Study On Simulation Modeling And Application Of 3D Radiation Therapy Planning System

Radiation therapy is an important mean of oncotherapy. Three dimensional radiation therapy planning system (3D-RTPS) is one of its key subsystems through which the whole irradiation therapeutic process can be simulated. Based on the analysis and evaluation of the patient body dose distribution data computed by the system a better treatment solution can be worked out, which will reduce side effects and increase tumor control rate.Starting from the product demands, the key techniques of 3D-RTPS was studied in detail. The verification and nanalysis works were carried out for all research works. The final goal of this thesis is to provide a scalable software product platform which can meet the clinical demand.The works of 3D visualization, tissue segmentation, dose calculation, IMRT, GPU speed-up and software development were done in this paper.In the aspect of 3D visualization two creative works were done. First a speed-up method by pre-computing voxel normal vectors which were indexed and stored with spherical coordinate was put forward based on phong model. Using this method method a lot of time was saved as normal vectors recomputing were avoided. Morerover less additional memory is needed to store the voxels surface. Second the application of volume rendering based on ray casting was used in dose 3D visualization. Because the color and opacity can be classified in different way due to the clinical need, the doctor can do a visual evaluation.In the aspect of tissue segmentation the automatic detection of body contour, lung and spinal cord were achieved. To contour spinal cord automatically three key subtasks were performed. A new diagnostic model was created to detected spinal cord point according to the structural knowledge of spinal cord and its surroundings in the subtask of detecting spinal cord possible region. From the spinal cord point the spinal cord possible region was obtained by region growing. With the adaptive diagnostic model, all spinal cords were detected automatically for the sixty clinical patients CT image series. Runing the software on a notepad computer the detection time was in three seconds, which meets the clinical requirements.In the aspect of dose calculation, a method of importing couch CT image pixels was put forward based on point kernel convolution/superposion model. With this method the attenuation of X ray beam from couch was considered, which reduced the error caused by couch and increased the system precise of dose calculation. For model matching, an automatic method was suggested based on simulated annealing arithmetic. It can reduce the dependence degree of the operator technique, reduce maintenance cost and improve the product competitiveness.In the aspect of IMRT a kind of pencil beam kernel generated by point kernel was suggested for dose calculation during the intensity optimization iteration, which reduced the dose calculation time. The direct aperture optimization method can be integrated in the planning system based on point kernel convolution/superposition model with this method. From the test results of the phantom and clinical patient, it was concluded that this method can be used for the intensity optimization iteration dose calculation as the satisfied precision due to the optimization result coherence was obtained.In the aspect of GPU speed-up technique, by modifying the old model the NVIDIA GPGPU model was first used in commercial 3D radiation therapy planning system with point kernel convolution/superposition model. The MFC DLL and export class technique was also used to avoid a mass of code migrating work. The most efficient number of parallel threads was determined by analyzing the result data.Based on works mentioned above the first commercial 3D radiation therapy planning system with point kernel convolution/superposition dose model was developed and used in clinic.