The Investigation On Patient Set-Up And Organs Motion Analysis In Radiotherapy

During the course of radiotherapy, organs motion may be caused due to the breath of lungs and there are patient set-up errors before therapy, which brings excursion of target region and has side effects on treatment. So there is need to investigate on the patient set-up and organs motion in order to improve the quality of radiotherapy.In this dissertation, we mainly use Digitally Reconstructed Radiograph (DRR) and Portal Image, combined with medical image match and registration, to investigate on the patient set-up and organs motion. The generation of DRR, the choice of similarity operator and mutual information based registration are also discussed.Firstly, we introduced the theory and generation of DRR with the algorithm of Ray Casting, through which we produce DRRs of different irradiation energy, projection orientation and apparatus (belly and head). Because the DRR generation is time-consuming, some improvements are proposed for acceleration, such as choice of interpolation and pretreatment of 3D data. Additionally we discuss the resolution and interaction of DRR. Also MIP and the related method are discussed.Next, the theory of medical image registration is introduced, and corresponding patient set-up errors can be got through the match of the template image of target region and Portal Film gathered before radiotherapy. The correlation ratio, correlation ratio combined with gradient and correlation coefficient are implemented for experiment and comparison. Additionally some improvement is utilized on the computation procedure to accelerate the process. The analysis is done on the real 3D patient set-up errors and the aspect needing advanced research is put forward.Finally, a verification image of therapy region is gathered before radiotherapy, and is implemented to register with real time portal images gathered during the treatment, so it can be used to inspect on the treatment region excursion caused by organs motion. Mutual information is commonly used in medical image registration. Usually the result and speed of mutual information based registration is related to grey-levels, resolution and interpolation, so we do some experiments and discussions on the aspect mentioned above and give some advice.