Development of real-time deformation methods of 3D volumetric images and their applications to medical image processing

在本研究中，我们研究了基于应用的实时图像变形、可视化和适当的算法，以开发临床领域高分辨率三维医学图像的新处理技术。(1)变形模型的构建。由于人体内部的变形包括器官和周围软组织的变形，我们建模的变形既可以代表器官和软组织之间的相互作用，也可以代表自身的相互作用。我们构建了简单的解剖模型，并将这里开发的变形模型应用于它们。除了采用有限元模型、质量-弹簧模型等常规模型外，还采用了解剖结构的具体模型。(2)医学图像人体解剖结构识别方法的发展器官形状及其内部结构等解剖结构在三维医学图像中用强度值表示。因此，通过分析医学图像中的强度分布，我们开发了识别器官位置、形状和内部结构的方法。(3)形变模型解剖结构注册方法的发展我们讨论了一种将形变模型注册到(2)中识别的解剖结构的方法。(4)保存形变过程的数据结构我们研究了保存每个形变过程的数据结构。我们在开发过程中也考虑到了可视化过程。(5)变形过程实时可视化方法的发展我们开发了能够处理变形过程中产生的大量图像数据的实时可视化方法。(6)人体图谱的构建我们构建了腹部器官图谱（一组分割结果），利用实时图像变形模型实现了图像分割系统。

In this research, we investigated on real-time image deformation, its visualization and adequate algorithms depending on applications to develop novel processing techniques of high resolution 3-D medical images in the clinical field.(1) Construction of deformation modelWe modeled deformation that can represent mutual interaction between organs and soft tissues as well as self-interaction, since deformation inside a human body includes deformation of organs and surrounding soft tissues. We constructed simple anatomical models and applied the deformation models developed here to them. We used not only conventional models such as finite element model and mass-spring model but also specific models to anatomical structures.(2) Development of recognition methods of human anatomical structures from medical imagesAnatomical structures such as organ shapes and their internal structures are represented by intensity values in 3D medical images. Therefore, by analyzing intensity profiles in medical images, we developed methods for recognizing organs' positions, shapes and internal structures.(3) Development of a method for registering deformation model anatomical structuresWe discussed on a method for registering deformation model to anatomical structures recognized in (2).(4) Data structure for holding deformation processWe investigated on data structure that holds each deformation process. Also we took visualization process into account in the development.(5) Development of real-time visualization method of deformation processWe developed real-time visualization methods that can treat tremendous amount of image data produced during deformation process.(6) Construction of human atlasWe constructed abdominal organ atlas (a set of segmentation results) to achieve an image segmentation system utilizing real-time image deformation model.