基于深度学习的医院海量档案特征快速查询算法研究  

Based on the Deep Learning Hospital Mass File Feature Quick Query Algorithm Research

作　　者：尹声声 YIN Sheng-sheng(The Second Affiliated Hospital of Shantou University Medical College,Shantou 515041 China)

机构地区：[1]汕头大学医学院第二附属医院,广东汕头515041

出　　处：《自动化技术与应用》2024年第5期114-117,共4页Techniques of Automation and Applications

基　　金：汕头市汕头大学医学院第二附属医院科技项目(220516166493276)。

摘　　要：为提高档案特征查询结果的查全率和查准率,提出基于深度学习的医院海量档案特征快速查询算法。通过深度学习,建立深度卷积神经网络,采用文本聚类算法,划分档案文本类别,利用图搜图的方式,通过神经网络,提取指定档案的图像特征,得到图像内容对应的文本注释,搜索该文本归属簇集,将簇集内所有文本的特征词和主题作为文本特征,将文本数据匹配到的图像作为图形特征。选取医学领域公共数据集作为实验数据,实验结果表明,针对海量档案文本特征和图像特征,设计算法提高了特征查询查全率和查准率,档案特征快速查询精度更优。In order to improve the recall and precision of archives feature query results,a fast query algorithm of massive archives features in hospital based on deep learning is proposed.Through deep learning,a deep convolution neural network is established,and the text clustering algorithm is used to divide the file text categories.By using the way of graph search,the image features of the specified file are extracted through the neural network,the text annotations corresponding to the image content are obtained,the text belongs to the cluster,and the characteristic words and topics of all texts in the cluster are taken as the text features.The image to which the text data is matched is taken as the graphic feature.The public data set in the medical field is selected as the experimental data.The experimental results show that for the massive file text features and image features,the design algorithm improves the recall and precision of feature query,and the fast query accuracy of file features is better.

关 键 词：深度学习 档案特征 查询算法 神经网络 文本聚类 特征词 

分 类 号：TP391.03[自动化与计算机技术—计算机应用技术]