Clinically Meaningful Interpretability of an AI Model for ECG Classification

Overview

Journal NPJ Digit Med

Specialty Medical Informatics

Date 2025 Feb 17

PMID 39962214

Authors

Vadim Gliner

Idan Levy

Kenta Tsutsui

Moshe Rav Acha

Jorge Schliamser

Assaf Schuster

Yael Yaniv

Affiliations

Soon will be listed here.

Abstract

Despite the high accuracy of AI-based automated analysis of 12-lead ECG images for classification of cardiac conditions, clinical integration of such tools is hindered by limited interpretability of model recommendations. We aim to demonstrate the feasibility of a generic, clinical resource interpretability tool for AI models analyzing digitized 12-lead ECG images. To this end, we utilized the sensitivity of the Jacobian matrix to compute the gradient of the classifier for each pixel and provide medical relevance interpretability. Our methodology was validated using a dataset consisting of 79,226 labeled scanned ECG images, 11,316 unlabeled and 1807 labeled images obtained via mobile camera in clinical settings. The tool provided interpretability for both morphological and arrhythmogenic conditions, highlighting features in terms understandable to physician. It also emphasized significant signal features indicating the absence of certain cardiac conditions. High correlation was achieved between our method of interpretability and gold standard interpretations of 3 electrophysiologists.

References

Rajkomar A, Dean J, Kohane I . Machine Learning in Medicine. N Engl J Med. 2019; 380(14):1347-1358. DOI: 10.1056/NEJMra1814259. View

Goldberger A, Amaral L, Glass L, Hausdorff J, Ivanov P, Mark R . PhysioBank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals. Circulation. 2000; 101(23):E215-20. DOI: 10.1161/01.cir.101.23.e215. View

Elul Y, Rosenberg A, Schuster A, Bronstein A, Yaniv Y . Meeting the unmet needs of clinicians from AI systems showcased for cardiology with deep-learning-based ECG analysis. Proc Natl Acad Sci U S A. 2021; 118(24). PMC: 8214673. DOI: 10.1073/pnas.2020620118. View

Siontis K, Suarez A, Sehrawat O, Ackerman M, Attia Z, Friedman P . Saliency maps provide insights into artificial intelligence-based electrocardiography models for detecting hypertrophic cardiomyopathy. J Electrocardiol. 2023; 81:286-291. DOI: 10.1016/j.jelectrocard.2023.07.002. View

Himmelreich J, Harskamp R . Diagnostic accuracy of the PMcardio smartphone application for artificial intelligence-based interpretation of electrocardiograms in primary care (AMSTELHEART-1). Cardiovasc Digit Health J. 2023; 4(3):80-90. PMC: 10282008. DOI: 10.1016/j.cvdhj.2023.03.002. View

Neves I, Folgado D, Santos S, Barandas M, Campagner A, Ronzio L . Interpretable heartbeat classification using local model-agnostic explanations on ECGs. Comput Biol Med. 2021; 133:104393. DOI: 10.1016/j.compbiomed.2021.104393. View

Sangha V, Mortazavi B, Haimovich A, Ribeiro A, Brandt C, Jacoby D . Automated multilabel diagnosis on electrocardiographic images and signals. Nat Commun. 2022; 13(1):1583. PMC: 8948243. DOI: 10.1038/s41467-022-29153-3. View

Smith S, Walsh B, Grauer K, Wang K, Rapin J, Li J . A deep neural network learning algorithm outperforms a conventional algorithm for emergency department electrocardiogram interpretation. J Electrocardiol. 2018; 52:88-95. DOI: 10.1016/j.jelectrocard.2018.11.013. View

Chung K, Walsh T, MASSIE E . WOLFF-PARKINSON-WHITE SYNDROME. Am Heart J. 1965; 69:116-33. DOI: 10.1016/0002-8703(65)90224-3. View

10.

Ribeiro A, Horta Ribeiro M, Paixao G, Oliveira D, Gomes P, Canazart J . Automatic diagnosis of the 12-lead ECG using a deep neural network. Nat Commun. 2020; 11(1):1760. PMC: 7145824. DOI: 10.1038/s41467-020-15432-4. View

11.

Gliner V, Makarov V, Avetisyan A, Schuster A, Yaniv Y . Using domain adaptation for classification of healthy and disease conditions from mobile-captured images of standard 12-lead electrocardiograms. Sci Rep. 2023; 13(1):14023. PMC: 10462630. DOI: 10.1038/s41598-023-40693-6. View

12.

Jastrzebski M . ECG and Pacing Criteria for Differentiating Conduction System Pacing from Myocardial Pacing. Arrhythm Electrophysiol Rev. 2021; 10(3):172-180. PMC: 8576513. DOI: 10.15420/aer.2021.26. View

13.

Madias J . Why the current diagnostic criteria of Takotsubo syndrome are outmoded: a proposal for new criteria. Int J Cardiol. 2014; 174(3):468-70. DOI: 10.1016/j.ijcard.2014.04.241. View

14.

Sun L, Lu Y, Yang K, Li S . ECG analysis using multiple instance learning for myocardial infarction detection. IEEE Trans Biomed Eng. 2012; 59(12):3348-56. DOI: 10.1109/TBME.2012.2213597. View

15.

Cheng Z, Zhu K, Tian Z, Zhao D, Cui Q, Fang Q . The findings of electrocardiography in patients with cardiac amyloidosis. Ann Noninvasive Electrocardiol. 2013; 18(2):157-62. PMC: 6932432. DOI: 10.1111/anec.12018. View

16.

Ayano Y, Schwenker F, Dufera B, Debelee T . Interpretable Machine Learning Techniques in ECG-Based Heart Disease Classification: A Systematic Review. Diagnostics (Basel). 2023; 13(1). PMC: 9818170. DOI: 10.3390/diagnostics13010111. View

17.

He J, Baxter S, Xu J, Xu J, Zhou X, Zhang K . The practical implementation of artificial intelligence technologies in medicine. Nat Med. 2019; 25(1):30-36. PMC: 6995276. DOI: 10.1038/s41591-018-0307-0. View

18.

Strodthoff N, Wagner P, Schaeffter T, Samek W . Deep Learning for ECG Analysis: Benchmarks and Insights from PTB-XL. IEEE J Biomed Health Inform. 2020; 25(5):1519-1528. DOI: 10.1109/JBHI.2020.3022989. View

19.

Siddiqi O, Ruberg F . Cardiac amyloidosis: An update on pathophysiology, diagnosis, and treatment. Trends Cardiovasc Med. 2017; 28(1):10-21. PMC: 5741539. DOI: 10.1016/j.tcm.2017.07.004. View