摘要
为实现对前壁心肌梗死、下壁心肌梗死、前间壁心肌梗死和正常心电信号进行分类识别,本研究以临床数据库为实验数据来源,从中提取训练集和测试集数据用于训练和测试网络模型,对传统的神经网络进行优化,设计出一种新的网络算法:多分辨率残差网络。将多分辨率残差网络与传统的网络进行可视化对比分析,以评价模型的识别效果。多分辨率残差网络的测试集准确率为91.8%,高于经典的神经网络。本研究的算法能够辅助医生进行心肌梗死类疾病的诊断,有一定的临床意义。
To realize the classification of the anterior myocardial infarction,inferior myocardial infarction,anterior intercostal myocardial infarction and normal electrocardiogram signal,we used the clinical database for experimental data source,extracted training set and test set used to train and test the network model,the classic neural networks was optimized and then a new network was designed:multi-resolution residual network.The multi-resolution residual network was compared with the classical convolution network model visually to evaluate the recognition effect of the model.The test set accuracy of multi-resolution residual network was 91.8%which was higher than the classical convolutional neural networks.The algorithm of this paper can assist doctors to diagnose the myocardial infarction diseases and has certain clinical significance.
作者
齐继
蒋华
张瑞卿
沈阳
佟彦妮
沙宪政
常世杰
QI Ji;JIANG Hua;ZHANG Ruiqing;SHEN Yang;TONG Yanni;SHA Xianzheng;CHANG Shijie(Biomedical Engineering Department of China Medical University,Shenyang 110122,China;Department of Medical Equipment,The Fourth Affiliated Hospital of China Medical University,Shenyang 110032;Department of Cardiovascular Medicine,The First Affiliated Hospital of China Medical University,Shenyang 110054)
出处
《生物医学工程研究》
2019年第4期387-392,共6页
Journal Of Biomedical Engineering Research
基金
辽宁省自然科学基金资助项目(2018055003)
辽宁省教育厅科学研究一般项目(L2015563)
中国医科大学首批健康资料大数据研究课题(重点课题第9号)
关键词
心肌梗死
心电图
深度学习
卷积神经网络
残差网络
Myocardial infarction
Electrocardiogram
Deep learning
Convolutional neural network
Residual network