"Artificial Intelligence Tackles the ECG: Detecting Atrial Fibrillation as It Strikes"

Recently, I asked my coworker if their smartwatch tracked their heart rate.  They showed me the screen, which had a little beating heart icon along with some numbers.  “Umm, is your heart rate normally that high?” I asked. They just laughed.

This got me thinking. Smartwatches and rings are great tools for fitness enthusiasts.  But wearable devices are also being used by consumers and healthcare providers to detect and diagnose heart issues [1].  Current technology has made it easier than ever to capture important data about the heart, but couldn’t this be creating a different kind of challenge for users?  The heart beats 24/7, meaning heart monitoring devices record enormous amounts of raw data.  For healthcare providers, this represents a huge workload of sorting, analyzing, and reaching a diagnosis for a heart problem, even with devices to assist them. 

In this edition of MedTech Momentum, I’ll dig deeper into this topic to provide an example of how AI technology is speeding up diagnosis of a common and significant heart rhythm disorder, atrial fibrillation.

But first, let’s talk about your heartbeat because it’s electrifying!

If you press on your wrist or the side of your neck, you’ll feel the steady thump-thump of your heartbeat.  But there’s much more to your pulse than the pressure on your fingertips.

The human heart is a collection of muscles that form two small chambers (atria) that sit on top of two much larger chambers (ventricles) [2,3].  During a single heartbeat, a pacemaker in the atria triggers their contraction to push blood into the ventricles.  The pacemaker’s signal next travels down to the ventricles, causing them to contract and squeeze blood into the rest of the body.  This pattern of atria-to-ventricle contractions defines a normal heart rhythm—also called a sinus rhythm [4,5].

If the heart is just a series of contractions, how can you detect and record how well it’s beating?  It turns out, the atria and ventricles generate a strong electrical current when they contract [6].  This electrical activity can be visualized and recorded on an electrocardiogram (ECG, or EKG) using a device that measures electrical activity through leads glued to a person’s chest [6,7].  ECG recordings give medical experts a detailed picture of a heart’s rhythm over time so they can more easily diagnose heart irregularity-related conditions.

Atrial Fibrillation: when the heart gets uncoordinated, life gets difficult

In a state of atrial fibrillation (AF), the muscles of the atria spasm or quiver rapidly (fibrillation) up to 400 times per minute!  This causes them to deliver less blood to the ventricles than normal.  In turn, the ventricles beat up to twice as fast as they usually do (120 vs 60 times per minute) [2,4,5].  As such, the atria deliver less blood to the ventricles again and the two sets of chambers begin to contract out of synch with each other.  The result?  The heart might not be able to pump enough blood to support a person’s physical activity in the moment.

Symptoms of AF can vary widely and be frightening.  Some people experience a racing or pounding heartbeat or feel like they are having a heart attack (i.e., heart palpitations) [6,7].  Others feel short of breath, tired, or dizzy, especially while exercising [8,9].  As such, AF symptoms are often mistaken for panic attacks or feelings of anxiety [10, 11].  More ominous is ‘silent’ AF, where a person experiences no symptoms at all and the condition isn’t noticed until that individual receives care for a different health problem entirely [5,7].

What about duration?  AF doesn’t have a fixed duration—it can last seconds, minutes, hours, or even days.  It's normal for people with AF to experience short, rare episodes lasting a few seconds or minutes [2,5].  However, more frequent or longer episodes can be considered more serious.  In persistent or chronic cases of AF, the heart can’t restore its own regular rhythm, requiring medical intervention [2,9].

Atrial fibrillation could happen to you or someone you know

It’s estimated that 1 in 5 people globally will experience AF that requires treatment in their lifetime (that’s 22 million people in the USA alone!) [12,13].  The risk for developing AF increases with age, especially over the age of 60 years.  Risk is further elevated by factors including high blood pressure, smoking, alcohol consumption, diabetes, and prior damage to the heart, such as from a heart attack or surgery [4,5,6,]. 

The most significant danger with AF is an increased risk of having a stroke.  When your atria quiver instead of contracting, blood can pool inside and form clots [2,5].  Parts of these blood clots could break off and get pushed into your circulation after which they could get lodged in your brain. This (having a stroke) is probably the worst time to discover you have AF, so it’s better to get ahead of the curve if you can!

Atrial fibrillation can be elusive to detect, but technology rises to the challenge

Because of the stroke risk, it’s essential to get a definitive diagnosis of AF as quickly as possible.  The challenge lies in catching enough episodes of AF with an ECG so an expert can diagnose the type of AF at play.  For many people, infrequent episodes may not be caught on an ECG during a short visit with a healthcare provider [5]. 

