AbstractHeart disease is the world’s leading cause of death. Diagnostic models based on electrocardiograms (ECGs) are often limited by the scarcity of high-quality data and issues of data imbalance. To address these challenges, we propose a conditional generative adversarial network (CECG-GAN). This strategy enables the generation of samples that closely approximate the distribution of ECG data.

摘要心脏病是世界上主要的死亡原因。基于心电图（ECG）的诊断模型通常受到高质量数据稀缺和数据不平衡问题的限制。为了应对这些挑战，我们提出了一个有条件的生成对抗网络（CECG-GAN）。该策略可以生成与ECG数据分布非常接近的样本。

Additionally, CECG-GAN addresses waveform jitter, slow processing speeds, and dataset imbalance issues through the integration of a transformer architecture. We evaluated this approach using two datasets: MIT-BIH and CSPC2020. The experimental results demonstrate that CECG-GAN achieves outstanding performance metrics.

此外，CECG-GAN通过集成变压器架构解决了波形抖动、处理速度慢和数据集不平衡问题。我们使用两个数据集评估了这种方法：MIT-BIH和CSPC2020。实验结果表明，CECG-GAN实现了出色的性能指标。

Notably, the percentage root mean square difference (PRD) reached 55.048, indicating a high degree of similarity between generated and actual ECG waveforms. Additionally, the Fréchet distance (FD) was approximately 1.139, the root mean square error (RMSE) registered at 0.232, and the mean absolute error (MAE) was recorded at 0.166..

值得注意的是，均方根差百分比（PRD）达到55.048，表明生成的ECG波形与实际ECG波形之间具有高度相似性。此外，Fréchet距离（FD）约为1.139，均方根误差（RMSE）记录为0.232，平均绝对误差（MAE）记录为0.166。。

IntroductionHeart disease has become the leading cause of death globally, with a notable increase in its prevalence among younger populations in recent years. Furthermore, significant global demographic shifts, such as population aging and growth, have been observed over the past three decades. According to the World Heart Report 2023 published by the World Heart Federation, cardiovascular disease (CVD) fatalities have escalated from approximately 12.1 million in 1990 to about 20.5 million in 2021.

引言心脏病已成为全球死亡的主要原因，近年来在年轻人群中的患病率显着增加。此外，在过去的三十年中，已经观察到了重大的全球人口变化，例如人口老龄化和增长。根据世界心脏联合会发布的《2023年世界心脏报告》，心血管疾病（CVD）死亡率已从1990年的约1210万上升到2021年的约2050万。

Sudden cardiac death and ischemic heart disease constitute 85% of these deaths worldwide. Diagnosis and treatment of such diseases predominantly depend on professional analysis of electrocardiograms (ECGs), which record the heart’s electrophysiological activity over time through skin-placed electrodes.

心脏猝死和缺血性心脏病占全球死亡人数的85%。此类疾病的诊断和治疗主要取决于心电图（ECG）的专业分析，心电图通过皮肤放置的电极记录心脏随时间的电生理活动。

ECGs are increasingly recognized as vital in cardiology therapeutics. However, the medical field faces notable challenges: (1) Human cardiac activity is constantly and rapidly changing, making manual data analysis by medical professionals highly challenging; (2) Machine learning-based detection algorithms necessitate extensive datasets for effective modeling, and manual data labeling incurs substantial time costs and raises patient privacy concerns.Machine learning algorithms are now gradually making a difference in the field of medical diagnostics with their automatic modelling benefits, such as BP neural networks, decision trees, temporal memory networks and other methods.

心电图在心脏病学治疗中越来越被认为是至关重要的。然而，医学领域面临着显着的挑战：（1）人类心脏活动不断快速变化，使得医学专业人员的手动数据分析极具挑战性；（2） 基于机器学习的检测算法需要大量数据集才能进行有效建模，而手动数据标记会产生大量的时间成本，并引发患者隐私问题。机器学习算法凭借其自动建模的优势，如BP神经网络、决策树、时间记忆网络和其他方法，正在医学诊断领域逐渐产生影响。

However, these methods necessitate extensive ECG data for training purposes. The classification and labeling of ECGs involve considerable time and resources from medical professionals. For instance, constructing a cardiac disease classification model requires a substantial dataset of ECG samples. This need .

