由于卷积滤波尺寸等限制,U-net无法学习到心电(Electrocardiographic,ECG)信号的长时序关联性以及标签间的相关性。对此提出一种基于U-net-BiLSTM-CRF的心律失常多目标检测方法,可同时输出目标心拍所属类型和位置信息。使用U-net学习融...由于卷积滤波尺寸等限制,U-net无法学习到心电(Electrocardiographic,ECG)信号的长时序关联性以及标签间的相关性。对此提出一种基于U-net-BiLSTM-CRF的心律失常多目标检测方法,可同时输出目标心拍所属类型和位置信息。使用U-net学习融合特征,再将其输入到双向长短时记忆网络(Bi-directional Long Short-Term Memory,BiLSTM)中学习长时序依赖特征,最后使用条件随机场(Conditional Random Field,CRF)对标签间的关系建模,优化分类结果。依据ANSI/AAMI EC57:2012的心搏分类标准,对MIT-BIH心律失常数据库中共85609个心拍记录进行划分,在划分后数据集上的实验结果表明,该方法对心拍分类的准确率达99.11%,特异性为99.76%,灵敏度为97.21%,优于传统U-net在MIT-BIH心律失常数据库上的分类性能。展开更多
Arrhythmia is a common type of cardiovascular disease,which has become the leading cause of global deaths.Recently,the automatic 12-lead ECG diagnosis system based on numerous labelled data has attracted increasing at...Arrhythmia is a common type of cardiovascular disease,which has become the leading cause of global deaths.Recently,the automatic 12-lead ECG diagnosis system based on numerous labelled data has attracted increasing attention.However,labelling 12-lead ECG recordings is a complex and time-consuming task for clinicians.And then,the existence of data distribution differences limits the direct cross domain use of the trained model.Enlighted by subdomain adaptation methods,this paper designs a novel subdomain adaptative deep network(SADN)for excavating diagnosis knowledge from source domain datasets.Firstly,the convolutional layer,residual blocks and SE-Residual blocks are utilized for extracting meaningful deep features automatically.Additionally,the feature classifier uses these deep features for obtaining the final diagnosis predictions.Further,designing a novel loss function with local maximum mean discrepancy is utilized for restricting data distribution discrepancy from different datasets.Finally,the Clinical Outcomes in Digital ECG and 1st China Physiological Signal Challenge datasets are utilized for evaluating the superiority of SADN,which presents that SADN enhances algorithm performance on the unlabelled target domain dataset.Further,compared with the existing methods,the proposed network structure acquires better performance with a F1-macro of 89.43%and a F1-macro1 of 87.09%.Besides,among the 4 kinds of ECG abnormalities,the diagnostic effect of the SADN is better than that of cardiology residents.Thus,SADN has promising potential as an auxiliary diagnostic tool for the clinical environment.展开更多
Arrhythmia is ubiquitous worldwide and cardiologists tend to provide solutions from the recent advancements in medicine.Detecting arrhythmia from ECG signals is considered a standard approach and hence,automating this...Arrhythmia is ubiquitous worldwide and cardiologists tend to provide solutions from the recent advancements in medicine.Detecting arrhythmia from ECG signals is considered a standard approach and hence,automating this process would aid the diagnosis by providing fast,costefficient,and accurate solutions at scale.This is executed by extracting the definite properties from the individual patterns collected from Electrocardiography(ECG)signals causing arrhythmia.In this era of applied intelligence,automated detection and diagnostic solutions are widely used for their spontaneous and robust solutions.In this research,our contributions are two-fold.Firstly,the Dual-Tree Complex Wavelet Transform(DT-CWT)method is implied to overhaul shift-invariance and aids signal reconstruction to extract significant features.Next,A neural attention mechanism is implied to capture temporal patterns from the extracted features of the ECG signal to discriminate distinct classes of arrhythmia and is trained end-to-end with the finest parameters.To ensure that the model’s generalizability,a set of five traintest variants are implied.The proposed model attains the highest accuracy of 98.5%for classifying 8 variants of arrhythmia on the MIT-BIH dataset.To test the resilience of the model,the unseen(test)samples are increased by 5x and the deviation in accuracy score and MSE was 0.12%and 0.1%respectively.Further,to assess the diagnostic model performance,AUC-ROC curves are plotted.At every test level,the proposed model is capable of generalizing new samples and leverages the advantage to develop a real-world application.As a note,this research is the first attempt to provide neural attention in arrhythmia classification using MIT-BIH ECG signals data with state-of-the-art performance.展开更多
