摘要
目的利用循环神经网络探索自动推断根本死因的可行性,为死因监测工作提供自动化工具。方法利用2016—2021年福建省国家级死因监测点经专家审核的死亡报告数据,基于门控循环单元(GRU)构建根本死因推断模型,通过训练、验证和测试确定最终模型;用准确率、加权查准率、加权查全率和F1分数评价模型性能。结果根本死因推断模型的验证集推断准确率达93.5%。测试集推断准确率为87.8%,加权查准率87.3%,加权查全率87.8%,加权F1分数为0.88。结论基于循环神经网络的根本死因推断模型具有较好性能,深度学习相关技术在辅助提升死因监测工作质量上能够发挥作用,降低人工审核压力。
Objective To explore the feasibility of automated inference of underlying cause of death using recurrent neural networks in order to provide an automated tool for cause of death surveillance.Methods An underlying cause of death inference model was constructed based on gated recurrent unit(GRU)using expert-vetted death report data from national cause-of-death surveillance sites in Fujian Province,2016-2021.The final model was determined through training,validation,and testing.The model performance was evaluated using accuracy,weighted precision,weighted recall and F1 score.Results The validation set of the underlying cause-of-death inference model had an inference accuracy of 93.5%.The test set had an inference accuracy of 87.8%,a weighted precision of 87.3%,a weighted recall of 87.8%,and a weighted F1score of 0.88.Conclusion The undering cause of death inference model based on recurrent neural networks showed good performance.Deep learning of related technologies can play a role in improving the quality of cause of death surveillance work and reducing the pressure of manual review.
作者
方欣
黄少芬
钟文玲
尹艳榕
陈铁晖
FANG Xin;HUANG Shaofen;ZHONG Wenling;YIN Yanrong;CHEN Tiehui(Fuzhou Center for Disease Control and Prevention,Fuzhou,Fujian 350012,China)
出处
《海峡预防医学杂志》
CAS
2023年第3期7-10,42,共5页
Strait Journal of Preventive Medicine
基金
福建省卫健委青年科研课题(2020QNB017)
关键词
根本死因
推断模型
机器学习
循环神经网络
门控循环单元
Underlying Cause of Death
Inference Model
Machine Learning
Recurrent Neural Network
Gated Recurrent Units
作者简介
第一作者:方欣,主管医师。专业:预防医学
