A novel method for clinical risk prediction with low-quality data
作者：
Highlights：
• A novel framework is designed to resolve four low quality issues of medical real-world data for clinical risk prediction, though the framework is not limited to medical domain.
• A strategy is proposed to map each sample to an informative embedding space with fixed dimensions.
• Two techniques from natural language processing and computer vision are employed to enhance the model performance significantly.
摘要
•A novel framework is designed to resolve four low quality issues of medical real-world data for clinical risk prediction, though the framework is not limited to medical domain.•A strategy is proposed to map each sample to an informative embedding space with fixed dimensions.•Two techniques from natural language processing and computer vision are employed to enhance the model performance significantly.
论文关键词：Clinical risk prediction,Incomplete data,Multi-instance learning,Transformer,Feature embedding
论文评审过程：Received 9 August 2020, Revised 16 February 2021, Accepted 9 March 2021, Available online 17 March 2021, Version of Record 25 March 2021.
论文官网地址：https://doi.org/10.1016/j.artmed.2021.102052