然而，这些方法需要大量的ECG数据用于训练目的。心电图的分类和标记涉及医学专业人员的大量时间和资源。例如，构建心脏病分类模型需要大量的ECG样本数据集。这需要。

(1) The imbalance in current heart rate abnormality datasets significantly hampers the effectiveness of existing classification methods. The current imbalance in heart rate anomaly datasets severely impacts the validity of existing classification methods. This leads to low actual accuracy of scarcity types when assessing heart rate metrics..

（1） 当前心率异常数据集的不平衡严重阻碍了现有分类方法的有效性。当前心率异常数据集的不平衡严重影响了现有分类方法的有效性。这导致在评估心率指标时，稀缺类型的实际准确性较低。。

(2) Existing generative ECG model effects still suffer from the problem of imbalance, which exacerbates the negative impact of model performance when training the classification model, resulting in the existing heart rate classification algorithms being heavily biased towards the majority class results, making it difficult to differentiate between new anomalous data..

（2） 现有的生成性ECG模型效应仍然存在不平衡的问题，这加剧了训练分类模型时模型性能的负面影响，导致现有的心率分类算法严重偏向大多数类结果，难以区分新的异常数据。。

(3) Predominantly, existing heart rate generation models utilize recurrent neural networks (RNNs) and convolutional neural networks (CNNs), with a primary focus on sequential output. This approach is both time-intensive and inefficient, leading to cumulative generation errors and resulting in jittery waveforms..

（3） 主要是，现有的心率生成模型利用递归神经网络（RNN）和卷积神经网络（CNN），主要关注顺序输出。这种方法既耗时又效率低下，导致累积的生成错误并导致波形抖动。。

To address these issues, we propose a novel heart rate generation strategy utilizing conditional generative adversarial networks. This model integrates a transformer architecture with conditional constraints, enabling the generative adversarial network to more accurately approximate real data distributions.

为了解决这些问题，我们提出了一种利用条件生成对抗网络的新型心率生成策略。该模型将变压器体系结构与条件约束相结合，使生成对抗网络能够更准确地近似真实数据分布。

This approach not only captures a broader range of scarce data distributions but also preserves data diversity. Consequently, it mitigates the performance degradation of classification models caused by data imbalances and addresses the issues related to prolonged output times and subpar results in existing models.MethodsAnalyses of imbalanced data distributionThe MIT-BIH arrhythmia dataset, widely utilized in arrhythmia classification research, comprises recordings from 47 individuals, each contributing a roughly 30 min arrhythmia recording.

这种方法不仅捕获了更广泛的稀缺数据分布，而且保留了数据的多样性。因此，它减轻了由数据不平衡引起的分类模型的性能下降，并解决了与现有模型中延长的输出时间和低于标准结果相关的问题。方法不平衡数据分布分析广泛用于心律失常分类研究的MIT-BIH心律失常数据集包括来自47个人的记录，每个人大约有30分钟的心律失常记录。

This dataset encapsulates a total of 109,500 cardiac beats, with approximately 30% classified as abnormal beats. It includes five types of cardiac beats: normal beats (N), atrial premature beats (A), ventricular premature beats (V), left bundle-branch block (L), and right bundle-branch block (R). Its validity has been established, making it a benchmark dataset in the study of cardiac arrhythmias.In this experiment, all cyclic waveforms collected were referenced to the R-peak identified within the dataset.

该数据集封装了总共109500次心跳，其中约30%被归类为异常心跳。它包括五种类型的心脏搏动：正常搏动（N），房性早搏（A），室性早搏（V），左束支传导阻滞（L）和右束支传导阻滞（R）。它的有效性已经建立，使其成为心律失常研究的基准数据集。在该实验中，收集的所有循环波形均参考数据集中识别的R峰。

One hundred time points were captured before the R-peak, and two hundred time points were captured following the R-peak. Thus, a complete cyclic waveform was constructed through these three hundred time points.In our study, we applied wavelet transform techniques to denoise signals in the MIT-BIH arrhythmia database, aiming to enhance the quality of the electrocardiogram (ECG) signals.

