The Electrocardiogram at 100 Years: History and Future

HomeCirculationVol. 149, No. 6The Electrocardiogram at 100 Years: History and Future Free AccessArticle CommentaryPDF/EPUBAboutView PDFView EPUBSections ToolsAdd to favoritesDownload citationsTrack citationsPermissions ShareShare onFacebookTwitterLinked InMendeleyReddit Jump toFree AccessArticle CommentaryPDF/EPUBThe Electrocardiogram at 100 Years: History and Future Paul A. Friedman Paul A. FriedmanPaul A. Friedman Correspondence to: Paul A. Friedman, MD, Department of Cardiovascular Medicine, Mayo Clinic, 200 First St SW, Rochester, MN 55905. Email E-mail Address: [email protected] https://orcid.org/0000-0001-5052-2948 Department of Cardiovascular Medicine, Mayo Clinic, Rochester, MN. Originally published5 Feb 2024https://doi.org/10.1161/CIRCULATIONAHA.123.065489Circulation. 2024;149:411–413The ECG is the cumulative recording on the body surface of cardiac electrical activity, generated by the action potentials of millions of cardiomyocytes. In contrast to neurons, cardiomyocytes are connected as a syncytium, in which activation of 1 cell activates adjoining cells, leading to wavefront propagation and coordinated electrical current generation. At the cellular level, individual myocytes generate action potentials through the carefully regulated activity of membrane channels, often ion specific, which create currents that propagate from cell to cell using pores formed by specific proteins. The multiple, complex, homeostatic mechanisms involved in the generation of the ECG enable it to serve as a unique identifying fingerprint for each individual, to function as an early detector of disease, and potentially to upend medical screening and remote monitoring. The power of this test could scarcely be imagined when it was first created at the turn of the 20th century.It had been recognized since the late 1800s that the beating heart produces electrical currents. However, available tools such as the capillary electrometer used by Waller lacked the sensitivity to adequately resolve millivolt level electrical signals. Einthoven, a Dutch physician and physiologist, thought to use a string galvanometer, in which the presence of a current deflects a filament positioned between magnets, to invent the first practical ECG in 1903, for which he was later awarded the Nobel Prize. He recognized signal features and assigned them the names still used today: P wave, QRS complex, and T wave. The ECG provided a way to objectively measure and diagnose heart disease, beyond signs and symptoms, and ultimately led to the birth of a new field: cardiology.Over the following decades, a growing number of observations linked human-named electrocardiographic signal features to cardiac and noncardiac diseases, including ischemia, arrhythmias, electrolyte abnormalities, liver disease, and others, with variations in the sensitivity and specificity for each disease in part as a function of its impact on cardiac electrical activity. With the development of increasing nuance and sophistication in these observations, the field of electrocardiography and, with it, cardiology grew. The ECG became central to the diagnosis of critical conditions such as acute myocardial infarction, atrial fibrillation (AF), and ventricular tachycardia. As cardiology matured and new technologies such as echocardiography were introduced, a deeper understanding of the relationships between electrocardiographic signal features and structural abnormalities (eg, hypertrophy, valvular lesions) developed, as did the strengths and limitations of these electrocardiographic signal features for diagnosis. A growing number of electrocardiographic criteria and rules defined by empirical associations between the ECG and disease conditions were described for the practitioner to remember to identify disease. In the 1970s, these feature-based rules were encoded into electrocardiographic machines to generate diagnostic reports as a clinical aid. Although debate exists about the accuracy and reliability of these feature-based electrocardiographic reading systems, they represented the pinnacle of computerized reads, until recent machine learning-based approaches were developed, now poised to reshape