August 02, 2019
2 min read

AI may identify AF in patients during normal sinus rhythm

You've successfully added to your alerts. You will receive an email when new content is published.

Click Here to Manage Email Alerts

We were unable to process your request. Please try again later. If you continue to have this issue please contact

Paul A. Friedman

An AI-enabled ECG that was obtained during normal sinus rhythm helped identify patients with atrial fibrillation at the point of care, according to a study published in The Lancet.

“Applying an AI model to the ECG permits detection of atrial fibrillation even if not present at the time the ECG is recorded,” Paul A. Friedman, MD, cardiac electrophysiologist, cardiologist and chair of cardiovascular medicine at the Mayo Clinic, said in a press release. “It is like looking at the ocean now and being able to tell that there were big waves yesterday.”

ECGs during normal sinus rhythm

Zachi I. Attia, MS, principal machine learning engineer and co-director of artificial intelligence in cardiology at the Mayo Clinic, and colleagues analyzed data from 180,922 patients with 649,931 normal sinus rhythm ECGs that were taken between Dec. 31, 1993 and July 21, 2017. AF was defined as any ECG with a rhythm of atrial flutter or AF.

Patients were assigned to the training dataset (454,789 ECGs from 126,526 patients), the internal validation dataset (64,340 ECGs from 18,116 patients) or the testing dataset (130,802 ECGs from 36,280 patients). A convolutional neural network was implemented to detect any indication of AF in ECGs. The primary outcome was the identification of patients with AF by the AI-enhanced ECG.

Verified AF before the normal sinus rhythm ECG was observed in 8.4% of patients in the testing dataset. A single AI-enabled ECG was able to identify AF with a sensitivity of 79% (95% CI, 77.5-80.4), an area under the curve of 0.87 (95% CI, 0.86-0.88), specificity of 79.5% (95% CI, 79-79.9), overall accuracy of 79.4% (95% CI, 79-79.9) and an F1 score of 39.2% (95% CI, 38.1-40.3).

When all of the acquired ECGs were analyzed, AI increased several factors including sensitivity (82.3%; 95% CI, 80.9-83.6), area under the curve (0.9; 95% CI, 0.9-0.91), specificity (83.4%; 95% CI, 83-83.8), overall accuracy (83.3%; 95% CI, 83-83.7) and F1 score (45.4%; 95% CI, 44.2-46.5).

“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 undermined source],” Attia and colleagues wrote. “This study shows the power of leveraging modern computing technology, large datasets, non-linear models and automated features extraction using convolution layers to potentially improve diagnosis and treatment of a highly prevalent and morbid disease state.”

Potential impact

“Given that AI algorithms have recently reached cardiologist level in diagnostic performance, this AI-ECG interpretation is groundbreaking in creating an algorithm to reveal the likelihood of atrial fibrillation in ECGs showing sinus rhythm,” Jeroen M.L. Hendriks, RN, PhD, FESC, FSCANZ, Derek Frewin Lectureship at the Centre for Heart Rhythm Disorders at the University of Adelaide and Royal Adelaide Hospital in Australia, and Larissa Fabritz, MD, reader at the Institute of Cardiovascular Science and honorary consultant in the department of cardiology at the University of Birmingham in the United Kingdom, wrote in a related editorial. – by Darlene Dobkowski

Disclosures: The authors report no relevant financial disclosures. Hendriks reports his institution received on his behalf consultant or lecture fees from Medtronic and Pfizer/Bristol-Myers Squibb. Fabritz reports she received institutional research grants and nonfinancial support from Gilead and is listed as an inventor for two patents held by the University of Birmingham.