Noval Methods Of Elastic Registration And Visualization For MRI-Man

In the space science and technology research, the manned space engineeringis a very important area. In this area, the life protection system for astronautsis the key project.Because of the strong radiation in the outer space, the evaluation of radiateharmness to the astronauts is necessary, and which can help to design theprotection scheme properly. Unfortunately, the dose received by astronautscannot be measured directly. The simulate calculation using 3-D human bodymodel is an appropriate approach to solve this problem.The creating of 3-D human body model is a complex and systemic work,which includes registration, segmentation, 3-D visualization, and so on. Thisthesis is devoted to the complete and penetrative research of the keytechniques applied to the 3-D human body model creating. The thoroughresearch is more concentrated on elastic registration. The 3-D contourreconstruction, point reconstruction, the arbitrary section plane display, andthe autometic segmentation of some organs are also deeply studied.The data for 3-D body reconstruction is obtained by MRI scanning, so thiswork can be called MRI-Man modeling. Due to the long time and the partlyscanning, the translation and deformation are appeared. The accurate elasticregistration is the key step for the MRI-Man modeling.For the deformation of the elastic object, the elastic mechanics can be used tosolve the problem well. Based on the elastic mechanics theory, an elasticmathematic model is studied and proposed for medical image registration.Much more accurate results are achieved by using the finite elementalgorithm.Multiquadric function is a useful reconstruction method for 3-D scattereddata. This approach is introduced into image elastic registration in this thesisfor the first time. By improving its smoothness character, the robustregistration is achieved. And a semi-automatic method is proposed forselecting the corresponding point, which simplified the points extractioneffectively.Elastic registration based on thin plate spline interpolation is deeply studied.The detailed analysis is performed in the fifth chapter and the disadvantagesof the previous methods are pointed out. Two new approaches are proposedto solve the problems. These methods improved the interpolation byincreasing the smoothness and compact support characters, which improvedthe robustness and applicability of the method. At the same time, theautomatic method for corresponding points selection is discussed.By using the above novel elastic registration method, the MRI-Man data areregistered accurately.3-D reconstruction is the important technique for visualization. To the actualproblem, the smooth contour reconstruction and smooth points reconstructionare researched and utilized. These reconstruction methods are fast, simpleand well working.More useful information of the organs can be obtained if the arbitrary sectionplane images can be showed. The section plane display algorithm is studiedin the thesis, and the interaction section and curve section display is achieved,which can provide the aids for observation and analysis of organs. Also, it isvery helpful for the dose distribution analysis.It is an important research topic to segment the regions of interesting. Aninvestigation of brain tissue automatic segmentation is made. The mathematicmorphological operators are used to conduct the task. The 3-D shape of thebrain is showed by the points rendering.