基于顺序耦合对抗学习的脓毒症序列生成方法

doi:10.19678/j.issn.1000-3428.0069967

摘要/Abstract

摘要： 脓毒症是一种由感染导致的危重症,是重症监护室(ICU)中患者死亡的主要原因之一。然而,在脓毒症治疗环境中,实际数据较难获取,存在临床数据匮乏的问题。为克服这些挑战,提出一种具有梯度惩罚的顺序耦合医疗Wasserstein生成对抗网络(SC-med WGAN),与现有工作侧重单步生成不同,强调对脓毒症患者状态和药物剂量的顺序生成,以更好地模拟临床数据的生成过程。该模型由两个耦合生成器组成,在统一模型中协调患者状态和药物剂量的生成。模型采用混合损失技巧,引入特征匹配损失和皮尔逊相关系数作为附加项,既考虑单个变量的实际分布,也考虑变量之间随时间的相关性。在包含17 898位脓毒症患者信息的重症监护医疗信息标记(MIMIC-Ⅲ)数据集上测试,并在贫血数据上进行验证,证明模型的准确性和鲁棒性。实验结果表明,该模型顺序生成的数据在质量和真实性上优于其他模型,揭示了患者状态和药物剂量数据的生成具有明显的相互影响这一临床事实。

关键词: 脓毒症, 数据生成, 顺序生成, 序列耦合对抗学习, 耦合生成器

Abstract: Sepsis is a critical condition caused by infection and is a leading cause of death in Intensive Care Units (ICUs). However, in the context of sepsis treatment, actual clinical data are challenging to obtain. To address this challenge, a Sequentially Coupled medical Wasserstein Generative Adversarial Network (SC-med WGAN) with a gradient penalty is proposed in this study. In contrast to existing models that focus on single-step generation, this model emphasizes the sequential generation of sepsis patient statuses and drug doses to improve the simulation of the process of generating clinical data. The SC-med WGAN consists of two coupled generators that coordinate the generation of patient status and drug dose in a unified model. Moreover, the model employs a mixed-loss technique that introduces feature-matched loss and Pearson's correlation coefficient as additional terms to account for the actual distribution of individual variables and the correlation between variables over time. Finally, the model is tested on the Medical Information Mark for Intensive Care-III (MIMIC-III) dataset, which contains 17 898 sepsis patient records. Additionally, the model is validated using anemia data, further demonstrating its accuracy and robustness. The experimental results show that the data generated sequentially by the proposed model are superior to those generated by other models in terms of quality and authenticity. The proposed method reveals a significant interaction between the generation of patient status and drug dose data.

Key words: sepsis, data generation, sequential generation, sequentially coupled adversarial learning, coupled generator

中图分类号:

TP183

张骐薇, 林彬, 刘云龙. 基于顺序耦合对抗学习的脓毒症序列生成方法[J]. 计算机工程, 2026, 52(2): 356-371.

ZHANG Qiwei, LIN Bin, LIU Yunlong. Sepsis Sequence Generation Method Based on Sequential Coupled Adversarial Learning[J]. Computer Engineering, 2026, 52(2): 356-371.

