基于双通道混合神经网络的房颤风险预测模型

doi:10.19678/j.issn.1000-3428.0066133

计算机工程 ›› 2023, Vol. 49 ›› Issue (11): 293-301. doi: 10.19678/j.issn.1000-3428.0066133

基于双通道混合神经网络的房颤风险预测模型

柯博文¹^,², 杨湘¹^,², 陈艳红³

1. 武汉科技大学计算机科学与技术学院, 武汉 430065
2. 国家新闻出版署富媒体数字出版内容组织与知识服务重点实验室, 北京 100038
3. 武汉亚洲心脏病医院心血管内科, 武汉 430022

收稿日期:2022-10-31 出版日期:2023-11-15 发布日期:2023-02-07
作者简介:
柯博文(1999—)，男，硕士研究生，主研方向为医疗大数据
杨湘，副教授、博士
陈艳红，博士、副主任医师
基金资助:
国家自然科学基金(U1836118); 武汉市重点研发计划(2022012202015070); 武汉市知识创新专项曙光计划项目(2023010201020409)

Risk Prediction Model of Atrial Fibrillation Based on Dual Channel Hybrid Neural Network

Bowen KE¹^,², Xiang YANG¹^,², Yanhong CHEN³

1. School of Computer Science and Technology, Wuhan University of Science and Technology, Wuhan 430065, China
2. The Key Laboratory of Rich-Media Knowledge Organization and Service of Digital Publishing Content Institute of Scientific and Technical Information of China, Beijing 100038, China
3. Internal Medicine-Cardiovascular Department, Wuhan Asia Heart Hospital, Wuhan 430022, China

Received:2022-10-31 Online:2023-11-15 Published:2023-02-07

摘要/Abstract

摘要：

心房颤动是一种具有隐秘性的心血管疾病，发病时具有较高的致命性，因此，对其进行预判和早筛尤为重要。电子健康病历(EHR)作为常规的检查结果记录方式，相较于心电图(ECG)数据，能使房颤的预测和筛查更具普遍性。现有的基于EHR的房颤风险预测方法缺乏对房颤重要指标的关注，同时传统的基于卷积的模型无法提取到医疗诊断之间的依赖关系。提出一种双通道混合神经网络学习模型FR-ANN。该模型分两个通道进行特征提取，一个通道利用注意力机制Attention提取医疗事件之间的潜在关系，另一个通道对房颤的部分重要指标进行特征提取，这些与房颤相关的重要指标在医生的协助下筛选得到。实验结果表明，所提模型在武汉亚洲心脏病医院的私有数据集上的AUC值为80.1%，F1值为68.1%，在MIT的公共数据集MIMIC-Ⅲ上的AUC值为71.4%，F1值为62.8%，相比基于EHR数据的疾病风险预测模型在房颤风险预测任务上的表现更好。此外，注意力机制的引入提供了事后可解释性，具有临床意义。

关键词: 心房颤动, 疾病分类, 电子健康病历, 注意力机制, 神经网络

Abstract:

Atrial fibrillation is a type of concealed cardiovascular disease with highly fatal consequences. Therefore, early prediction and screening of atrial fibrillation are particularly important. The Electronic Health Record (EHR), which is a routine recording method for examination results, can make prediction and screening of atrial fibrillation more universal than Electrocardiogram(ECG)data. Existing EHR-based risk prediction methods for atrial fibrillation do not consider important indicators of atrial fibrillation, and traditional convolution-based models cannot extract the dependence between medical diagnoses. This study proposes a dual-channel hybrid neural network learning model, called FR-ANN. In this model, two channels are used for feature extraction. One channel uses attention to extract the potential relationship between medical events, and the other channel extracts important indicators of atrial fibrillation. The atrial fibrillation-related important indicators were screened by doctors. The experimental results show that the proposed model has an Area Under Curve(AUC)value of 80.1% and an F1 value of 68.1% on the private data set of Wuhan Asian Heart Hospital, and an AUC value of 71.4% and an F1 value of 62.8% on the public data set of MIT-Ⅲ. Compared with the disease risk prediction model based on EHR data, the proposed method performed better in atrial fibrillation risk prediction tasks. In addition, the introduction of attention mechanisms provides ex-post interpretability and has clinical significance.

Key words: atrial fibrillation, disease classification, electronic medical record, attention mechanism, neural network

柯博文, 杨湘, 陈艳红. 基于双通道混合神经网络的房颤风险预测模型[J]. 计算机工程, 2023, 49(11): 293-301.

