摘要
为了进一步提升心肌梗死的诊断效果,提出一种基于卷积神经网络和长短时记忆网络的心肌梗死检测方法,用于准确地从心电图信号中检测心肌梗死。具体来讲,提出3种分别基于卷积神经网络、卷积神经网络结合长短时记忆网络以及它们的集成模型的检测方法,以期从心电信号中检测心肌梗死和正常搏动。此外,采用数据重采样方法,即合成少数类过采样方法和Tomek Link解决数据集不平衡问题。最终与其他方法的实验结果相比,经过数据重采样的集成卷积神经网络模型的结果取得了明显优势,证明提出方法的有效性。
A screening method for myocardial infarction based on convolutional neural network and long short-term memory network is proposed to accurately detect myocardial infarction from electrocardiogram(ECG)signals,thereby further improving the diagnostic efficacy of myocardial infarction.Based on convolutional neural network,convolution neural network combined with long short-term memory network,and their integration,3 different models are put forward to detect myocardial infarction and normal beats from ECG signals.In addition,the data resampling methods,namely synthetic minority oversampling technique and Tomek Link,are used to solve the class imbalance problem of data sets.The data resampled integrated convolutional neural network has obtained better experimental results than other methods,which proves the effectiveness of the proposed method.
作者
刘建华
吕建峰
蔡金丹
LIU Jianhua;LÜJianfeng;CAI Jindan(Department of Cardiovascular Medicine,Renhe Hospital Affiliated to China Three Gorges University,Yichang 443000,China)
出处
《中国医学物理学杂志》
CSCD
2022年第11期1448-1452,共5页
Chinese Journal of Medical Physics
基金
湖北省卫生健康科研项目(WJ2021F061)。
关键词
心肌梗死
心电图
卷积神经网络
长短时记忆网络
myocardial infarction
electrocardiogram
convolutional neural network
long short-term memory network
