论文部分内容阅读
目的创建一个基于内部影像报告的高级图像检索和数据挖掘系统。方法首先,采用自然语言处理(NLP)技术对影像报告进行语义分析,并将其储存于当前最先进的搜索引擎中。然后,在图像存储与传输系统(PACS)中利用报告中的序列和影像号检索图像,并存储以供今后查看。采用基于网络前端作为查询图像、显示检索结果和报告文本的界面。以一本全面的放射学词典作为基本术语来源,搜索算法可同时搜索同义词、缩写及相关主题。结果采用不同系统设置对108份人工书写的报告进行分析。其中,全句法和语义分析效果最佳,其精确度和查全率分别为0.929和0.952。此系统自2010年10月成功运行以来,已有258824份报告被索引,405146幅预览图像存储于数据库中。结论采用数据挖掘和NLP技术能够快速访问大量的图像和影像报告资源库,具有很高的精确度和查全率。因此,该系统已成为日常临床工作、教学和研究的重要工具。
Objective To create an advanced image retrieval and data mining system based on internal image reporting. Methodology First, NLP is used to semantic analysis of video reports and stored in the state-of-the-art search engine. The images are then retrieved using sequences and image numbers from the report in the Image Storage and Transmission System (PACS) and stored for future review. An interface based on web front end to query images, display search results, and report text. With a comprehensive radiology dictionary as a basic source of terminology, search algorithms can search for synonyms, abbreviations, and related topics simultaneously. Results The reports of 108 manuscripts were analyzed using different system settings. Among them, the full syntactic and semantic analysis of the best results, the accuracy and recall rate were 0.929 and 0.952. Since the system was successfully operational in October 2010, 258,824 reports have been indexed and 405,146 preview images have been stored in the database. Conclusion Using data mining and NLP technology provides fast access to a large number of image and image reporting repositories with high accuracy and recall. Therefore, the system has become an important tool for routine clinical work, teaching and research.