https://www.ecice06.com/CN/Y2026/V52/I2/356

参考文献

[1] JIA Y, BURDEN J, LAWTON T, et al. Safe reinforcement learning for sepsis treatment[C]//Proceedings of the IEEE International Conference on Healthcare Informatics. Washington D. C., USA: IEEE Press, 2020: 1-7.
[2] ZAHER S R, ELGOHARY D A, ANTONIOS M A M. Comparison between early and late mortalities due to severe sepsis in a pediatric intensive care unit: five-years-experience[J]. Egyptian Pediatric Association Gazette, 2022, 70(1): 17.
[3] SINGER M, DEUTSCHMAN C S, SEYMOUR C W, et al. The third international consensus definitions for sepsis and septic shock (sepsis-3)[J]. The Journal of the American Medical Association, 2016, 315(8): 801-810.
[4] KOMOROWSKI M, CELI L A, BADAWI O, et al. The artificial intelligence clinician learns optimal treatment strategies for sepsis in intensive care[J]. Nature Medicine, 2018, 24(11): 1716-1720.
[5] BUCZAK A L, BABIN S, MONIZ L. Data-driven approach for creating synthetic electronic medical records[J]. BMC Medical Informatics and Decision Making, 2010, 10: 59.
[6] WANG K F, GOU C, DUAN Y J, et al. Generative adversarial networks: introduction and outlook[J]. IEEE/CAA Journal of Automatica Sinica, 2017, 4(4): 588-598.
[7] KUO N I, POLIZZOTTO M N, FINFER S, et al. The health gym: synthetic health-related datasets for the development of reinforcement learning algorithms[J]. Scientific Data, 2022, 9: 693.
[8] CRESWELL A, WHITE T, DUMOULIN V, et al. Generative adversarial networks: an overview[J]. IEEE Signal Processing Magazine, 2018, 35(1): 53-65.
[9] ZHANG Z, YAN C, MESA D A, et al. Ensuring electronic medical record simulation through better training, modeling, and evaluation[J]. Journal of the American Medical Informatics Association, 2020, 27(1): 99-108.
[10] THANH-TUNG H, TRAN T, VENKATESH S. Improving generalization and stability of generative adversarial networks[EB/OL].[2024-05-05]. https://arxiv.org/pdf/1902.03984.
[11] 宋航, 周凤, 熊伟. 基于自相关-变分对抗学习的物理系统异常检测[J]. 计算机工程, 2024, 50(12): 358-366. SONG H, ZHOU F, XIONG W. Anomaly detection of physical systems based on autocorrelation-variance adversarial learning[J]. Computer Engineering, 2024, 50(12): 358-366. (in Chinese)
[12] CHALLEN R, DENNY J, PITT M, et al. Artificial intelligence, bias and clinical safety[J]. BMJ Quality Safety, 2019, 28(3): 231-237.
[13] KUO N I, JORM L, BARBIERI S. Synthetic health-related longitudinal data with mixed-type variables generated using diffusion models[EB/OL].[2024-05-05]. https://arxiv.org/pdf/2303.12281.
[14] SOHL-DICKSTEIN J, WEISS E, MAHESWARANATHAN N, et al. Deep unsupervised learning using nonequilibrium thermodynamics[C]//Proceedings of the International Conference on Machine Learning. Washington D. C., USA: IEEE Press, 2015: 2256-2265.
[15] BEAULIEU-JONES B K, WU Z S, WILLIAMS C, et al. Privacy-preserving generative deep neural networks support clinical data sharing[J]. Cardiovascular Quality and Outcomes, 2019, 12(7): e005122.
[16] CHOI E, BISWAL S, MALIN B, et al. Generating multi-label discrete patient records using generative adversarial networks[EB/OL].[2024-05-05]. https://arxiv.org/pdf/1703.06490.
[17] MACIAS-FASSIO E, MORALES A, PRUENZA C, et al. Privacy-preserving tabular data generation: application to sepsis detection[EB/OL].[2024-05-05]. https://arxiv.org/pdf/2404.16638.
[18] KILLIAN T W, PARBHOO S, GHASSEMI M. Risk sensitive dead-end identification in safety-critical offline reinforcement learning[EB/OL].[2024-05-05]. https://arxiv.org/pdf/2301.05664.
[19] BAOWALY M K, LIN C C, LIU C L, et al. Synthesizing electronic health records using improved generative adversarial networks[J]. Journal of the American Medical Informatics Association, 2019, 26(3): 228-241.
[20] ESTEBAN C, HYLAND S L, RATSCH G. Real-valued (medical) time series generation with recurrent conditional GANs[EB/OL].[2024-05-05]. https://arxiv.org/pdf/1706.02633.
