A novel convolutional neural network for reconstructing surface electrocardiograms from intracardiac electrograms and vice versa
作者：
Highlights：
• Proposed a generalized formulation on the problem of multivariate multiple nonlinear regression for neural networks.
• Developed a novel CNN architecture to perform multivariate multiple nonlinear regression with only one model.
• Provided state of the art results for 12-lead ECG reconstruction from a set of EGM leads.
• Allowed the reverse mapping of 12-lead ECG to reconstructed EGM.
• Created an interpretable classifier for ECG arrhythmia classification from the weights learned in the CNN model.
摘要
•Proposed a generalized formulation on the problem of multivariate multiple nonlinear regression for neural networks.•Developed a novel CNN architecture to perform multivariate multiple nonlinear regression with only one model.•Provided state of the art results for 12-lead ECG reconstruction from a set of EGM leads.•Allowed the reverse mapping of 12-lead ECG to reconstructed EGM.•Created an interpretable classifier for ECG arrhythmia classification from the weights learned in the CNN model.
论文关键词：ECG reconstruction,Intracardiac electrogram,Implantable devices,Nonlinear regression,Convolutional multivariate multiple regression,Deep neural network
论文评审过程：Received 11 November 2020, Revised 7 July 2021, Accepted 11 July 2021, Available online 16 July 2021, Version of Record 29 July 2021.
论文官网地址：https://doi.org/10.1016/j.artmed.2021.102135
