基于内容的医学图像检索方法综述

A Survey of Content-Based Medical Image Retrieval Methods

随着医学影像设备产生的数字图像日益增加,如何从大规模海量医学图像中有效检索出相关图像成为当前图像处理与分析领域研究的热点。为了给相关研究者提供有益参考,有必要对近年来医学图像检索领域提出的方法进行全面的综述。通过查阅国内外相关研究资料,发现医学图像检索的研究点主要包括特征提取和特征匹配2个方面。在特征提取方面,主要概述了传统特征提取方法和新近提出的基于深度学习的特征表示方法;而对于特征匹配方面,则通过对词汇树、哈希等方法的概述进行介绍。最后提出,特征的高效表示法和检索的实时性是医学图像检索领域未来研究的主要趋势。

Medical image retrieval is important in clinical application.With the increasing number of digital images,how to effectively retrieve relevant images from large-scale images has become a hot topic in medical image analysis and image processing.In order to provide useful references for relevant researchers,it is necessary to give a comprehensive review of the methods proposed in medical image retrieval in recent years.Based on the relevant domestic and foreign research data,the research points of medical image retrieval mainly include feature extraction and feature matching.In terms of feature extraction,we have summarized the traditional features and learned features.For feature matching,it is introduced by an overview of the vocabulary tree,hashing,etc.Finally,the future directions in medical image retrieval focus on the effective feature representation method and real-time retrieval.