RelCAT：推动从非结构化电子健康记录中提取临床 inter-Entity关系
摘要
本研究介绍了 RelCAT (Relation Concept Annotation Toolkit)，一个用于对从临床叙述中提取的实体之间的关系进行分类的交互式工具、库和工作流程。基于 CogStack MedCAT 框架，RelCAT 旨在捕获分散于文本中的完整临床关系。该 toolkit 实现了最先进的机器学习模型，如 BERT 和 Llama along with proven evaluation and training methods。我们 demonstrate a dataset annotation tool (built within MedCATTrainer), model training, and evaluate our methodology on both openly available gold-standard and real-world UK National Health Service (NHS) hospital clinical datasets。我们 perform extensive experimentation and a comparative analysis of the various publicly available models with varied approaches selected for model fine-tuning。最终我们 achieve macro F1-scores of 0.977 on the gold-standard n2c2, surpassing the previous state-of-the-art performance, and achieve performance of >=0.93 F1 on our NHS gathered datasets。
引用
@article{arxiv.2501.16077,
title = {RelCAT: Advancing Extraction of Clinical Inter-Entity Relationships from Unstructured Electronic Health Records},
author = {Shubham Agarwal and Vlad Dinu and Thomas Searle and Mart Ratas and Anthony Shek and Dan F. Stein and James Teo and Richard Dobson},
journal= {arXiv preprint arXiv:2501.16077},
year = {2025}
}