In the 1960’s, people began wearing mobile ECG monitors at home, such as the Holter [14,15].  These mobile devices quickly became a standard method for longer ECG recordings [2,13].  While reliable, the limited recording duration of the Holter (up to 48 hours) often required repeated testing and a delay in the diagnosis and treatment of AF [14,16].

In recent decades, medical device companies have invented improved ECG recording devices that are comfortable to wear and can collect data over several days or even weeks at a time [1, 13].  The most recent innovation is possibly the most exciting: using artificial intelligence to shorten the period between analyzing enormous amounts of ECG data to reach an accurate diagnosis and get a patient on an ideal treatment plan [16,17]. 

Artificial intelligence tools are making it easier to diagnose AF:  A spotlight on iRhythm’s Zeus AI technology

Using software to analyze ECG recordings is nothing new.  Early attempts at automating heart arrhythmia analysis go back to the 1970’s [14,15].  Today, AI technology is being developed to detect and classify AF to an unprecedented degree [17,18].  Trained on large libraries of ECG data, these systems can rapidly detect and classify AF from a patient’s data at the level of medical experts [17, 18, 19, 20].  When combined with wearable ECG monitors, AI is a game-changer for AF diagnosis.

An example of this is the Zeus AI by iRhythm.  iRhythm’s Zeus is an FDA-cleared deep-learning algorithm designed to analyze ECG data from iRhythm’s wearable monitor, the Zio: a chest patch worn for up to two weeks at a time [21].  The Zeus algorithm is based on a machine learning system trained on a large set of ECG data representing scans of both normal and AF heart rhythms.  As such, the Zio has been shown to be highly effective at detecting AF, being able to identify episodes of AF with an accuracy comparable to expert cardiologists [16, 22, 23].

The Zio is not the only AI-powered ECG medical device on the market [13]. However, the Zeus system illustrates how AI-driven analysis of complex clinical data can help patients and doctors by shortening the time to diagnose AF and determine an appropriate treatment plan [23].

Will AI-assisted diagnosis improve long-term health outcomes for people with AF?

In the short term, using AI will improve the speed with which people can receive personalized treatment for AF.  The deeper question is: will the use of AI to diagnose AF improve long-term clinical outcomes, such as reducing the incidence of stroke or additional health problems attributed to AF?

It’s too soon to be certain.

A recent study showed that while iRhythm’s Zio XT patch can speed up AF diagnosis, it doesn’t correlate with a reduction in hospitalization for stroke.  However, this study was interrupted by the COVID-19 pandemic, meaning no conclusive answer could be drawn [24]. 

ECG and other heart rhythm-tracking technology are evolving, and multiple medical technology companies and academic research programs are currently evaluating the efficacy of long-term monitoring paired with AI-detection.  Nevertheless, this is undoubtedly an area where AI-assisted medical technologies have the potential to improve healthcare for our population.  It might not be long until a medical drama on television shows a doctor saying, “What does {insert AI name} thinkI about the patient’s heart rhythm?”

Citations

1.  DAIC B. BLOG: A Double-Edged Sword: How Over-the-Counter ECG Devices are Impacting Cardiac Care. DAIC. August 24, 2022. Accessed October 8, 2025. http://www.dicardiology.com/content/blogs/blog-double-edged-sword-how-over-counter-ecg-devices-are-impacting-cardiac-care

 2. Waktare JEP. Atrial Fibrillation. Circulation. 2002;106(1):14-16. doi:10.1161/01.CIR.0000022730.66617.D9

 3. Association AH. Cardiovascular Media Library. Watch. Learn. Live. Accessed October 3, 2025. https://watchlearnlive.heart.org/index.php?moduleSelect=atrfib

 4. Atrial fibrillation and atrial flutter. Accessed October 16, 2025. https://ssl.adam.com/content.aspx?productid=117&isarticlelink=false&pid=1&gid=000184

 5. CDC. About Atrial Fibrillation. Heart Disease. May 20, 2024. Accessed October 9, 2025. https://www.cdc.gov/heart-disease/about/atrial-fibrillation.html

 6. Friedman PA. The Electrocardiogram at 100 Years: History and Future. Circulation. 2024;149(6):411-413. doi:10.1161/CIRCULATIONAHA.123.065489

 7. Vincent R. From a laboratory to the wearables: a review on history and evolution of electrocardiogram. From a laboratory to the wearables: a review on history and evolution of electrocardiogram. 2022;4(4):248-255. doi:10.53986/ibjm.2022.0038

 6. Clinic M. Atrial fibrillation - Symptoms and causes. Mayo Clinic. Accessed June 4, 2025. https://www.mayoclinic.org/diseases-conditions/atrial-fibrillation/symptoms-causes/syc-20350624

