GenECG: a synthetic image-based ECG dataset to augment artificial intelligence-enhanced algorithm development

Neil Bodagh; Kyaw Soe Tun; Adam Barton; Malihe Javidi; Darwon Rashid; Rachel Burns; Irum Kotadia; Magda Klis; Ali Gharaviri; Vinush Vigneswaran; Steven Niederer; Mark O'Neill; Miguel O Bernabeu; Steven E Williams

doi:10.1136/bmjhci-2024-101335

GenECG: a synthetic image-based ECG dataset to augment artificial intelligence-enhanced algorithm development

Neil Bodagh^*, Kyaw Soe Tun, Adam Barton, Malihe Javidi, Darwon Rashid, Rachel Burns, Irum Kotadia, Magda Klis, Ali Gharaviri, Vinush Vigneswaran, Steven Niederer, Mark O'Neill, Miguel O Bernabeu, Steven E Williams

^*Corresponding author for this work

Research output: Contribution to journal › Article › peer-review

Abstract

OBJECTIVES: An image-based ECG dataset incorporating visual imperfections common to paper-based ECGs, which are typically scanned or photographed into electronic health records, could facilitate clinically useful artificial intelligence (AI)-ECG algorithm development. This study aimed to create a high-fidelity, synthetic image-based ECG dataset.

METHODS: ECG images were recreated from the PTB-XL database, a signal-based dataset and image manipulation techniques were applied to mimic imperfections associated with ECGs in real-world settings. Clinical Turing tests were conducted to evaluate the fidelity of the synthetic images, and the performance of current AI-ECG algorithms was assessed using synthetic images containing visual imperfections.

RESULTS: GenECG, an image-based dataset containing 21 799 ECGs with visual imperfections encountered in routine clinical care paired with imperfection-free images, was created. Turing tests confirmed the realism of the images: expert observer accuracy of discrimination between real-world and synthetic ECGs fell from 63.9% (95% CI 58.0% to 69.8%) to 53.3% (95% CI 48.6% to 58.1%) over three rounds of testing, indicating that observers could not distinguish between synthetic and real ECGs. The performance of pre-existing algorithms on synthetic (area under the curve (AUC) 0.592, 95% CI 0.421 to 0.763) and real-world (AUC 0.647, 95% CI 0.520 to 0.774) ECG images containing imperfections was limited. Algorithm fine-tuning with GenECG data improved real-world ECG classification accuracy (AUC 0.821, 95% CI 0.730 to 0.913) demonstrating its potential to augment image-based algorithm development.

DISCUSSION/CONCLUSION: GenECG is the first synthetic image-based ECG dataset to pass a clinical Turing test. The dataset will enable image-based AI-ECG algorithm development, ensuring utility in low resource areas, prehospital settings and hospital environments where signal data are unavailable.

Original language	English
Article number	e101335
Number of pages	8
Journal	BMJ Health & Care Informatics
Volume	32
Issue number	1
DOIs	https://doi.org/10.1136/bmjhci-2024-101335
Publication status	Published - 31 May 2025

Keywords / Materials (for Non-textual outputs)

Humans
Electrocardiography/methods
Algorithms
Artificial Intelligence
Electronic Health Records
Databases, Factual

Access to Document

10.1136/bmjhci-2024-101335Licence: Creative Commons: Attribution (CC-BY)

GenECG_BODAGH_DOA12052025_VOR_CC-BYFinal published version, 1.15 MBLicence: Creative Commons: Attribution (CC-BY)

https://informatics.bmj.com/content/32/1/e101335Licence: Creative Commons: Attribution (CC-BY)

Cite this

Bodagh, N., Tun, K. S., Barton, A., Javidi, M., Rashid, D., Burns, R., Kotadia, I., Klis, M., Gharaviri, A., Vigneswaran, V., Niederer, S., O'Neill, M., Bernabeu, M. O., & Williams, S. E. (2025). GenECG: a synthetic image-based ECG dataset to augment artificial intelligence-enhanced algorithm development. BMJ Health & Care Informatics, 32(1), Article e101335. https://doi.org/10.1136/bmjhci-2024-101335

