Abstract
Background
Brugada syndrome is a major cause of sudden cardiac death in young people and has
distinctive electrocardiographic (ECG) features. We aimed to develop a deep learning–enabled
ECG model for automatic screening for Brugada syndrome to identify these patients
at an early point in time, thus allowing for life-saving therapy.
Methods
A total of 276 ECGs with a type 1 Brugada ECG pattern (276 type 1 Brugada ECGs and
another randomly retrieved 276 non-Brugada type ECGs for 1:1 allocation) were extracted
from the hospital-based ECG database for a 2-stage analysis with a deep learning model.
After trained network for identifying right bundle branch block pattern, we transferred
the first-stage learning to the second task to diagnose the type 1 Brugada ECG pattern.
The diagnostic performance of the deep learning model was compared with that of board-certified
practicing cardiologists. The model was further validated in an independent ECG data
set collected from hospitals in Taiwan and Japan.
Results
The diagnoses by the deep learning model (area under the receiver operating characteristic
curve [AUC] 0.96, sensitivity 88.4%, specificity 89.1%) were highly consistent with
the standard diagnoses (kappa coefficient 0.78). However, the diagnoses by the cardiologists
were significantly different from the standard diagnoses, with only moderate consistency
(kappa coefficient 0.63). In the independent ECG cohort, the deep learning model still
reached a satisfactory diagnostic performance (AUC 0.89, sensitivity 86.0%, specificity
90.0%).
Conclusions
We present the first deep learning–enabled ECG model for diagnosing Brugada syndrome,
which appears to be a robust screening tool with a diagnostic potential rivalling
trained physicians.
Résumé
Contexte
Le syndrome de Brugada est une cause majeure de mort subite cardiaque chez les jeunes
et présente des caractéristiques électrocardiographiques (ECG) distinctives. Nous
avons cherché à développer un modèle d'ECG basé sur l'apprentissage profond pour le
dépistage automatique du syndrome de Brugada afin d'identifier ces patients à un stade
précoce, ce qui permettrait de mettre en place un traitement salutaire.
Méthodes
Un total de 276 ECG avec un modèle d'ECG de Brugada de type 1 (276 ECG de Brugada
de type 1 et 276 autres ECG, non Brugada, collectés aléatoirement, pour une allocation
1:1) ont été extraits de la base de données d'ECG de l'hôpital pour une analyse en
deux étapes avec un modèle d'apprentissage profond. Après avoir entraîné le réseau
à identifier le tracé du bloc de branche droit, nous avons transposé l'apprentissage
de cette première étape vers une seconde tâche afin de diagnostiquer le tracé de l'ECG
d'un Brugada de type 1. La performance diagnostique du modèle d'apprentissage profond
a été comparée à celle de cardiologues praticiens certifiés. Le modèle a ensuite été
validé dans un ensemble indépendant de données ECG recueillies dans des hôpitaux de
Taïwan et du Japon.
Résultats
Les diagnostics établis par le modèle d'apprentissage profond (aire sous la courbe
de la fonction d'efficacité du récepteur [ASC] 0,96, sensibilité 88,4 %, spécificité
89,1 %) étaient très cohérents avec les diagnostics standard (coefficient kappa 0,78).
Cependant, les diagnostics des cardiologues étaient significativement différents des
diagnostics standards, moyennant une cohérence modérée (coefficient kappa 0,63). Dans
la cohorte d'ECG indépendants, le modèle d'apprentissage profond atteignait encore
une performance diagnostique satisfaisante (ASC 0,89, sensibilité 86,0 %, spécificité
90,0 %).
Conclusions
Nous présentons le premier modèle d'ECG basé sur l'apprentissage profond pour le diagnostic
du syndrome de Brugada, qui s'avère être un outil de dépistage robuste avec un potentiel
diagnostique rivalisant avec celui des médecins formés.
