A method for the early prediction of chronic diseases based on short sequential medical data

作者：

Highlights：

• We outline the real-world short sequential medical data and the importance of using such data for early NCD prediction.

• We propose a novel early NCD prediction method SSEPM specifically designed to process short sequential medical data.

• We propose a novel network to extract information from multilabel, temporal and nontemporal aspects for NCD prediction.

• We propose a clinical NCD knowledge-based data augmentation process to enrich the data quantity and quality during training.

• Evaluation results show that SSEPM outperforms the SOTA algorithms commonly used for early NCD prediction.

摘要

•We outline the real-world short sequential medical data and the importance of using such data for early NCD prediction.•We propose a novel early NCD prediction method SSEPM specifically designed to process short sequential medical data.•We propose a novel network to extract information from multilabel, temporal and nontemporal aspects for NCD prediction.•We propose a clinical NCD knowledge-based data augmentation process to enrich the data quantity and quality during training.•Evaluation results show that SSEPM outperforms the SOTA algorithms commonly used for early NCD prediction.

论文关键词：Chronic disease early prediction,Short sequential data,Temporal learning,Neural network,Data augmentation

论文评审过程：Received 9 September 2020, Revised 18 February 2022, Accepted 23 February 2022, Available online 3 March 2022, Version of Record 15 March 2022.

论文官网地址：https://doi.org/10.1016/j.artmed.2022.102262