@article{bd212b3817dc429bb2d4e59ef70c285e,

title = "GenECG: a synthetic image-based ECG dataset to augment artificial intelligence-enhanced algorithm development",

abstract = "OBJECTIVES: An image-based ECG dataset incorporating visual imperfections common to paper-based ECGs, which are typically scanned or photographed into electronic health records, could facilitate clinically useful artificial intelligence (AI)-ECG algorithm development. This study aimed to create a high-fidelity, synthetic image-based ECG dataset.METHODS: ECG images were recreated from the PTB-XL database, a signal-based dataset and image manipulation techniques were applied to mimic imperfections associated with ECGs in real-world settings. Clinical Turing tests were conducted to evaluate the fidelity of the synthetic images, and the performance of current AI-ECG algorithms was assessed using synthetic images containing visual imperfections.RESULTS: GenECG, an image-based dataset containing 21 799 ECGs with visual imperfections encountered in routine clinical care paired with imperfection-free images, was created. Turing tests confirmed the realism of the images: expert observer accuracy of discrimination between real-world and synthetic ECGs fell from 63.9\% (95\% CI 58.0\% to 69.8\%) to 53.3\% (95\% CI 48.6\% to 58.1\%) over three rounds of testing, indicating that observers could not distinguish between synthetic and real ECGs. The performance of pre-existing algorithms on synthetic (area under the curve (AUC) 0.592, 95\% CI 0.421 to 0.763) and real-world (AUC 0.647, 95\% CI 0.520 to 0.774) ECG images containing imperfections was limited. Algorithm fine-tuning with GenECG data improved real-world ECG classification accuracy (AUC 0.821, 95\% CI 0.730 to 0.913) demonstrating its potential to augment image-based algorithm development.DISCUSSION/CONCLUSION: GenECG is the first synthetic image-based ECG dataset to pass a clinical Turing test. The dataset will enable image-based AI-ECG algorithm development, ensuring utility in low resource areas, prehospital settings and hospital environments where signal data are unavailable.",

keywords = "Humans, Electrocardiography/methods, Algorithms, Artificial Intelligence, Electronic Health Records, Databases, Factual",

author = "Neil Bodagh and Tun, \{Kyaw Soe\} and Adam Barton and Malihe Javidi and Darwon Rashid and Rachel Burns and Irum Kotadia and Magda Klis and Ali Gharaviri and Vinush Vigneswaran and Steven Niederer and Mark O'Neill and Bernabeu, \{Miguel O\} and Williams, \{Steven E\}",

note = "{\textcopyright} Author(s) (or their employer(s)) 2025. Re-use permitted under CC BY. Published by BMJ Group.",

year = "2025",

month = may,

day = "31",

doi = "10.1136/bmjhci-2024-101335",

language = "English",

volume = "32",

journal = "BMJ Health \& Care Informatics",

issn = "2632-1009",

publisher = "BMJ Publishing Group",

number = "1",

}

Bodagh, N, Tun, KS, Barton, A, Javidi, M, Rashid, D, Burns, R, Kotadia, I, Klis, M, Gharaviri, A, Vigneswaran, V, Niederer, S, O'Neill, M, Bernabeu, MO & Williams, SE 2025, 'GenECG: a synthetic image-based ECG dataset to augment artificial intelligence-enhanced algorithm development', BMJ Health & Care Informatics, vol. 32, no. 1, e101335. https://doi.org/10.1136/bmjhci-2024-101335

TY - JOUR

T1 - GenECG

T2 - a synthetic image-based ECG dataset to augment artificial intelligence-enhanced algorithm development