文摘由于卷积滤波尺寸等限制,U-net无法学习到心电(Electrocardiographic,ECG)信号的长时序关联性以及标签间的相关性。对此提出一种基于U-net-BiLSTM-CRF的心律失常多目标检测方法,可同时输出目标心拍所属类型和位置信息。使用U-net学习融合特征,再将其输入到双向长短时记忆网络(Bi-directional Long Short-Term Memory,BiLSTM)中学习长时序依赖特征,最后使用条件随机场(Conditional Random Field,CRF)对标签间的关系建模,优化分类结果。依据ANSI/AAMI EC57:2012的心搏分类标准,对MIT-BIH心律失常数据库中共85609个心拍记录进行划分,在划分后数据集上的实验结果表明,该方法对心拍分类的准确率达99.11%,特异性为99.76%,灵敏度为97.21%,优于传统U-net在MIT-BIH心律失常数据库上的分类性能。
基金supported by the National Key R&D Program of China(Grant No.2018YFB1307005)the State Key Laboratory of Mechanical System and Vibration(Grant No.MSVZD202103)Shanghai Municipal Science and Technology Major Project(Grant No.2021SHZDZX0102)。
文摘Arrhythmia is a common type of cardiovascular disease,which has become the leading cause of global deaths.Recently,the automatic 12-lead ECG diagnosis system based on numerous labelled data has attracted increasing attention.However,labelling 12-lead ECG recordings is a complex and time-consuming task for clinicians.And then,the existence of data distribution differences limits the direct cross domain use of the trained model.Enlighted by subdomain adaptation methods,this paper designs a novel subdomain adaptative deep network(SADN)for excavating diagnosis knowledge from source domain datasets.Firstly,the convolutional layer,residual blocks and SE-Residual blocks are utilized for extracting meaningful deep features automatically.Additionally,the feature classifier uses these deep features for obtaining the final diagnosis predictions.Further,designing a novel loss function with local maximum mean discrepancy is utilized for restricting data distribution discrepancy from different datasets.Finally,the Clinical Outcomes in Digital ECG and 1st China Physiological Signal Challenge datasets are utilized for evaluating the superiority of SADN,which presents that SADN enhances algorithm performance on the unlabelled target domain dataset.Further,compared with the existing methods,the proposed network structure acquires better performance with a F1-macro of 89.43%and a F1-macro1 of 87.09%.Besides,among the 4 kinds of ECG abnormalities,the diagnostic effect of the SADN is better than that of cardiology residents.Thus,SADN has promising potential as an auxiliary diagnostic tool for the clinical environment.
基金This research was partially supported by JNTU Hyderabad,India under Grant proceeding number:JNTUH/TEQIP-III/CRS/2019/CSE/08.The authors are grateful for the support provided by the TEQIP-III team.
文摘Arrhythmia is ubiquitous worldwide and cardiologists tend to provide solutions from the recent advancements in medicine.Detecting arrhythmia from ECG signals is considered a standard approach and hence,automating this process would aid the diagnosis by providing fast,costefficient,and accurate solutions at scale.This is executed by extracting the definite properties from the individual patterns collected from Electrocardiography(ECG)signals causing arrhythmia.In this era of applied intelligence,automated detection and diagnostic solutions are widely used for their spontaneous and robust solutions.In this research,our contributions are two-fold.Firstly,the Dual-Tree Complex Wavelet Transform(DT-CWT)method is implied to overhaul shift-invariance and aids signal reconstruction to extract significant features.Next,A neural attention mechanism is implied to capture temporal patterns from the extracted features of the ECG signal to discriminate distinct classes of arrhythmia and is trained end-to-end with the finest parameters.To ensure that the model’s generalizability,a set of five traintest variants are implied.The proposed model attains the highest accuracy of 98.5%for classifying 8 variants of arrhythmia on the MIT-BIH dataset.To test the resilience of the model,the unseen(test)samples are increased by 5x and the deviation in accuracy score and MSE was 0.12%and 0.1%respectively.Further,to assess the diagnostic model performance,AUC-ROC curves are plotted.At every test level,the proposed model is capable of generalizing new samples and leverages the advantage to develop a real-world application.As a note,this research is the first attempt to provide neural attention in arrhythmia classification using MIT-BIH ECG signals data with state-of-the-art performance.