 7. Health A. Patient Success Story - AdmIRE and AFib - Atlantic Health System. Accessed October 9, 2025. https://ahs.atlantichealth.org/conditions-treatments/heart-care/treatment-services/cardiac-electrophysiology/joe-p-story.html

 8. StopAfib. StopAfib.org - Atrial Fibrillation Patient Story: Michele Straube. Accessed October 9, 2025. https://www.stopafib.org/patient-stories/michele-straube/

 9. Hopkins J. Atrial Fibrillation (AFib): Sarann’s Story. Accessed October 9, 2025. https://www.hopkinsmedicine.org/health/conditions-an

10. Presbyterian R. Laura’s Story | Rush. Accessed October 9, 2025. https://www.rush.edu/news/lauras-story?utm_source=chatgpt.com

 11. StopAfib. StopAfib.org - Atrial Fibrillation Patient: Jason Mattingly. Accessed October 9, 2025. https://www.stopafib.org/patient-stories/jason-mattingly/

 12. Benjamin EJ, Muntner P, Alonso A, et al. Heart disease and stroke statistics—2019 update: a report from the American Heart Association. Circulation. 2019;139(10):e56–e528

 13. Atrial fibrillation estimated to affect about 1 in 22 Americans | NHLBI, NIH. September 12, 2024. Accessed June 25, 2025. https://www.nhlbi.nih.gov/news/2024/atrial-fibrillation-estimated-affect-about-1-22-americans

 14. Noubiap JJ, Tang JJ, Teraoka JT, Dewland TA, Marcus GM. Minimum National Prevalence of Diagnosed Atrial Fibrillation Inferred From California Acute Care Facilities. JACC. 2024;84(16):1501-1508. doi:10.1016/j.jacc.2024.07.014

 13. DAIC B. BLOG: Moving Beyond Holter ECG: The Rise in Next Generation Technology. DAIC. August 24, 2022. Accessed October 8, 2025.

 14. What Is a Holter Monitor? Cleveland Clinic. Accessed June 18, 2025. https://my.clevelandclinic.org/health/diagnostics/21491-holter-monitor

 15. Vincent R. From a laboratory to the wearables: a review on history and evolution of electrocardiogram. From a laboratory to the wearables: a review on history and evolution of electrocardiogram. 2022;4(4):248-255. doi:10.53986/ibjm.2022.0038\\

 16. iRhythm Unveils New Data Demonstrating Benefits of Long-Term Continuous Monitoring for Arrhythmia Detection. DAIC. March 31, 2025. Accessed September 5, 2025. http://www.dicardiology.com/content/irhythm-unveils-new-data-demonstrating-benefits-long-term-continuous-monitoring-arrhythmia

 17. Tarakji K. Digital Health and the Care of the Patient With Arrhythmia | Circulation: Arrhythmia and Electrophysiology. Accessed October 9, 2025.

 18. Rahul J, Sharma LD. Advancements in AI for cardiac arrhythmia detection: A comprehensive overview. Computer Science Review. 2025;56:100719.

 19. Karakasis P, Theofilis P, Sagris M, et al. Artificial Intelligence in Atrial Fibrillation: From Early Detection to Precision Therapy. Journal of Clinical Medicine.

 20.  Hannun AY, Rajpurkar P, Haghpanahi M, et al. Cardiologist-level arrhythmia detection and classification in ambulatory electrocardiograms using a deep neural network. Nat Med. 2019;25(1):65-69. doi:10.1038/s41591-018-0268-3

 21. iRhythm Gains FDA Clearance for its Clinically Integrated ZEUS System |. Accessed October 9, 2025. https://www.irhythmtech.com/us/en/who-we-are/news-events/irhythm-gains-fda-clearance-for-its-clinically-integrated-zeus-system

 22. Rosenberg MA, Samuel M, Thosani A, Zimetbaum PJ. Use of a Noninvasive Continuous Monitoring Device in the Management of Atrial Fibrillation: A Pilot Study. Pacing Clin Electrophysiol. 2013;36(3):328-333. doi:10.1111/pace.12053

 23. Taylor P. iRhythm wearable heart monitor raises AFib diagnoses by 52%. Pharmaphorum. September 2, 2024. Accessed September 5, 2025. https://pharmaphorum.com/news/irhythm-wearable-heart-monitor-raises-afib-diagnoses-52

 24. Lopes RD, Atlas SJ, Go AS, et al. Effect of Screening for Undiagnosed Atrial Fibrillation on Stroke Prevention. JACC. 2024;84(21):2073-2084. doi:10.1016/j.jacc.2024.08.019