|本期目录/Table of Contents|

[1]董孝彤,曲新亮,魏守水△.用于睡眠呼吸暂停检测的心电特征稳定性分析*[J].生物医学工程研究,2020,01:6-10.
 DONG Xiaotong,QU Xinliang,WEI Shoushui.Stability analysis of electrocardiogram features for sleep apnea detection[J].Journal of Biomedical Engineering Research,2020,01:6-10.
点击复制

用于睡眠呼吸暂停检测的心电特征稳定性分析*(PDF)

《生物医学工程研究》[ISSN:1006-6977/CN:61-1281/TN]

期数:
2020年01期
页码:
6-10
栏目:
出版日期:
2020-03-25

文章信息/Info

Title:
Stability analysis of electrocardiogram features for sleep apnea detection
文章编号:
1672-6278 (2020)01-0006 -05
作者:
董孝彤1曲新亮2魏守水1△
1. 山东大学控制科学与工程学院,济南 250061;2.山东第一医科大学第一附属医院,济南 250014
Author(s):
DONG Xiaotong1 QU Xinliang2 WEI Shoushui1
1.School of Control Science and Engineering,Shandong University, Jinan 250061,China;2.The First Affiliated Hospital of Shandong First Medical University,Jinan 250014
关键词:
心电信号睡眠呼吸暂停稳定特征选择支持向量机最小冗余最大相关
Keywords:
ElectrocardiogramSleep apneaStability feature selectionSupport vector machine Minimal-redundancy-maximal-relevance
分类号:
R318
DOI:
10.19529/j.cnki.1672-6278.2020.01.02
文献标识码:
A
摘要:
基于心电信号(electrocardiogram,ECG)的睡眠呼吸暂停检测具有十分重要的意义,很多研究致力于提高检测的准确率却忽视了特征的稳定性。本研究对用于睡眠呼吸暂停检测的心电特征进行稳定性分析,并建立呼吸暂停事件检测模型。基于集成稳定特征选择策略,将最小冗余最大相关(minimal-redundancy-maximal-relevance,mRMR)特征选择方法与稳健排序聚合(robust rank aggregation,RRA)方法结合,对45个心电特征进行稳定性分析。使用10折交叉验证及支持向量机(SVM)进行特征验证及检测模型建立。最终使用14个特征建立分类模型,在独立测试集上实现Acc=90.03%,Se=86.71%,Sp=91.73%,所选特征在稳定性及检测准确率方面有明显提高。
Abstract:
The detection of sleep apnea based on electrocardiogram(ECG) signals has great significance. However, many studies focus on improving the accuracy of detection but ignore the stability of features. We analyzed the stability of ECG features for sleep apnea detection and established the apnea event detection model.Based on the integrated stable feature selection strategy, the minimal-redundancy-maximal-relevance (mRMR) feature selection method was combined with the robust rank aggregation (RRA) method to achieve stable ranking of 45 ECG features. The 10-fold cross-validation and support vector machine (SVM) were used for feature verification and detection model establishment. Finally 14 features were used to establish the classification model, and Acc=90.03%, Se=86.71%, Sp=91.73% were achieved on the independent test set. The stability and detection accuracy of the selected features are significantly improved.

参考文献/References

备注/Memo

备注/Memo:
(收稿日期:2020-02-19)国家自然科学基金资助项目(81871444);山东省重点研发计划项目(2018GSF118133)。△通信作者Email:sswei@sdu.edu.cn
更新日期/Last Update: 2020-04-14