| Summary: | 碩士 === 長庚大學 === 醫學物理暨影像科學研究所 === 96 === Image registration is a process of finding spatial transform that maps points from one image to the same points in another image. A proper image registration can bring images from different modalities or different time frames into spatial alignment. Medical image registration has many clinical and research applications. These clinical and research applications may face registration problems which might need different solutions and different algorithms. A research platform with flexible choice of parameters and algorithms may enhance the image registration quality of many clinical and research applications. The purpose of the study is to develop a platform for 3D image registration with ITK (insight toolkit) coupled with Visual C++. Users can adjust images or algorithms for their specific application. This platform not only provides three types of registration methods: Translation Transform, Multi-Resolution, and Versor rigid 3D transform, but also provides users many options of image processing functions. Besides, we assess parameters for registration optimization with this platform. The results show that the registration errors of the Translation Transform method are always less then 1 mm, and the multi-resolution approach could improve accuracy and speed. The registration errors using Versor rigid 3D transform are always less than 3.13 mm. The platform that we developed could only perform 3D rigid body image registration currently, and will be able to perform the deformable registration in the future.
|
|---|
