A single AI-enabled ECG identified AF with a sensitivity of 79%, a specificity of 79.5%, and an overall accuracy of 79.4%, according to a new study.
Artificial intelligence (AI)-enabled electrocardiography (ECG) software is capable of identifying patients with atrial fibrillation (AF) even while they are in normal sinus rhythm, according to a study published in the September issue of The Lancet.
“This is the first study to our knowledge to use a convolution neural network to identify the electrocardiographic signature of atrial fibrillation present during sinus rhythm,” stated researchers led by Paul Friedman, MD, professor of medicine and chair of the Department of Cardiovascular Medicine, Mayo Clinic, Rochester, Minnesota.
Prior to this study, no one individual feature present on the ECG of patients with intermittent AF has been identified during periods of normal sinus rhythm that could reliably predict or diagnose AF.
AF screening is complex
AF is a very common and often underdiagnosed cardiac arrhythmia that puts patients at higher risk for stroke, heart failure, and death.
Screening can be a challenge, however, because performing a single ECG may not catch the arrhythmia since it can come and go. Finding a low-cost, readily available, noninvasive test with a high sensitivity for diagnosing AF would have exciting implications for both diagnosis and therapy aimed at preventing morbidity and mortality.
The authors trained, tested, and validated a neural network that would identify very subtle features present in a standard 10-second, 12-lead ECG during normal sinus rhythm-features that may accurately point to a history of or risk for developing AF.
AI-enabled ECG proves accurate
The researchers conducted a retrospective review of >180 000 Mayo Clinic patients aged ≥18 years who had at least 1 digital, normal sinus rhythm, standard 10-second, 12-lead ECG between 1993 and 2017. Among the 180 922 patients, nearly 650 000 ECGs were evaluated by trained technicians.
Researchers applied the AI algorithm to the analysis of the ECGs, which were also read by trained personnel for validation.
A single AI-enabled ECG identified AF with a sensitivity of 79%, a specificity of 79.5%, and an overall accuracy of 79.4%.
When all of the ECGs acquired during the first month of each patient’s window of interest were accounted for--rather than just a single one--the sensitivity, specificity, and overall accuracy climbed to 82.3%, 83.4%, and 83.3%, respectively.
The authors stated that the AI-enabled ECG as a diagnostic tool for uncovering AF performs on par with other established tests such as the brain-type natriuretic peptide test for heart failure and the Papanicolaou smear for cervical cancer.
Friedman and colleagues point out that, “The ability to identify undetected atrial fibrillation with an inexpensive, widely available, point-of-care test-an ECG recorded during normal sinus rhythm-has important practical implications, particularly for atrial fibrillation screening efforts or for the management of patients with embolic stroke of unknown source.”
Finally, this technology offers a readily available, low cost, noninvasive, point-of-care test that could have a large impact on the quality of patients’ lives.
Source: Attia ZI, Noseworthy PA, Lopez-Jimenez F, et al. An artificial intelligence-enabled ECG algorithm for the identification of patients with atrial fibrillation during sinus rhythm: a retrospective analysis of outcome prediction. Lancet. 2019;394:861-867.