Development of intelligent computer aided diagnosis system of medical images based on navigation of virtual human body

本研究的目的是开发一种基于虚拟人体导航的医学图像智能导航诊断系统。我们讨论了一种全新的系统，通过自动检测异常区域的位置和状态，并将其叠加在虚拟人体的内部视图上，使医生能够有效地诊断医学图像。(1)基于高维医学图像的实时导航方法开发该系统的输入图像为三维或四维图像，其中包含精确的人体信息。研究了虚拟人体内部的快速导航方法。(2)解剖知识库的构建。探讨了胸腹器官解剖知识的存储方法。本文研究了解剖学知识在计算机上的表示方法。以解剖学知识库为例，建立了支气管名称数据库和器官图谱。(3)发展了更多的器官识别方法利用上述构建的解剖学知识库，我们开发了一种多器官同时识别的方法。(4)高维医学图像识别和理解方法的发展。利用解剖学知识库，研究了异常区域的检测、分类和比较算法。以多模态图像集成为例，研究了PET-CT图像的处理方法。(5)诊断辅助信息实时显示算法的开发。讨论了异常区域检测结果、解剖结构识别结果等辅助信息的实时显示算法。在这里，辅助信息不仅包括异常区域的标记结果，还包括与异常区域相关的解剖名称。(6)系统在临床领域的实际评估我们通过整合我们已经开发的方法，创建了一个导航诊断的原型系统。临床医生对该系统进行了评估。少

Purpose of this research was to develop an intelligent navigation based diagnosis system of medical images based on navigation of virtual human body. We discussed a completely novel system which enables a medical doctor to diagnose medical images efficiently by automatically detecting the locations and status of abnormal regions and by superimposing them on internal views of virtualized human body.(1) Development of real-time navigation method based on high dimensional medical imagesInput image of this system is three or four dimensional image which includes precise human body information. We investigated fast navigation methods inside virtualized human body(2) Construction of anatomical knowledge databaseStorage methods of anatomical knowledge of chest and abdominal organs were discussed. Here, we investigated representing methods of anatomical knowledge on computer. As examples of anatomical knowledge database, bronchus name database and organ atlases were constructed.(3) Development … More of recognition methods of organsUsing anatomical knowledge database constructed above, we developed a method for recognizing multi organs simultaneously.(4) Development of methods for recognizing and understanding high dimensional medical imagesUsing the anatomical knowledge database, algorithms for detecting, classifying and comparing abnormal regions were examined. As an example of integration of multi modal images, image processing methods for PET-CT images were investigated.(5) Development of real-time displaying algorithm of assistance information of diagnosisWe discussed algorithms for displaying assistance information such as results of abnormal region detection and anatomical structure recognition in real time. Here, assistance information includes not only marking results of abnormal regions but also anatomical names concerning with the regions.(6) Practical evaluation of the system in the clinical fieldWe created a prototype system of navigation diagnosis by integrating the methods we have developed. The system was evaluated by clinicians. Less