the clinical prowess of the ECG and clinical workflows.Machine learning has been used to train deep neural networks, a series of mathematical equations constructed to mimic human cerebral cortex, with each neuron represented by a nonlinear equation that, like a neuron, may "discharge" according to its input. Large labeled data sets have trained networks that identify subtle, nonlinear electrocardiographic patterns, potentially in unnamed electrocardiographic segments not used in traditional feature analysis, to enable 3 broad capabilities: (1) performing tasks that humans perform such as rhythm classification but at massive scale, (2) exceeding human capabilities by identifying conditions (eg, ventricular dysfunction) not reliably recognized by expert human readers, and (3) identifying individuals at high risk of developing disease not yet detected with standard tests.1 The mechanism by which the artificial intelligence (AI) ECG performs these predictive tasks is not confirmed but likely reflects disease impact on ion channels, resulting in subtle electrocardiographic changes that precede abnormalities detectable by imaging studies such as echocardiography. It has long been known that ischemia on the ECG precedes echocardiographic wall motion abnormalities.AI ECGs have been developed to effectively detect left ventricular dysfunction, the presence of AF from an electrocardiographic recording in during sinus rhythm, aortic stenosis, pulmonary hypertension, cirrhosis, hypertrophic cardiomyopathy, and hyperkalemia, among other conditions.1 Because it is software based, the AI ECG has been rapidly and inexpensively tested in clinical workflows. In the EAGLE study (ECG AI-Guided Screening for Low Ejection Fraction), 22 640 adults were enrolled in ≈8 months during the pandemic using a cluster randomized design. In the intervention arm, when routine ECGs were ordered by primary care practitioners, AI analysis results were presented to clinicians; in the control arm, the results were not disclosed. The availability of the AI electrocardiographic result increased the new diagnosis of left ventricular dysfunction by one-third.2 Top of license practice was seen as nurse practitioners and physician assistants were twice as likely to follow the AI recommendation and thus make the diagnosis, recognizing that their panels were less complex.To test whether an ECG from a wearable form factor collected in nonmedical environments would suffice for AI analysis, Attia et al,3 working with the Mayo Clinic Center for Digital Health, created an application (app) to allow collection of Apple Watch electrocardiographic recordings and retrained the 12-lead network to classify watch ECGs. A single part-time study coordinator digitally enrolled 2544 subjects from 46 states and 11 countries and transmitted >125 000 ECGs in ≈6 months. The watch ECGs identified left ventricular dysfunction with an area under the curve of 0.89 (Figure).Download figureDownload PowerPointFigure. The evolution of the ECG. AI indicates artificial intelligence; App, application; and AUC, area under the curve. A is in the public domain (PD-US).Previously acquired and stored digital ECGs have also been used to identify disease in individuals not known to have it. In the BEAGLE study (Batch Enrollment for AI-Guided Intervention to Lower Neurologic Events in Unrecognized AF), 669 032 patients with no known AF and elevated CHADSVASC risk score for stroke with ECGs recorded in the past (January 1, 2017–June 30, 2021) were computer screened with an AI electrocardiographic algorithm to detect AF from an electrocardiographic recording during sinus rhythm and invited to enroll in the study that included use of a wearable monitor for a month. Those patients with a positive AI ECG for silent AF had a 5-fold increased likelihood of AF detection at 1 month.4 The AI ECG has also been shown to reflect therapeutic drug effect. Tison et al5 demonstrated that the AI electrocardiographic scores effectively detected hypertrophic cardiomyopathy, were associated with outflow gradient and NT-proBNP (N-terminal pro-B-type natriuretic peptide) levels, and fell in response to mevacamten therapy.In short, the ECG has evolved from a primitive detection of