STUDY OF KNOWLEDGE ACQUISITION FROM PICTURES BY INTEGRATION OF CASE DATA,ALGORITHMS AND SKETCHES

(A)使用自动生成算法我们扩展了我们已经开发的图像处理专家系统IMPRESS，以便可以考虑“负样本”（我们不想提取的样本图形）。通过这一点，我们展示了知识获取的可能性，其形式为<通过程序@92@92@92>从 * 等灰色图片中提取的内容>。(B)草图的自动生成从三维（3D）图像中提取组件图形的程序被开发并应用于胸部X线CT图像。提取血管、支气管、肋骨和肋骨架，以生成肺中3D结构的草图。(C)虚拟化和经验的情况下，样品作为一个新的工具，体验三维的情况下，如三维CT图像的人体，我们开发的程序来执行分割和重建的三维图像的器官，我们要检查在计算机上。通过在生成的虚拟人体中自由导航，我们可以了解每个患者身体的状态。将此思想应用于虚拟支气管镜的实现中，并在临床试验中取得了良好的评价。

(A) Use of automated generation of algorithms We extended the image processing expert system IMPRESS which we already developed so that "negative samples" (sample figures which we do not want to extract) can be considered. By this we showed the possibility of knowledge acquisition in the form of <things that are extracted from a gray pictures like **** by the procedure @92@92@92>.(B) Automated generation of sketches Procedures which extract components figures from a three dimensional (3D) pictures were developed and applied to chest X-ray CT image. Extraction of blood vessels, bronchus, ribs and rib cages were extracted to generate a sketch of 3D structures in lung.(C) Virtualization and experience of case samples As a new tool of experiencing 3D case examples like 3D CT images of human body, we developed procedures to perform segmentation and reconstruct a 3D images of organs we want to examine in computer. By navigating freely through the generated virtualized human body, we can understand the state of each patients body. This idea was applied to realize the virtualized bronchoscopy which achieved good evaluation in clinical test.