在R峰之前捕获了100个时间点，在R峰之后捕获了200个时间点。因此，通过这300个时间点构建了完整的循环波形。在我们的研究中，我们将小波变换技术应用于MIT-BIH心律失常数据库中的信号去噪，旨在提高心电图（ECG）信号的质量。

We chose the fifth-order Daubechies wavelet.

我们选择了五阶Daubechies小波。

Data availability

数据可用性

The MIT-BIH and CSPC2020 datasets mentioned in this paper are both public datasets. They can be downloaded from the following addresses: https://www.physionet.org/content/mitdb/1.0.0/ and http://2020.icbeb.org/CSPC2020.

本文提到的MIT-BIH和CSPC020数据集都是公共数据集。可从以下地址下载：https://www.physionet.org/content/mitdb/1.0.0/和http://2020.icbeb.org/CSPC2020.

ReferencesDelaney AM, Brophy E, Ward TE. Synthesis of realistic ECG using generative adversarial networks. arxiv preprint https://arxiv.org/abs/1909.09150. (2019).Hazra, D. & Byun, Y. C. SynSigGAN: Generative adversarial networks for synthetic biomedical signal generation. Biology 9(12), 441 (2020).Article .

参考文献Delaney AM，Brophy E，Ward TE。使用生成对抗网络合成真实心电图。arxiv预印本https://arxiv.org/abs/1909.09150.（2019年）。Hazra，D。＆Byun，Y.C。SynSigGAN：用于合成生物医学信号产生的生成对抗网络。生物学9（12），441（2020）。。

PubMed

PubMed

PubMed Central

公共医学中心

Google Scholar

谷歌学者

Zhu, F. et al. Electrocardiogram generation with a bidirectional LSTM-CNN generative adversarial network. Sci. Rep. 9(1), 6734 (2019).Article

Zhu，F。等。双向LSTM-CNN生成对抗网络的心电图生成。科学。代表9（1），6734（2019）。文章

ADS

广告

PubMed

PubMed

PubMed Central

公共医学中心

Google Scholar

谷歌学者

Li, X. et al. Tts-gan: A transformer-based time-series generative adversarial network. In International Conference on Artificial Intelligence in Medicine 133–143 (Springer International Publishing, 2022).

Li，X.等人，《Tts-gan：一种基于变压器的时间序列生成对抗网络》，载于国际医学人工智能会议133-143（Springer International Publishing，2022）。

Google Scholar

谷歌学者

Adib, E., Afghah, F. & Prevost, J. J. Arrhythmia classification using cgan-augmented ecg signals. In 2022 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) 1865–1872 (IEEE, 2022).Chapter

Adib，E.，Afghah，F。＆Prevost，J。J。使用cgan增强的ecg信号进行心律失常分类。2022年IEEE生物信息学和生物医学国际会议（BIBM）1865-1872（IEEE，2022）。第章

Google Scholar

谷歌学者

Yoon J, Jarrett D, Van der Schaar M. Time-series generative adversarial networks. Adv. Neural Inf. Process. Syst., 32. (2019).Goodfellow I, Pouget-Abadie J, Mirza M et al. Generative adversarial nets. Adv. Neural Inf. Process. Syst., 27. (2014).Mirza M, Osindero S. Generative adversarial networks.

Yoon J，Jarrett D，Van der Schaar M.时间序列生成对抗网络。高级神经信息处理。系统。，32（2019年）。Goodfellow I，Pouget-Abadie J，Mirza M等人。生成对抗网。高级神经信息处理。系统。，27（2014年）。Mirza M，Osindero S.生成对抗网络。

Comput. Sci., 2672–2680. (2014).Vaswani A, Shazeer N, Parmar N et al. Attention is all you need. Adv. Neural Inf. Process. Syst., 30. (2017).Download referencesFundingThis work is supported by the National Natural Science Foundation of China, Ye Qisun Foundation, Project Grant No. U2141231, the Fund of Education Department of Jilin Province, Scientific Research Program of Jilin Provincial Department of Education, Fund No.