To read this article in full you will need to make a payment
Purchase one-time access:
Academic & Personal: 24 hour online accessCorporate R&D Professionals: 24 hour online accessOne-time access price info
- For academic or personal research use, select 'Academic and Personal'
- For corporate R&D use, select 'Corporate R&D Professionals'
Subscribe:
Subscribe to Canadian Journal of CardiologyAlready a print subscriber? Claim online access
Already an online subscriber? Sign in
Register: Create an account
Institutional Access: Sign in to ScienceDirect
References
- Present status of Brugada syndrome: JACC state-of-the-art review.J Am Coll Cardiol. 2018; 72: 1046-1059
- 2017 AHA/ACC/HRS guideline for management of patients with ventricular arrhythmias and the prevention of sudden cardiac death: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines and the Heart Rhythm Society.J Am Coll Cardiol. 2018; 72: e91-e220
- 2015 ESC guidelines for the management of patients with ventricular arrhythmias and the prevention of sudden cardiac death: the Task Force for the Management of Patients with Ventricular Arrhythmias and the Prevention of Sudden Cardiac Death of the European Society of Cardiology (ESC). Endorsed by: Association for European Paediatric and Congenital Cardiology (AEPC).Eur Heart J. 2015; 36: 2793-2867
- HRS/EHRA/APHRS expert consensus statement on the diagnosis and management of patients with inherited primary arrhythmia syndromes: document endorsed by HRS, EHRA, and APHRS in May 2013 and by ACCF, AHA, PACES, and AEPC in June 2013.Heart Rhythm. 2013; 10: 1932-1963
- Brugada syndrome: clinical and genetic findings.Genet Med. 2016; 18: 3-12
- A meta-analytic review of prevalence for Brugada ECG patterns and the risk for death.Medicine. 2016; 95: e5643
- About Brugada syndrome and its prevalence.Europace. 2012; 14: 925-928
- Sudden unexplained death: heritability and diagnostic yield of cardiological and genetic examination in surviving relatives.Circulation. 2005; 112: 207-213
- Pathogenesis and management of Brugada syndrome.Nat Rev Cardiol. 2016; 13: 744-756
- Programmed ventricular stimulation for risk stratification in the Brugada syndrome: a pooled analysis.Circulation. 2016; 133: 622-630
- Long-term prognosis of patients diagnosed with Brugada syndrome: results from the FINGER Brugada Syndrome Registry.Circulation. 2010; 121: 635-643
- Asymptomatic Brugada syndrome: clinical characterization and long-term prognosis.Circ Arrhythm Electrophysiol. 2015; 8: 1144-1150
- Brugada syndrome presenting as an “acute myocardial infarction.”.J Emerg Med. 2009; 37: 15-20
- Propafenone-induced Brugada-like ECG changes mistaken as acute myocardial infarction.Emerg Med J. 2008; 25: 117-118
- Cardiologist-level arrhythmia detection and classification in ambulatory electrocardiograms using a deep neural network.Nat Med. 2019; 25: 65-69
- Deep learning in cardiology.IEEE Rev Biomed Engin. 2019; 12: 168-193
- Unlocking the power of artificial intelligence and big data in medicine.J Med Internet Res. 2019; 21e16607
- Right bundle branch block: prevalence, risk factors, and outcome in the general population: results from the Copenhagen City Heart Study.Eur Heart J. 2013; 34: 138-146
- Interrater reliability: the kappa statistic.Biochem Med (Zagreb). 2012; 22: 276-282
- Deep Learning with Python.Manning Publications, 2017
- Brugada syndrome: report of the second consensus conference: endorsed by the Heart Rhythm Society and the European Heart Rhythm Association.Circulation. 2005; 111: 659-670
- Prevalence of the congenital long-QT syndrome.Circulation. 2009; 120: 1761-1767
- Long QT syndrome and associated gene mutation carriers in Japanese children: results from ECG screening examinations.Clin Sci (Lond). 2009; 117: 415-424
- Epidemiology of arrhythmias and sudden cardiac death in Asia.Circ J. 2013; 77: 2419-2431
- Criteria for evaluating rSr′ patterns due to high precordial ECG lead placement accurately confirm absence of a Brugada ECG pattern.J Electrocardiol. 2016; 49: 182-186
- Right bundle branch block, persistent ST segment elevation and sudden cardiac death: a distinct clinical and electrocardiographic syndrome. A multicenter report.J Am Coll Cardiol. 1992; 20: 1391-1396
- A survey on transfer learning.IEEE Trans Knowl Data Eng. 2010; 22: 1345-1359
- ECGNet: deep network for arrhythmia classification.2018 IEEE Int Symp Med Meas Appl (MeMeA) (2018). 2018; : 1-6
- Transfer learning in ECG classification from human to horse using a novel parallel neural network architecture.Sci Rep. 2020; 10: 186
Article info
Publication history
Published online: August 27, 2021
Accepted:
August 24,
2021
Received:
March 24,
2021
Footnotes
See editorial by Bleijendaal and Wilde, pages 149–151 of this issue.
See page 158 for disclosure information.
Identification
Copyright
© 2021 Canadian Cardiovascular Society. Published by Elsevier Inc. All rights reserved.
ScienceDirect
Access this article on ScienceDirectLinked Article
- From a Polish 3-Year-Old Boy Who Visited Maastricht to Automatic Detection Using Deep Learning: Brugada Syndrome Is Being RevolutionisedCanadian Journal of CardiologyVol. 38Issue 2
- PreviewIn 1992, Pedro and Josep Brugada described a typical electrocardiographic (ECG) pattern including right bundle branch block (RBBB) and persistent ST-segment elevation associated with sudden cardiac disease, without underlying electrolyte disturbances, ischemia, or structural heart disease. The syndrome, which later became widely known as the Brugada syndrome (BrS), first came to their attention in 1986 when a 3-year-old Polish boy named Lech and his father Andreas were referred to Professor Wellens in Maastricht, The Netherlands, owing to multiple episodes of syncope.
- Full-Text
- Preview
