Automatic quantitative stroke severity assessment based on Chinese clinical named entity recognition with domain-adaptive pre-trained large language model

Zhanzhong Gu, Xiangjian He, Ping Yu, Wenjing Jia, Xiguang Yang, Gang Peng, Penghui Hu, Shiyan Chen, Hongjie Chen, Yiguang Lin

Abstract: Highlights•An automatic method quantitatively assesses stroke severity from clinical notes.•Domain-adaptive pre-trained large language model boosts named entity recognition.•An end-to-end pipeline enables fast and reliable objective scoring of stroke severity.•A densely annotated Chinese stroke dataset supports accurate clinical entity mining.•A novel entity-to-NIHSS mapping supports quantitative stroke research.