人脑图谱中图象三维重建方法的研究与实现

【作者】 万里洲；

【导师】 王博亮；

【作者基本信息】 厦门大学 ， 计算机应用， 2001， 硕士

【摘要】 利用人体组织的CT、MRI或生理切片扫描图像对人体组织进行重建，在医学中拥有非常广阔的应用前景。本文主要对基于人脑图谱中利用组织切片进行三维重建的方法进行了初步研究。 三维重建方法一般可分为表面重建和体重建，两类算法具有各自不同的特点。前者更为简单直接，后者则提供了所观察组织更多的信息，相应的，需要更多的数据保存和计算，比较适合于绘制比较复杂的场景。 本课题的目标是在PC机和Windows操作系统平台上设计并实现了一个简单的可视化实验系统，为用户提供了一个具有实时性的交互的浏览人体组织的工具。课题进行了以下的工作： ●研究了采用直接体素绘制的方法，实现最简单的生长法进行图形的绘制。它需要处理的数据量十分巨大，是非常耗时的解决方案。由于PC机自身性能的限制，无法获得比较好的用户交互性，不能达到期望的目标。 ●采用VRML语言实现了利用几何重构来绘制组织图象。借助于VRML浏览器，可以做到远程的浏览。 ●采用Visual C++和OpenGL实现了一个实验系统，其中用体数据来构造等值面进行投影绘制，重建人体组织的图像。通过扫描输入的人体组织的切片图象，获得组织的二维信息，再将这些二维数据经过分类和转换，把它们转换成对应的三维体数据，在数据库中保存：然后采用ContourConnection算法进行几何重构，用三角面进行等值面拟合，实现基于等值面的体数据的三维图像绘制。 ●对三维重建中几个常用的插值算法进行比较研究，并在实验系统中进行运用：采用线性插值恢复生理切片断层间丢失的信息数据；采用二次贝塞尔方法拟合，来调整物体表面轮廓，用Gouraud算法来处理一些特殊的表面的法向量，获得较好的重建效果。实验系统证明这种基于等值面的体数据的重建绘制方法计算量较小，在性能并不高的微机上面也能实现体数据的浏览和绘制，并且交互性比较良好，具有一定的实用性；但是这种方法在编程上具有相当的复杂性，使得对具有复杂形体的组织的绘制具有一定的难度，这也是该方法的主要局限性。更多还原

【Abstract】 It is very popular to apply the CF. MRI or photograph of tissue slices to reconstruct them in medical practice now. We do some research elementarily in 3D reconstruction algorithms based on slices of human tissues in this article.3D reconstruction algorithms can be broadly classified into two categories. The first one is based on surface of physiological organizations and the other is base on cell, and each one has its own specialty. The former is more simple and direct, while the latter can offer more information for the tissues with the compensation of more data saving and computation.The object of this article is to design a simple visual reconstruct experimental system based volume data on PC and Windows OS platform, which should supply the users with an inter-communicating tool to watch the physical tissues. We have done the following jobs:We have tried to render the tissue directly by cell. But the quantity of datum is large, and the method is quite time wasting. With the limit of PC’s performance, we cannot reach our object, intercommumon.We implemented geometric construct by VRML. By dint of VRML browser, we can view tissues from remote location.We design and implement a experimental system by Visual C++ and OpenGL. Within it, we employed way to construct equivalence planes using volume data and project them to render tissues. We scan photographs of tissue slices to get 2D data and then these data are classified and transfer into 3D data which are saved into database. We do geometric construction by Contour Connection algorithm and render the tissue using Volume Data based on surface.Some algorithms used widely are compared in this article and implemented in the experimental system: simulating lost datum by linear interpolation, adjusting surface profile by twice Bezier interpolation, dealing with some special normalized vector.Experimental system proved that this method is quite easy, so it can be run well on a normal PC to browse and show volume data, and gain goodintercommunion. and practicability. But programming in this way is hard, it is quite difficult to draw a complicated tissue, this is its limitation.更多还原