Bowen KE, Xiang YANG, Yanhong CHEN. Risk Prediction Model of Atrial Fibrillation Based on Dual Channel Hybrid Neural Network[J]. Computer Engineering, 2023, 49(11): 293-301.

http://www.ecice06.com/CN/Y2023/V49/I11/293

图/表 9

图1 数据切割过程

Fig.1 Process of data cutting

图2 FR-ANN模型结构

Fig.2 Structure of FR-ANN model

图3 LSTM模块的结构

Fig.3 Structure of LSTM module

图4 单指标特征分析通道的模块结构

Fig.4 Module structure of single index feature analysis channel

图5 消融实验评分图

Fig.5 Rating chart of ablation experiment

图6 FR-ANN自注意力层输出的注意力分布

Fig.6 Attention distribution outputing by FR-ANN self attention level

参考文献 26

1	LLOYD-JONES D M, WANG T J, LEIP E P, et al. Lifetime risk for development of atrial fibrillation: the Framingham heart study. Circulation, 2004, 110(9): 1042- 1046. doi: 10.1161/01.CIR.0000140263.20897.42
2	刘晓勃, 马长生. 2019 HRS/ACP心房颤动筛查和教育倡议的解读. 中国循环杂志, 2019, 34(S01): 4- 6.
	LIU X B, MA C S. Interpretation of the 2019 HRS/ACP atrial fibrillation screening and education initiative. Chinese Circulation Journal, 2019, 34(S01): 4- 6.
3	RAJKOMAR A, OREN E, CHEN K, et al. Scalable and accurate deep learning with electronic health records. Npj Digital Medicine, 2018, 1, 18- 23. doi: 10.1038/s41746-018-0029-1
4	LECUN Y, BOTTOU L, BENGIO Y, et al. Gradient-based learning applied to document recognition. Proceedings of the IEEE, 1998, 86(11): 2278- 2324. doi: 10.1109/5.726791
5	LEUNG PATRICK CHEUNG B, DAHL D. Deep learning from electronic medical records using attention-based cross-modal convolutional neural networks[C]//Proceedings of IEEE EMBS International Conference on Biomedical & Health Informatics. Washington D. C., USA: IEEE Press, 2018: 222-225.
6	HOCHREITER S, SCHMIDHUBER J. Long short-term memory. Neural Computation, 1997, 9(8): 1735- 1780. doi: 10.1162/neco.1997.9.8.1735
7	AN Y, HUANG N J, CHEN X L, et al. High-risk prediction of cardiovascular diseases via attention-based deep neural networks. IEEE/ACM Transactions on Computational Biology and Bioinformatics, 2021, 18(3): 1093- 1105. doi: 10.1109/TCBB.2019.2935059
8	HE R, WU X, SUN Z N, et al. Wasserstein CNN: learning invariant features for NIR-VIS face recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019, 41(7): 1761- 1773. doi: 10.1109/TPAMI.2018.2842770
9	GAN Z Y, YU Y, WANG R, et al. CNN-based speaker verification and speech recognition in Tibetan. Journal of Physics: Conference Series, 2020, 1693(1): 012180. doi: 10.1088/1742-6596/1693/1/012180
10	CICONTE G, CONTI M, EVANGELISTA M, et al. Atrial fibrillation in autoimmune rheumatic diseases: from pathogenesis to treatment. Reviews on Recent Clinical Trials, 2018, 13(3): 170- 175. doi: 10.2174/1574887113666180418110721
11	DE VOS C B, PISTERS R, NIEUWLAAT R, et al. Progression from paroxysmal to persistent atrial fibrillation clinical correlates and prognosis. Journal of the American College of Cardiology, 2010, 55(8): 725- 731. doi: 10.1016/j.jacc.2009.11.040
12	JOHNSON A E W, POLLARD T J, SHEN L, et al. MIMIC-III, a freely accessible critical care data base. Scientific Data, 2016, 3, 160035. doi: 10.1038/sdata.2016.35
13	XIE F, YUAN H, NING Y, et al. Deep learning for temporal data representation in electronic health records: a systematic review of challenges and methodologies. Journal of Biomedical Informatics, 2022, 126, 103980. doi: 10.1016/j.jbi.2021.103980
14	GAO J Y, XIAO C, GLASS L M, et al. COMPOSE: cross-modal pseudo-siamese network for patient trial matching[C]//Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. New York, USA: ACM Press, 2020: 803-812.
15	LANDI I, GLICKSBERG B S, LEE H C, et al. Deep representation learning of electronic health records to unlock patient stratification at scale. Npj Digital Medicine, 2020, 3, 96- 105. doi: 10.1038/s41746-020-0301-z
16	WANG Z, ZHU Y, LI D, et al. Feature rearrangement based deep learning system for predicting heart failure mortality. Computer Methods and Programs in Biomedicine, 2020, 191, 105383. doi: 10.1016/j.cmpb.2020.105383
17	AN Y, MAO Y, ZHANG L, et al. RAHM: relation augmented hierarchical multi-task learning framework for reasonable medication stocking. Journal of Biomedical Informatics, 2020, 108, 103502. doi: 10.1016/j.jbi.2020.103502
18	BARBIERI S, KEMP J, PEREZ-CONCHA O, et al. Benchmarking deep learning architectures for predicting readmission to the ICU and describing patients-at-risk. Scientific Reports, 2020, 10, 1111- 1119. doi: 10.1038/s41598-020-58053-z
19	CHU J B, DONG W, HUANG Z X. Endpoint prediction of heart failure using electronic health records. Journal of Biomedical Informatics, 2020, 109, 103518. doi: 10.1016/j.jbi.2020.103518
20	MA F L, CHITTA R, ZHOU J, et al. Dipole: diagnosis prediction in healthcare via attention-based bidirectional recurrent neural networks[C]//Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York, USA: ACM Press, 2017: 1903-1911.
21	XU E L, ZHAO S W, MEI J, et al. Multiple MACE risk prediction using multi-task recurrent neural network with attention[C]//Proceedings of IEEE International Conference on Healthcare Informatics. Washington D. C., USA: IEEE Press, 2019: 1-2.
22	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all You need[EB/OL]. [2022-09-12]. https://arxiv.org/abs/1706.03762.
23	CHEN P P, DONG W, WANG J L, et al. Interpretable clinical prediction via attention-based neural network. BMC Medical Informatics and Decision Making, 2020, 20(3): 1- 9.
24	XIANG Y, JI H Y, ZHOU Y J, et al. Asthma exacerbation prediction and risk factor analysis based on a time-sensitive, attentive neural network: retrospective cohort study. Journal of Medical Internet Research, 2020, 22(7): e16981. doi: 10.2196/16981
25	WANG Z C, LI H R, LIU L C, et al. Predictive multi-level patient representations from electronic health records[C]//Proceedings of IEEE International Conference on Bioinformatics and Biomedicine. Washington D. C., USA: IEEE Press, 2020: 987-990.
26	LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(2): 318- 327. doi: 10.1109/TPAMI.2018.2858826

