自心臟CT影像重建大血管之3-D模型

Abstract

在醫學影像的領域中，3-D模型重建是一項重要的應用。藉由精確之3-D模型，醫師可以得到更多的資訊，進而做出正確的診斷。本論文提出一套完整的方法，針對連續平行取樣的2-D醫像建立 3-D立體模型。首先由2-D影像上找出物體輪廓後，我們利用TOROIDAL圖建立輪廓間的表面。其中我們也加入了DELAUNAY三角化的技術。由於其共球體的特性，我們預期能夠找出更合理的表面。實驗證明DELAUNAY三角化的確能夠防止表面相交的問題並維持物體空間對稱性。另外，我們也以兩組實際CT醫學影像為實驗，並有良好的結果。

3D soilid obtained from a set of CT images has many applications. It helps physicians to do diagnosis and allow the surgeons to plan a surgery. In this thesis, we study implement a system to reconstruct the 3D solid from a stack of CT images. We first extract the boundaries of the area of interest from each CT image. We then connect the boundary by using triangle patches. In order to obtain a set of reasonable triangle patches, we construct to “toroidal graph”. The triangles are obtianed by finding the shortest path in the toroidal graph. We also take advantage of the Delaunay triangles to prevent the existance of the intersection between the triangle patches. In order to solve the one many structure, we implement an user interface system to help physicians to reconstruct the solid. Experiments show that we can find good triangulation in a reasonable computing time.