摘要
随着电子病历数据开放共享的需求越来越大,电子病历去隐私性问题亟需解决。利用自然语言处理技术,提出一种基于Bert-BiLSTM-CRF模型的电子病历隐私信息识别方法。采用某三甲中医院的电子病历作为数据来源,结合当前公开的数据集进行训练,得到正确率为94.02%、召回率为94.25%、F1为93.98%的中医电子病历隐私信息识别模型。与其他传统模型进行对比实验表明,Bert-BiLSTM-CRF模型能有效识别并保护电子病历中的隐私数据,有助于医疗数据的开放共享。
With the increasing demand for open sharing of electronic medical record data,the privacy of electronic medical record needs to be solved urgently.Using natural language processing technology,a privacy information recognition method of electronic medical record based on Bert-BiLSTM-CRF model is proposed.Using the electronic medical record of a three-tier traditional Chinese medicine hospital as the data source,combined with the current public data set for training,we get the privacy information recognition model of traditional Chinese medicine electronic medical record with the accuracy rate of 94.02%,the recall rate of 94.25%and F1 of 93.98%.Compared with other traditional models,the experiment shows that Bert-BiLSTM-CRF model can effectively identify and protect the private data in EMR,which is conducive to the open sharing of medical data.
作者
陈逸墨
叶辉
易珺
周华文
方丹丹
曹东
CHEN Yimo;YE Hui;YI jun;ZHOU Huawen;FANG Dandan;CAO Dong(School of Medical Information Engineering,Guangzhou University of Chinese Medicine,Guangzhou 510006,China;College of Medical Information Engineering,Guangdong Pharmaceutical University,Guangzhou 510006,China)
出处
《自动化与信息工程》
2022年第2期35-40,共6页
Automation & Information Engineering
基金
国家重点研发计划资助(2019YFC1710400):广东省普通高校重点领域专项(2020ZDZX3080)。