[1]	朱孟栩, 张文豪, 李国洪, 顾行发, 余涛, 郑逢杰, 张丽丽, 吴俣, 邴芳飞, 唐健雄. 基于卷积神经网络的高分六号卫星多光谱图像压缩[J]. 计算机工程, 2023, 49(9): 287-294.
[2]	杨静, 陆铭华, 马洁琼, 吴金平, 刘星璇. 基于交替循环神经网络的水下防御态势预测方法[J]. 计算机工程, 2023, 49(9): 69-78.
[3]	李现国, 李滨. 基于Transformer和多尺度CNN的图像去模糊[J]. 计算机工程, 2023, 49(9): 226-233, 245.
[4]	孙龙, 张荣芬, 刘宇红, 饶庭漓. 监控视角下密集人群口罩佩戴检测算法[J]. 计算机工程, 2023, 49(9): 313-320.
[5]	隋国华, 李陶然, 刘昊, 陈林, 汪卫. 基于图表示学习的领域知识图谱推理技术研究[J]. 计算机工程, 2023, 49(9): 89-98.
[6]	苏晓东, 李世洲, 赵佳圆, 亮洪宇, 张玉荣, 徐红岩. 基于多级叠加和注意力机制的图像语义分割[J]. 计算机工程, 2023, 49(9): 265-271, 278.
[7]	杜逸潇, 王红军, 李修和. 基于频谱地图的辐射源指纹定位方法研究[J]. 计算机工程, 2023, 49(9): 183-190, 198.
[8]	高御尧, 石明全, 秦渝, 陈建平, 周喜, 张鹏. 基于改进非全连接神经网络的站点客流预测模型[J]. 计算机工程, 2023, 49(9): 43-51.
[9]	胡水. 基于深度强化学习的智能兵棋推演决策方法[J]. 计算机工程, 2023, 49(9): 303-312.
[10]	刘晓黎, 王轶彤. 基于自监督学习的多密度图会话推荐[J]. 计算机工程, 2023, 49(9): 60-68, 78.
[11]	韩璐, 霍纬纲, 张永会, 刘涛. 基于多尺度特征融合与双注意力机制的多元时间序列预测[J]. 计算机工程, 2023, 49(9): 99-108.
[12]	孔凌辉, 饶哲恒, 徐彦彦, 潘少明. 基于深度强化学习的无线网络智能路由算法[J]. 计算机工程, 2023, 49(9): 199-207, 216.
[13]	李哲铭, 王晋东, 侯建中, 李伟, 张世华, 张恒巍. 基于显著区域优化的对抗样本攻击方法[J]. 计算机工程, 2023, 49(9): 246-255, 264.
[14]	龙玉江, 卫薇, 舒彧, 张正刚, 王道累, 李峰. 基于自适应关键点的破损旋转绝缘子检测方法[J]. 计算机工程, 2023, 49(9): 272-278.
[15]	丰芳宇, 罗晓曙, 蒙志明, 王广宇. 基于抗混叠残差注意力网络的人脸表情识别[J]. 计算机工程, 2023, 49(8): 190-198.

选择文件类型/文献管理软件名称

选择包含的内容

基于双通道混合神经网络的房颤风险预测模型

Risk Prediction Model of Atrial Fibrillation Based on Dual Channel Hybrid Neural Network

RichHTML

PDF

可视化

被引次数

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 26

相关文章 15

编辑推荐

Metrics

本文评价

模态框（Modal）标题

选择文件类型/文献管理软件名称

选择包含的内容

基于双通道混合神经网络的房颤风险预测模型

Risk Prediction Model of Atrial Fibrillation Based on Dual Channel Hybrid Neural Network

RichHTML

PDF

可视化

被引次数

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 26

相关文章 15

编辑推荐

Metrics

本文评价