Development of intelligent image database system for medical x-ray images

在本研究中，我们试图开发一个高度智能化的医学X光图像数据库，通过整合图像理解，计算机辅助诊断，和图像检索。（1）转移过去的系统：我们已经开发了一个独特的胸部X射线图像数据库，该数据库具有操作存储在数据库中的图像草图的能力。通过这个系统，我们可以检索所需的图像，通过使用图形键上的草图和<query-by-pictorial-example>方案。由于该系统是在传统的大型计算机上实现的，我们首先将该系统转移到EWS（SUN-4）上，并增加了将诊断记录写入原始X射线图像的新功能。（2）X线图像的计算机辅助诊断：应用图像理解技术研究了肺癌和胃癌的计算机诊断。关于胃癌，我们开发了一个程序来检测可疑的小区域从双对比度胃X射线图像异常。本文介绍了一种荧光照相自动加网系统，对目前用于大规模加网的100 mm胶片进行了性能测试，并指出了该系统的局限性和今后改进的基本方针。为了定量诊断尘肺病，我们开发了一个程序，自动分类为正常和9类疾病根据国际劳工组织的标准，通过检测小圆形阴影观察到的尘肺病的X射线图像的肺领域。（3）支持软件工具：为了支持计算机辅助图像诊断系统的开发，我们对以前开发的图像处理专家系统IMPRESS进行了改进，并开始了智能图像接口的研究。对于前者，我们在系统IMPRESS中增加了目标转换功能。

In this research, we tried to develope a highly intelligent image database for medical x-ray images , by integrating image understanding computer-aided diagnosis, and image retrieval.(1)Transferring the past system : We already developed a unique database for chest x-ray images which has the ability to manipulate sketches of the images stored in the database. By this system we can retrieve desired images by using pictorial keys given on a sketch and <query-by-pictorial-example> schemes. Since this system was implemented on a conventional large scale computer, we first transferred this system to the EWS(SUN-4) , and added the new function to write down diagnostic records into an original x-ray image.(2)Computer-aided diagnosis of cancer by x-ray images : We studied comuter diagnosis of lung and stomach cancer by applying image understanding techniques. Concerning stomach cancer we developed a procedure to detect small regions which are suspected to the abnormal from a double-contrast stomach x-ray image. Automated screening system for a photofluorogram was tested its performance to diagnose a 100mm films now used for mass screening and its limitations and basic policy for future improvement were made clear. For quantitative diagnosis of pneumoconiosis, we developed a procedure to automatically classify diseases into normal and nine classes according to the criteria given by ILO by detecting small rounded opacities observed in the lung field of pneumoconiosis x-ray images.(3)Supporting software tool : In order to support development of computer-aided image diagnosis system, we improved the image processing expert system IMPRESS we had previously developed , and began to study the intelligent image interface. Concerning the former, we added a goal transformation function to the system IMPRESS.