ISeeU2: Visually interpretable mortality prediction inside the ICU using deep learning and free-text medical notes

作者：

Highlights：

• Deep Learning can outperform traditional scores such as SAPS-II.

• Results suggest Deep Learning uses metadata to predict patient mortality.

• Performance is competitive with the state of the art using simpler pre-processing.

• Shapley Values offer interpretability without sacrificing predictive performance.

摘要

•Deep Learning can outperform traditional scores such as SAPS-II.•Results suggest Deep Learning uses metadata to predict patient mortality.•Performance is competitive with the state of the art using simpler pre-processing.•Shapley Values offer interpretability without sacrificing predictive performance.

论文关键词：ICU,Clinical notes,Deep learning,MIMIC-III,Shapley Value

论文评审过程：Received 29 December 2020, Revised 28 June 2021, Accepted 2 April 2022, Available online 22 April 2022, Version of Record 27 April 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.117190