基于DAM_XGCN的疾病辅助诊断模型  

Disease Auxiliary Monitoring Model Based on DAM_XGCN

作　　者：陈燕[1] 龚庆悦[1] 李铁军 王红云 鲍剑洋[1] 胡孔法[1] CHEN Yan;GONG Qing-yue;LI Tie-jun;WANG Hong-yun;BAO Jian-yang;HU Kong-fa(College of Artificial Intelligence and Information Technology,Nanjing University of Chinese Medicine,Nanjing 210023,China;The Second Affiliated Hospital of Nanjing University of Chinese Medicine(Jiangsu Second Hospital of Traditional Chinese Medicine),Nanjing 210017,China;College of Nursing,Nanjing University of Chinese Medicine,Nanjing 210023,China)

机构地区：[1]南京中医药大学人工智能与信息技术学院,江苏南京210023 [2]南京中医药大学第二附属医院(江苏省第二中医院),江苏南京210017 [3]南京中医药大学护理学院,江苏南京210023

出　　处：《软件导刊》2021年第8期70-74,共5页Software Guide

摘　　要：为挖掘患者治疗中诊断疾病指标变化特点,提出一种疾病辅助诊断模型(DAM_XGCN)。首先使用XGBoost预处理中西医指标数据,将指标诊断重要性排序结果可视化;其次将高等级指标及患者就诊时序信息输入到图卷积神经网络中进行训练。以三甲医院名医亲自记载近17年的临床医案为基础,得到一/二/三维乙肝小、大三阳指标组合的交/差集结果,该模型预测结果准确率达72%,比传统方法提高10%以上。DAM_XGCN模型不仅可有效学习中西医诊断指标间的复杂关联信息,辅助医生临床诊断,而且能追踪模型中各环节输出结果。In order to explore the characteristics of the index changes in the diagnosis of diseases in the treatment of patients,a disease-assisted diagnosis model(DAM_XGCN)is proposed.First,XGBoost is used to preprocess the index data of Chinese and Western medicine to visualize the importance of index diagnosis;secondly,the high-level index and the time sequence information of patient’visit are input into GCNs for training.Based on the clinical records of the famous doctors in the top three hospitals personally recording nearly 17 years of clinical medical records,the results of the intersection/difference set of one/two/three-dimensional hepatitis B small and large three positive indicators are obtained.The accuracy of the model is 72%,which is more than 10%higher than the traditional method.The DAM_XGCN model not only effectively learns the complex correlation information of Chinese and Western medicine diagnostic indicators to assist doctors in clinical diagnosis,but also tracks the output results of each link in the model.

关 键 词：图卷积网络 XGBoost 中西医诊断 乙肝 智慧医疗 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]