Electrocardiogram analysis of post-stroke elderly people using one-dimensional convolutional neural network model with gradient-weighted class activation mapping

作者：

Highlights：

• Our proposed model has achieved ~90 % prediction accuracy and 0.95 area under the Receiver Operating Characteristic curve.

• We have devoted substantial effort to model interpretation, unraveling ECG features contributing to the model's success.

• Our model is developed entirely using open-source tools to promote open, accessible medical sciences.

摘要

•Our proposed model has achieved ~90 % prediction accuracy and 0.95 area under the Receiver Operating Characteristic curve.•We have devoted substantial effort to model interpretation, unraveling ECG features contributing to the model's success.•Our model is developed entirely using open-source tools to promote open, accessible medical sciences.

论文关键词：Electrocardiogram,Stroke,Cardioembolism,Deep neural network,Convolutional neural network,GRAD-CAM

论文评审过程：Received 27 September 2021, Revised 1 May 2022, Accepted 27 June 2022, Available online 30 June 2022, Version of Record 2 July 2022.

论文官网地址：https://doi.org/10.1016/j.artmed.2022.102342