疾病关联网络的构建及其对并发症预测方法的研究

目的:通过基于Bert深度学习技术处理的电子病历数据构建疾病关联网络,使用链路预测技术对患者的并发症进行预测.方法:使用Bert-WWM模型进行命名实体识别,提取来自非结构化中文电子病历中的疾病诊断信息,使用关联规则挖掘技术计算疾病关联系数,在此基础上构建疾病关联网络.最后,使用链路预测,预测患者潜在的并发症.结果:通...

The Construction of the Disease Association Network and Research on the Prediction Method of its Complications

Objective:To predict the complications of patients by data mining technology based on electronic medical records,so as to assist in the prevention and treatment of clinical patients.Methods:In this paper,the diagnosis from unstructured electronic medical records was extracted by named entity recognition,and the possible complications of patients were predicted by constructing disease association network and link prediction.Results:By mining association rules,81 important disease concurrent rules were found.By using link prediction technology to predict the disease association network,we found that common respiratory diseases are often accompanied by other respiratory diseases or endocrine system diseases.At the same time,it also predicts the concurrence of some unrelated system diseases.Finally,the predicted complications were verified by literature and pathogenesis analysis.Conclusion:The link prediction based on disease association network constructed in this study can predict clinical potential complications and provide service for effective clinical decision making.