计算机。科学。，2672–2680。（2014年）。Vaswani A，Shazeer N，Parmar N等人。注意力就是你所需要的。高级神经信息处理。系统。，30（2017年）。下载参考文献基金这项工作得到了国家自然科学基金、叶启孙基金、项目资助号U2141231、吉林省教育厅基金、吉林省教育厅科研计划、基金号。

JJKH20241673KJ, and Jilin Science and Technology Development Program Project, Project No.:20230201076GX, 20240305046YY.Author informationAuthors and AffiliationsSchool of Electronic Information Engineering, Changchun University of Science and Technology, Changchun, 130022, ChinaYang Yang, Tianyu Lan, Yang Wang, Fengtian Li, Liyan Liu, Xupeng Huang & Shuhua JiangChangchun University of Architecture and Civil Engineering, Changchun, 130607, ChinaYang Yang & Fei GaoJilin Province Advanced Control Technology and Intelligent Automation Equipment Research Engineering Lab, Changchun, 130022, ChinaTianyu Lan, Yang Wang, Fengtian Li, Liyan Liu & Xupeng HuangTongfang Nuctech Co., Beijing, 100084, ChinaZhijun ZhangDepartment of Cardiology, FAW General Hospital, Changchun, 130011, Jilin, ChinaZhijun Zhang & Xing ChenAuthorsYang YangView author publicationsYou can also search for this author in.

JJKH20241673KJ和吉林省科学技术发展计划项目，项目编号：20230201076GX，20240305046YY。作者信息作者和附属机构长春科技大学电子信息工程学院，长春，130022，中国杨扬，田玉兰，杨旺，李凤田，刘丽艳，黄旭鹏和舒华江长春建筑与土木工程大学，长春，130607，中国杨扬和高飞吉林省先进控制技术和智能自动化设备研究工程实验室，长春，130022，中国田玉兰，杨旺，李凤田，刘丽艳和徐鹏黄同芳Nuctech Co.，北京，100084，张志军一汽总医院心内科，长春，130011，吉林，张志军和陈兴作者Yang YangView作者出版物您也可以在中搜索此作者。

PubMed Google ScholarTianyu LanView author publicationsYou can also search for this author in

PubMed Google ScholarTianyu LanView作者出版物您也可以在

PubMed Google ScholarYang WangView author publicationsYou can also search for this author in

PubMed Google ScholarYang WangView作者出版物您也可以在

PubMed Google ScholarFengtian LiView author publicationsYou can also search for this author in

PubMed Google ScholarFengtian LiView作者出版物您也可以在

PubMed Google ScholarLiyan LiuView author publicationsYou can also search for this author in

PubMed Google ScholarLiyan LiuView作者出版物您也可以在

PubMed Google ScholarXupeng HuangView author publicationsYou can also search for this author in

PubMed谷歌学者黄旭鹏查看作者出版物您也可以在

PubMed Google ScholarFei GaoView author publicationsYou can also search for this author in

PubMed Google ScholarFei GaoView作者出版物您也可以在

PubMed Google ScholarShuhua JiangView author publicationsYou can also search for this author in

PubMed Google ScholarShuhua JiangView作者出版物您也可以在

PubMed Google ScholarZhijun ZhangView author publicationsYou can also search for this author in

PubMed谷歌学者张志军查看作者出版物您也可以在

PubMed Google ScholarXing ChenView author publicationsYou can also search for this author in

PubMed Google ScholarXing ChenView作者出版物您也可以在

PubMed Google ScholarContributionsConceptualization: Yang Yang, Tianyu Lan. Formal analysis: Yang Wang, Xupeng Huang. Funding acquisition: Fei Gao. Methodology: Yang Yang, Tianyu Lan, DianLi Wang. Project administration: Zhijun Zhang. Supervision: Yang Yang, Tianyu Lan, Xing Chen.