[21] 刘帅威, 李智, 王国美, 等. 基于Transformer和GAN的对抗样本生成算法[J]. 计算机工程, 2024, 50(2): 180-187. LIU S W, LI Z, WANG G M, et al. Adversarial sample generation algorithm based on Transformer and GAN[J]. Computer Engineering, 2024, 50(2): 180-187. (in Chinese)
[22] DASH S, YALE A, GUYON I, et al. Medical time-series data generation using generative adversarial networks[M]. Berlin, Germany: Springer, 2020.
[23] ARJOVSKY M, CHINTALA S, BOTTOU L. Wasserstein generative adversarial networks[C]//Proceedings of the International Conference on Machine Learning. Washington D. C., USA: IEEE Press, 2017: 214-223.
[24] RASHIDIAN S, WANG F S, MOFFITT R, et al. SMOOTH-GAN: towards sharp and smooth synthetic EHR data generation[C]//Proceedings of the 18th International Conference on Artificial Intelligence in Medicine. Berlin, Germany: Springer, 2020: 37-48.
[25] WEI X, GONG B Q, LIU Z X, et al. Improving the improved training of Wasserstein GANs: a consistency term and its dual effect[EB/OL].[2024-05-05]. https://arxiv.org/pdf/1803.01541.
[26] YAHI A, VANGURI R, ELHADAD N, et al. Generative adversarial networks for electronic health records: a framework for exploring and evaluating methods for predicting drug-induced laboratory test trajectoriest[EB/OL].[2024-05-05]. https://arxiv.org/pdf/1712.00164.
[27] ODENA A, OLAH C, SHLENS J. Conditional image synthesis with auxiliary classifier GANs[C]//Proceedings of the International Conference on Machine Learning. Washington D. C., USA: IEEE Press, 2017: 2642-2651.
[28] RAGHU A, KOMOROWSKI M, CELI L, et al. Continuous state-space models for optimal sepsis treatment: a deep reinforcement learning approach[C]//Proceedings of the International Conference on Machine Learning for Healthcare. Washington D. C., USA: IEEE Press, 2017: 147-163.
[29] HOANG Q, NGUYEN T D, LE T, et al. Multi-generator generative adversarial nets[EB/OL].[2024-05-05]. https://arxiv.org/pdf/1708. 02556.
[30] GRAVES A. Supervised sequence labelling with recurrent neural networks[M]. Berlin, Germany: Springer, 2012.
[31] RUMELHART D E, HINTON G E, WILLIAMS R J. Learning representations by back-propagating errors[J]. Nature, 1986, 323(6088): 533-536.
[32] SALIMANS T, GOODFELLOW I, ZAREMBA W, et al. Improved techniques for training GANs[C]//Proceedings of the Advances in Neural Information Processing Systems. Cambridge, USA: MIT Press, 2016: 29.
[33] MUKAKA M M. A guide to appropriate use of correlation coefficient in medical research[J]. Malawi Medical Journal, 2012, 24(3): 69-71.
[34] VILLANI C. Optimal transport: old and new[M]. Berlin, Germany: Springer, 2009.
[35] CHEN Y Q, LIN Z H, MVLLER H G. Wasserstein regression[J]. Journal of the American Statistical Association, 2023, 118(542): 869-882.
[36] JOHNSON A E, POLLARD T J, SHEN L, et al. MIMIC-III, a freely accessible critical care database[J]. Scientific Data, 2016, 3: 160035.
[37] MUYAMA L, NEURAZ A, COULET A. Extracting diagnosis pathways from electronic health records using deep reinforcement learning[EB/OL].[2024-05-05]. https://arxiv.org/pdf/2305.06295.
[38] TAI C Y, WANG W J, HUANG Y M. Using time-series generative adversarial networks to synthesize sensing data for pest incidence forecasting on sustainable agriculture[J]. Sustainability, 2023, 15(10): 7834.
[39] GAO F, YANG Y, WANG J, et al. A deep convolutional generative adversarial networks-based semi-supervised method for object recognition in synthetic aperture radar images[J]. Remote Sensing, 2018, 10(6): 846.
[40] LIU Y S, WIJEWICKREMA S, LI A, et al. Time-transformer: integrating local and global features for better time series generation[C]//Proceedings of the 2024 SIAM International Conference on Data Mining. Philadelphia, USA: Society for Industrial and Applied Mathematics, 2024: 325-333.
[41] SOLATORIO A V, DUPRIEZ O. REaLTabFormer: generating realistic relational and tabular data using transformers[EB/OL].[2024-05-05]. https://arxiv.org/pdf/2302.02041.

选择文件类型/文献管理软件名称

选择包含的内容