AU - Bodagh, Neil

AU - Tun, Kyaw Soe

AU - Barton, Adam

AU - Javidi, Malihe

AU - Rashid, Darwon

AU - Burns, Rachel

AU - Kotadia, Irum

AU - Klis, Magda

AU - Gharaviri, Ali

AU - Vigneswaran, Vinush

AU - Niederer, Steven

AU - O'Neill, Mark

AU - Bernabeu, Miguel O

AU - Williams, Steven E

PY - 2025/5/31

Y1 - 2025/5/31

N2 - OBJECTIVES: An image-based ECG dataset incorporating visual imperfections common to paper-based ECGs, which are typically scanned or photographed into electronic health records, could facilitate clinically useful artificial intelligence (AI)-ECG algorithm development. This study aimed to create a high-fidelity, synthetic image-based ECG dataset.METHODS: ECG images were recreated from the PTB-XL database, a signal-based dataset and image manipulation techniques were applied to mimic imperfections associated with ECGs in real-world settings. Clinical Turing tests were conducted to evaluate the fidelity of the synthetic images, and the performance of current AI-ECG algorithms was assessed using synthetic images containing visual imperfections.RESULTS: GenECG, an image-based dataset containing 21 799 ECGs with visual imperfections encountered in routine clinical care paired with imperfection-free images, was created. Turing tests confirmed the realism of the images: expert observer accuracy of discrimination between real-world and synthetic ECGs fell from 63.9% (95% CI 58.0% to 69.8%) to 53.3% (95% CI 48.6% to 58.1%) over three rounds of testing, indicating that observers could not distinguish between synthetic and real ECGs. The performance of pre-existing algorithms on synthetic (area under the curve (AUC) 0.592, 95% CI 0.421 to 0.763) and real-world (AUC 0.647, 95% CI 0.520 to 0.774) ECG images containing imperfections was limited. Algorithm fine-tuning with GenECG data improved real-world ECG classification accuracy (AUC 0.821, 95% CI 0.730 to 0.913) demonstrating its potential to augment image-based algorithm development.DISCUSSION/CONCLUSION: GenECG is the first synthetic image-based ECG dataset to pass a clinical Turing test. The dataset will enable image-based AI-ECG algorithm development, ensuring utility in low resource areas, prehospital settings and hospital environments where signal data are unavailable.

AB - OBJECTIVES: An image-based ECG dataset incorporating visual imperfections common to paper-based ECGs, which are typically scanned or photographed into electronic health records, could facilitate clinically useful artificial intelligence (AI)-ECG algorithm development. This study aimed to create a high-fidelity, synthetic image-based ECG dataset.METHODS: ECG images were recreated from the PTB-XL database, a signal-based dataset and image manipulation techniques were applied to mimic imperfections associated with ECGs in real-world settings. Clinical Turing tests were conducted to evaluate the fidelity of the synthetic images, and the performance of current AI-ECG algorithms was assessed using synthetic images containing visual imperfections.RESULTS: GenECG, an image-based dataset containing 21 799 ECGs with visual imperfections encountered in routine clinical care paired with imperfection-free images, was created. Turing tests confirmed the realism of the images: expert observer accuracy of discrimination between real-world and synthetic ECGs fell from 63.9% (95% CI 58.0% to 69.8%) to 53.3% (95% CI 48.6% to 58.1%) over three rounds of testing, indicating that observers could not distinguish between synthetic and real ECGs. The performance of pre-existing algorithms on synthetic (area under the curve (AUC) 0.592, 95% CI 0.421 to 0.763) and real-world (AUC 0.647, 95% CI 0.520 to 0.774) ECG images containing imperfections was limited. Algorithm fine-tuning with GenECG data improved real-world ECG classification accuracy (AUC 0.821, 95% CI 0.730 to 0.913) demonstrating its potential to augment image-based algorithm development.DISCUSSION/CONCLUSION: GenECG is the first synthetic image-based ECG dataset to pass a clinical Turing test. The dataset will enable image-based AI-ECG algorithm development, ensuring utility in low resource areas, prehospital settings and hospital environments where signal data are unavailable.

KW - Humans

KW - Electrocardiography/methods

KW - Algorithms

KW - Artificial Intelligence

KW - Electronic Health Records

KW - Databases, Factual

U2 - 10.1136/bmjhci-2024-101335

DO - 10.1136/bmjhci-2024-101335

M3 - Article

C2 - 40451261

SN - 2632-1009

VL - 32

JO - BMJ Health & Care Informatics

JF - BMJ Health & Care Informatics

IS - 1

M1 - e101335

ER -

GenECG: a synthetic image-based ECG dataset to augment artificial intelligence-enhanced algorithm development

Abstract

Keywords / Materials (for Non-textual outputs)

Access to Document

Fingerprint

Cite this