DeepMI: Deep multi-lead ECG fusion for identifying myocardial infarction and its occurrence-time

作者：

Highlights：

• Spectral representation of time-series data offers robustness to variability in sampling rates and device specifications

• Cross-domain transfer learning to extract spectral features, reduce spectrogram dimension and minimise redundancy

• Fusion approaches at different levels of data representation aiming to utilise distinctive characteristics of ECG leads

• Predicting the occurrence time of heart attack, utilising a large scale dataset (>15,000 patients and >323,000 samples).

摘要

•Spectral representation of time-series data offers robustness to variability in sampling rates and device specifications•Cross-domain transfer learning to extract spectral features, reduce spectrogram dimension and minimise redundancy•Fusion approaches at different levels of data representation aiming to utilise distinctive characteristics of ECG leads•Predicting the occurrence time of heart attack, utilising a large scale dataset (>15,000 patients and >323,000 samples).

论文关键词：Deep learning,Heart disease,Myocardial infarction,Onset time detection,Transfer learning

论文评审过程：Received 11 March 2021, Revised 7 July 2021, Accepted 5 October 2021, Available online 12 October 2021, Version of Record 2 November 2021.

论文官网地址：https://doi.org/10.1016/j.artmed.2021.102192