Research On Two Dimensional And Three Dimensional Medical Image Registration Algorithms In Radiation Therapy

With the development of therapy equipments and imaging technology, Radiation Therapy has entered into precise era. In this area, Image Guided Radiation Therapy (IGRT) has been becoming a hot research domain for reducing the position error and organ displacement error effectively. For IGRT, registration method must be accurate, robust, fast and automatic. 2D/3D medical image registration is one of the key techniques in IGRT for its rapid guide and accurate error verification capability. The project consists of three contents. The first content is to research on the performance difference of several registration methods; in the second part, we realize the 2D/3D registration which is joined with Image Pyramid method; the third content dedicates the building of medical image registration platform which can run many 2D/2D and 2D/3D registration algorithms in windows operation system.This paper pays attention to intensity-based registration algorithms, we studied the performance evaluation method of registration algorithm, and discussed the interrelationship of registration's four elements. Based on the registration platform that we developed, from two sides: veracity and speed, we used Integration Registration Error evaluation method to study three similarity measures'influence on registration, included Mutual Information metric, Normalized Correlation Coefficient metric and Mean Squares metric. Experiment shows that Mutual Information is the best measure in integrated performance, by joining with image pyramid algorithm, its disadvantage about easy to get into local extremum can be solved in some situations.After deep analysis of 2D/3D medical image registration theory, We used Ray-casting method to reconstruct DRRs from CT and MR volume ,and realized DRR-based and Gradient Difference-based 2D/3D registration which is joined with Image Pyramid algorithm. Experiment shows that our method is faster than the common registration algorithm, we also discussed 2D image size, image pyramid and other factor's influence on the algorithm.At last, we integrated the intensity-based 2D/2D registration and 2D/3D registration realization into the registration platform which developed as Graphic User Interface (GUI) in Windows OS. User can construct various registration algorithms by selecting four different elements under this platform, so it can extremely facilitate the comparison research between registration algorithms. In the 2D/3D registration platform, we show the patient positioning verification by GUI, which lays a foundation for development of IGRT software platform.