Although 12-lead electrocardiograms(ECGs)provide a wide range of spatiotemporal characteristics,interpreting them for arrhythmia detection is difficult due to a lack of reliable large-scale clinical datasets.Herein,we...Although 12-lead electrocardiograms(ECGs)provide a wide range of spatiotemporal characteristics,interpreting them for arrhythmia detection is difficult due to a lack of reliable large-scale clinical datasets.Herein,we proposed an innovative lightweight computerized ECG interpretation approach based on 12-lead data.Our model was trained,validated,and tested on 53845 standard 12-lead ECG records collected at Shanghai First People’s Hospital in affiliation with Shanghai Jiao Tong University.The experiments revealed that our approach had a classification accuracy of 94.41%in the classification task of seven types of rhythms,which was markedly superior to related single-lead and 12-lead ECG classification methods.Moreover,the average receiver operating characteristic area under the curve reached a value of 0.940,and the precision values for sinus tachycardia and sinus bradycardia were 0.945 and 0.91,respectively,with specificity values of 0.996 and 0.994.By employing our boosting method,we were able to improve the accuracy to 94.85%.To investigate the performance degradation of the proposed neural network in some classes,an ECG cardiologist was enlisted to review questionable ECGs;this process provides a promising direction for network performance improvement.Therefore,the proposed computerized ECG interpretation approach has practical significance because it could help professional physicians analyze patients’heart conditions based on real-time 12-lead ECG or grade their disease severity in advance.展开更多
The Lyapunov exponents of synchronous 12-lead ECG signals have been investigated for the first time using a multi-sensor (electrode) technique. The results show that the Lyapunov exponents computed from different loca...The Lyapunov exponents of synchronous 12-lead ECG signals have been investigated for the first time using a multi-sensor (electrode) technique. The results show that the Lyapunov exponents computed from different locations on the body surface are not the same, but have a distribution characteristic for the ECG signals recorded from coronary artery disease (CAD) patients with sinus rhythms and for signals from healthy older people. The maximum Lyapunov exponent L1 of all signals is positive. While all the others are negative, so the ECG signal has chaotic characteristics. With the same leads, L1 of CAD patients is less than that of healthy people, so the CAD patients and healthy people can be classified by L1, L1 therefore has potential values in the diagnosis of heart disease.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
基金supported by Shanghai Municipal Science and Technology Major Project(Grant No.2021SHZDZX0102)the National Key Technology R&D Program of China(Grant No.SQ2018YFB130700)。
文摘Although 12-lead electrocardiograms(ECGs)provide a wide range of spatiotemporal characteristics,interpreting them for arrhythmia detection is difficult due to a lack of reliable large-scale clinical datasets.Herein,we proposed an innovative lightweight computerized ECG interpretation approach based on 12-lead data.Our model was trained,validated,and tested on 53845 standard 12-lead ECG records collected at Shanghai First People’s Hospital in affiliation with Shanghai Jiao Tong University.The experiments revealed that our approach had a classification accuracy of 94.41%in the classification task of seven types of rhythms,which was markedly superior to related single-lead and 12-lead ECG classification methods.Moreover,the average receiver operating characteristic area under the curve reached a value of 0.940,and the precision values for sinus tachycardia and sinus bradycardia were 0.945 and 0.91,respectively,with specificity values of 0.996 and 0.994.By employing our boosting method,we were able to improve the accuracy to 94.85%.To investigate the performance degradation of the proposed neural network in some classes,an ECG cardiologist was enlisted to review questionable ECGs;this process provides a promising direction for network performance improvement.Therefore,the proposed computerized ECG interpretation approach has practical significance because it could help professional physicians analyze patients’heart conditions based on real-time 12-lead ECG or grade their disease severity in advance.
基金This work was supported by Tsinghua University (Grant No. 0009).
文摘The Lyapunov exponents of synchronous 12-lead ECG signals have been investigated for the first time using a multi-sensor (electrode) technique. The results show that the Lyapunov exponents computed from different locations on the body surface are not the same, but have a distribution characteristic for the ECG signals recorded from coronary artery disease (CAD) patients with sinus rhythms and for signals from healthy older people. The maximum Lyapunov exponent L1 of all signals is positive. While all the others are negative, so the ECG signal has chaotic characteristics. With the same leads, L1 of CAD patients is less than that of healthy people, so the CAD patients and healthy people can be classified by L1, L1 therefore has potential values in the diagnosis of heart disease.
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.
