Research On Key Technology Of Image Processing For Heavy-ion 3D Conformal Radiotherapy

As an effective treatment measure to locally control and kill tumor, radiotherapy together with operationtherapy and chemotherapy are considered as most powerful remedy for tumor. Compared with conventional beam radiotherapy withχ,γand photons etc., heavy-ion radiotherapy introduces less damage on healthy tissues, more destruction on tumor, more accurate positioning on tumor volume and more precise controlling radiation dose. Therefore, heavy-ion has become most advanced and effective radiation beam for radiotherapy.3D conformal radiation treatment planning system(RTPS) based on the medical image processing is the key connection between the radiotherapy hardware device and clinical treatment. It is necessary for radiotherapist to make conformal treatment plan on the system. The dissertation focused on the key medical image processing algorithm for heavy-ion RTPS and development of the RTPS.In order to improve the registration speed of breast dynamic contrast enhancement magnetic resonance images(DCE MRI), the dissertation presents a registration model for breast DCE MRI using fast Demons non-rigid registration algorithm with intensity correction. Original Demons is based on intensity change to get deformation parameters and unsuitable for breast DCE MRI. Intensity correction between pre and post contrast images based on polynomial is suggested to overcome the problem according to the signal enhancement of the breast model. The presented approach provides a novel method for the registration of DCE MRI.Based on the method of Fuzzy inference and radial basis function neural networks (RBFNN), the dissertation puts forward a new method for multimodal medical image fusion. Medical images are inherently ambiguous and non-uniform, especially for tumor,the border of which is usually fuzzy because of the soakage from the tumor to healthy tissue. The superiority of intelligent algorithm of fuzzy inference and RBFNN is integrated to perform auto-adaptive image fusion. Global genetic algorithm (GA) is employed to train the networks. Experimental results show that the proposed approach is more superiorer for fusion of multimodal medical images, especially for blurry source images.The paper improves interactive Live-Wire segmentation algorithm in two aspects: heap sort is used for searching the globally optimal path from the start node to the goal node, and the time complexity of the algorithm can be reduced from O[n2] to O[nlog2 n] by the algorithm, the restriction condition that the search stops immediately with the find of the destination node,is set up to greatly reduce searched nodes, thus the time complexity is reduced to less than O[nlog2 n]. Algorithm analysis and experiments indicate that the presented search strategy can evidently improve the speed of Live-Wire algorithm. A fast volume rendering algorithm based on Ray Casting is presented in the dissertation. Firstly, the large medical volume data is divided into equal-size blocks, and then a set of visibility tests such as empty block space skipping, early block termination and early ray termination are used to speed up the whole rendering process. Finally, volume rendering pre-integration is utilized to improve the performance of volume rendering. The experiment results show that the proposed fast volume rendering has picked up the speed of rendering for a large medical volume data without loss of image quality.The dissertation applies parallel projection theory to realize the coordinate transformation from 2-D to 3-D, and proposes a 3-D arbitrary superimposed rotation algorithm which uses one measurement extreme point as origin to pick another one. 3-D precise measurement is realized in heavy-ion RTPS. The algorithm is accurate enough to reach the requirement of 3D conformal radiation therapy.By analyzing the model of heavy-ion dose calculation, based on the transformational relation between CT value and water-equivalent path length, this work designs and realizes the calculation and visualization of heavy dose distribution for 3D superimposed conformal irradiation and dose evaluation on CT image.By conducting how respiratory motion affects the dosimetric distribution on target and critical tissues during heavy ion radiotherapy, the work illuminates that because of the high RBE and positioning accuracy of heavy-ion beam, it is more necessary for heavy ion radiotherapy than traditional radiotherapy to take respiratory control measures to reduce the affect on exact radioation therapy.Based on the systemical requirement analysis of heavy-ion radiotherapy TPS in Institute of Modern Physics, overall design of the TPS and detailed design is conducted in the dissertation. And heavy-ion TPS based on medical image processing has developed to provide a platform for heavy ion radiotherapy clinical trials. Radiotherapist can import multimodal medical images to the system to design conveniently a clinical treatment plan for patient.