PubMed谷歌学术贡献概念：杨洋，田玉兰。形式分析：王阳，黄旭鹏。资金收购：高飞。方法论：杨洋，田玉兰，王殿丽。项目管理：张志军。监督：杨洋、田玉兰、邢晨。

Validation: Xupeng Huang, DianLi Wang. Visualization: Yang Wang, Fengtian Li. Writing—original draft: Yang Yang, Tianyu Lan. Writing—review & editing: Yang Yang, Tianyu Lan, Yang Wang, Xupeng Huang, Fengtian Li, Fei Gao, DianLi Wang, Zhijun Zhang, Xing Chen. We, the undersigned authors of the manuscript titled “Data Imbalance in Cardiac Health Diagnostics Using CECG-GAN” confirm our consent to publish this work.

验证：黄旭鹏，王殿丽。可视化：杨旺，李凤田。撰写原稿：杨洋，田玉兰。写作评论和编辑：杨扬，田玉兰，杨旺，黄旭鹏，李凤田，高飞，王殿丽，张志军，陈兴。我们，即题为“使用CECG-GAN进行心脏健康诊断的数据不平衡”的手稿的署名作者，确认同意发表这项工作。

We affirm that we have all contributed significantly to the research and the preparation of the manuscript, and we have approved the final version to be published. We agree that this manuscript is original, not previously published, nor under consideration for publication elsewhere. We collectively take responsibility for the integrity of the work as a whole.Corresponding authorCorrespondence to.

我们确认，我们都为手稿的研究和准备做出了重大贡献，并且我们已经批准了最终版本的发布。我们同意这份手稿是原创的，以前没有发表过，也没有考虑在其他地方发表。我们集体对整个工作的完整性负责。对应作者对应。

Xing Chen.Ethics declarations

邢晨。道德宣言

Competing interests

相互竞争的利益

The authors declare no competing interests.

作者声明没有利益冲突。

Additional informationPublisher's noteSpringer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Supplementary InformationSupplementary Information.Rights and permissions

Additional informationPublisher的noteSpringer Nature在已发布地图和机构隶属关系中的管辖权主张方面保持中立。补充信息补充信息。权限和权限

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.

。

The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

本文中的图像或其他第三方材料包含在文章的知识共享许可中，除非在材料的信用额度中另有说明。如果材料未包含在文章的知识共享许可中，并且您的预期用途不受法律法规的许可或超出许可用途，则您需要直接获得版权所有者的许可。

To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/..

要查看此许可证的副本，请访问http://creativecommons.org/licenses/by/4.0/..

Reprints and permissionsAbout this articleCite this articleYang, Y., Lan, T., Wang, Y. et al. Data imbalance in cardiac health diagnostics using CECG-GAN.

转载和许可本文引用本文Yang，Y.，Lan，T.，Wang，Y。等人。使用CECG-GAN进行心脏健康诊断的数据不平衡。

Sci Rep 14, 14767 (2024). https://doi.org/10.1038/s41598-024-65619-8Download citationReceived: 29 December 2023Accepted: 21 June 2024Published: 26 June 2024DOI: https://doi.org/10.1038/s41598-024-65619-8Share this articleAnyone you share the following link with will be able to read this content:Get shareable linkSorry, a shareable link is not currently available for this article.Copy to clipboard.

Sci Rep 1414767（2024）。https://doi.org/10.1038/s41598-024-65619-8Download收到引文日期：2023年12月29日接受日期：2024年6月21日发布日期：2024年6月26日OI：https://doi.org/10.1038/s41598-024-65619-8Share本文与您共享以下链接的任何人都可以阅读此内容：获取可共享链接对不起，本文目前没有可共享的链接。复制到剪贴板。

Provided by the Springer Nature SharedIt content-sharing initiative

由Springer Nature SharedIt内容共享计划提供

KeywordsHeart diseaseGenerative adversarial networksUnbalanced dataMulti-class classificationElectrocardiogram

关键词心脏病生成对手网络不平衡数据多类别分类心电图

CommentsBy submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

评论通过提交评论，您同意遵守我们的条款和社区指南。如果您发现有虐待行为或不符合我们的条款或准则，请将其标记为不合适。