多特征融合的远程会诊智能分诊研究  

Intelligent Triage of Teleconsultation Based on Multi-Feature Fusion

作　　者：翟运开 李俊珂 李金林[4] 乔岩 ZHAI Yunkai;LI Junke;LI Jinlin;QIAO Yan(School of Management,Zhengzhou University,Zhengzhou 450001,China;National Engineering Laboratory for Internet Medical Systems and Applications,Zhengzhou 450052,China;Henan International Joint Laboratory for Intelligent Healthcare Information Systems,Zhengzhou 450001,China;School of Management and Economics,Beijing Institute of Technology,Beijing 100081,China)

机构地区：[1]郑州大学管理学院,郑州450001 [2]互联网医疗系统与应用国家工程实验室,郑州450052 [3]河南省智能健康信息系统国际联合实验室,郑州450001 [4]北京理工大学管理学院,北京100081

出　　处：《计算机工程与应用》2025年第3期326-335,共10页Computer Engineering and Applications

基　　金：国家自然科学基金(72202217,71972012);河南省高等学校哲学社会科学基础研究基金(2022-JCZD-21);河南省科技攻关项目(242102210023)。

摘　　要：为解决调研中发现的现阶段大多数远程会诊平台仍依赖于人工分诊的问题,探索提出一种融合多特征的BTB智能分诊模型。假设数据中所安排的会诊科室为实际最优会诊科室,选取患者年龄、性别、基层医生初步诊断作为分诊依据。针对远程会诊申请信息复杂,重点分布不均的特点,使用BERT模型训练词向量,在此基础上,利用文本卷积神经网络(TextCNN)和双层双向门控循环神经网络(double-layer BiGRU)捕捉远程会诊申请的局部特征信息和上下文信息。将BERT特征与TextCNN特征、双层BiGRU特征相融合,最后输入到全连接层中,为基层医生推荐多个合适的会诊科室。模型在国家远程医疗中心大数据上进行多次训练和测试,当推荐科室为5时,模型的准确率可达90.77%,MRR达到76.45%。对比实验表明,提出的BTB智能分诊模型的性能比传统的机器学习模型更优,可为远程医疗平台实现智能化提供理论支撑和实践指导。A multi-feature BTB intelligent triage model is proposed to address the problem that most teleconsultation platforms still rely on manual triage.This model selects the actual optimal consultation departments from the data,and uses age,gender,and initial diagnosis of primary doctors as the basis for triage.In response to the complexity and uneven distribution of teleconsultation application information,the BERT model is used to train word vectors.On this basis,text convolutional neural network(TextCNN)and double-layer bidirectional gated recurrent unit(BiGRU)are utilized to capture local feature information and context information of teleconsultation application.Additionally,the BERT features are combined with TextCNN features and double-layer BiGRU features.Finally,these features are fed into the fully connected layer to recommend multiple suitable consultation departments for primary care doctors.The proposed BTB intelligent triage model is trained and tested on a large data set from the National Telemedicine Center of China,and when the recommended department is 5,the model can reach 90.77%in accuracy and 76.45%in MRR.Comparative experiments demon-strate that the proposed model outperforms traditional machine learning models and can serve as a useful theoretical reference and practical guide to enhance the intelligence of telemedicine platforms.

关 键 词：智能分诊 多特征融合 远程会诊 神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]