cardiac electrical activity to a sophisticated test to detect occult disease, to predict impeding disease, and to monitor therapeutic effect. It has become inexpensive, ubiquitous, integrated into clinical workflows, and available in nonmedical environments. Its role as a power tool in the pockets of clinicians is destined to continue to grow, even after 100 years.ARTICLE INFORMATIONSources of FundingNone.Disclosures A number of AI electrocardiographic algorithms have been licensed by Mayo Clinic to Anumana, Eko Health, and AliveCor, and Mayo Clinic and Dr Friedman may benefit financially from their commercialization.FootnotesThe American Heart Association celebrates its 100th anniversary in 2024. This article is part of a series across the entire AHA Journal portfolio written by international thought leaders on the past, present, and future of cardiovascular and cerebrovascular research and care. To explore the full Centennial Collection, visit https://www.ahajournals.org/centennialThe opinions expressed in this article are not necessarily those of the editors or of the American Heart Association.For Sources of Funding and Disclosures, see page 413.Circulation is available at www.ahajournals.org/journal/circCorrespondence to: Paul A. Friedman, MD, Department of Cardiovascular Medicine, Mayo Clinic, 200 First St SW, Rochester, MN 55905. Email friedman.paul@mayo.eduREFERENCES1. Somani S, Russak AJ, Richter F, Zhao S, Vaid A, Chaudhry F, De Freitas JK, Naik N, Miotto R, Nadkarni GN, et al. Deep learning and the electrocardiogram: review of the current state-of-the-art.Europace. 2021; 23:1179–1191. doi: 10.1093/europace/euaa377CrossrefMedlineGoogle Scholar2. Yao X, Rushlow DR, Inselman JW, McCoy RG, Thacher TD, Behnken EM, Bernard ME, Rosas SL, Akfaly A, Misra A, et al. Artificial intelligence-enabled electrocardiograms for identification of patients with low ejection fraction: a pragmatic, randomized clinical trial.Nat Med. 2021; 27:815–819. doi: 10.1038/s41591-021-01335-4CrossrefMedlineGoogle Scholar3. Attia ZI, Harmon DM, Dugan J, Manka L, Lopez-Jimenez F, Lerman A, Siontis KC, Noseworthy PA, Yao X, Klavetter EW, et al. Prospective evaluation of smartwatch-enabled detection of left ventricular dysfunction.Nat Med. 2022; 28:2497–2503. doi: 10.1038/s41591-022-02053-1CrossrefMedlineGoogle Scholar4. Noseworthy PA, Attia ZI, Behnken EM, Giblon RE, Bews KA, Liu S, Gosse TA, Linn ZD, Deng Y, Yin J, et al. Artificial intelligence-guided screening for atrial fibrillation using electrocardiogram during sinus rhythm: a prospective non-randomised interventional trial.Lancet. 2022; 400:1206–1212. doi: 10.1016/S0140-6736(22)01637-3CrossrefMedlineGoogle Scholar5. Tison GH, Siontis Konstantinos C, Abreau S, Attia Z, Agarwal P, Balasubramanyam A, Li Y, Sehnert AJ, Edelberg JM, Friedman PA, et al. Assessment of disease status and treatment response with artificial intelligence−enhanced electrocardiography in obstructive hypertrophic cardiomyopathy.J Am Coll Cardiol. 2022; 79:1032–1034. doi: 10.1016/j.jacc.2022.01.005CrossrefMedlineGoogle Scholar eLetters(0) eLetters should relate to an article recently published in the journal and are not a forum for providing unpublished data. Comments are reviewed for appropriate use of tone and language. Comments are not peer-reviewed. Acceptable comments are posted to the journal website only. Comments are not published in an issue and are not indexed in PubMed. Comments should be no longer than 500 words and will only be posted online. References are limited to 10. Authors of the article cited in the comment will be invited to reply, as appropriate. Comments and feedback on AHA/ASA Scientific Statements and Guidelines should be directed to the AHA/ASA Manuscript Oversight Committee via its Correspondence page. Sign In to Submit a Response to This Article Previous Back to top Next FiguresReferencesRelatedDetails February 6, 2024Vol 149, Issue 6 Advertisement Article Information Metrics © 2024 American Heart Association, Inc.https://doi.org/10.1161/CIRCULATIONAHA.123.065489PMID: 38315763 Originally publishedFebruary 5, 2024 Keywordsartificial intelligencecardiologyelectrocardiographyhistory, 20th centuryPDF download Advertisement Subjects Electrophysiology