG recording and generating only 12 false positives per 1,000 recording days.
The study’s findings have significant implications for the future of healthcare, particularly in addressing the worldwide shortage of trained staff capable of interpreting long-term ECG monitoring. With AI-powered diagnostics on the horizon, patients may benefit from faster and more accurate diagnoses, while also reducing the workload of healthcare professionals.
Source: https://medicalxpress.com/news/2025-02-ai-humans-term-ecg-large.html