Letter from the Editor

Use of Electrocardiography to Detect AF: Hard to Beat

Bradley P. Knight, MD, FACC, FHRS, Editor-in-Chief

Bradley P. Knight, MD, FACC, FHRS, Editor-in-Chief

This weekend, a colleague sent an email related to a mutual patient who has a history of atrial fibrillation (AF). Attached to the email was a patient-acquired rhythm strip in the form of a PDF clearly showing AF (Figure 1).

The rhythm strip was recorded by the patient using the smartphone Kardia application (AliveCor, Inc). The email and the attached rhythm strip were easily forwarded by email to our clinic nurse with a request to contact the patient on Monday and arrange for an electrical cardioversion.

The market has exploded with new technologies to make electrocardiographic recordings. Although many have argued that the ability of patients to make their own EKG recordings and to contact their doctors directly 24/7 will lead to a tsunami of uncompensated work for doctors and nurses, no one can argue that these new smartphone- and watch-based applications do not provide an efficient and cost-effective way for patients be engaged in their heart rhythm care.

In 2018, Yan et al published a paper showing that facial photoplethysmographic (FPPG) signals can be used to detect variations in facial blood flow to generate pulse waveforms that can be used to detect heart rate irregularity.1 A neural network approach has been applied to these data to diagnose atrial fibrillation. In their latest research, Yan et al provided additional information2 related to the potential of this new technology when applied to groups of patients from a distance. Filming multiple permutations of five patients (some in AF) using a digital camera 1.5 meters away for one minute while the patients were still, they were able to detect which patients were in AF with over a 95% reliability. This concept is novel, and has the potential to diagnose AF in a person without ever touching him or her. It could be used for population screening of AF, or simply to determine heart rates when triaging casualties.

Initial reactions to this recent publication have focused on the privacy risks.3 Could this technology be applied to patients without their knowledge? Who would control the data? However, the first question should be whether or not this type of technology would ever have a role in the diagnosis of AF.

Video plethysmography has an advantage over electrocardiography because it does not require physical contact and could be done remotely. However, very few things are as timeless in medicine, and have proven to be as useful and cost-effective, as electrocardiography. Electrocardiography is not only the gold standard for determining the rhythm of the heart, it can be used to inspect for other cardiac conditions such as prior myocardial infarction, chamber hypertrophy, conduction disorders, and QT prolongation. In fact, a group recently showed that hyperkalemia can be detected in stable patients by applying deep learning to a routine 12-lead EKG.4 For the average patient with a heart rhythm disorder, a simple rhythm strip provides far more insight than inspection of venous pulsations and cardiac auscultation. FPPG is clever, but it is hard to imagine that it will ever be the preferred tool to apply to a person to determine their heart rhythm given that there are already multiple products available to quickly and inexpensively make an electrocardiographic heart rhythm recording.


Bradley P. Knight, MD, FACC, FHRS

Editor-in-Chief, EP Lab Digest

Disclosure: Dr. Knight reports that he is a consultant, speaker, investigator, and offers fellowship support for Abbott, Baylis Medical, Biosense Webster, Inc., BIOTRONIK, Boston Scientific, Medtronic, and SentreHEART.

  1. Yan BP, Lai WHS, Chan CKY, et al. Contact-free screening of atrialfibrillation by a smartphone using facial pulsatile photoplethysmographic signals. J Am Heart Assoc. 2018;7(8):e008585. doi:10.1161/JAHA.118.008585
  2. Yan  BP, Lai  WHS, Chan  CKY,  et al.  High-throughput, contact-free detection of atrial fibrillation from video with deep learning. JAMA Cardiol. 2019 Nov 27.  doi:10.1001/jamacardio.2019.4004. [Epub ahead of print]
  3. Turakhia MP. Diagnosing with a camera from a distance — proceed cautiously and responsibly. JAMA Cardiol. 2019 Nov 27. doi:10.1001/jamacardio.2019.4572. [Epub ahead of print]
  4. Galloway CD, Valys AV, Shreibati JB, et al. Development and validation of a deep-learning model to screen for hyperkalemia from the electrocardiogram. JAMA Cardiol. 2019;4(5):428-436. doi:10.1001/jamacardio.2019.0640.