Study On Some Key Issues Of Online Adaptive Radiotherapy

The aim of radiotherapy is to raise the tumor local control probability by increasing dose on tumor target and drop the radiation-induced complications of normal tissue by decreasing dose on organs at risk. But, in several weeks for radiation therapy, both the position and the shape of the tumor and organs at risk vary throughout the course of treatment comparing the initial radiotherapy plan. This will result in more low-dose overflows to the tumor and high-dose spills to the adjacent normal tissues and organs, which led to less cure rate and more radiation-induced complication.Now, the positional and morphological variation of the tumor and organs at risk can be systematically monitored and characterized by the volumetric imaging equipment. The treatment plan can be modified using a systematic feedback of measurements. Thus, the dose deviation induced by the positional and morphological variation of the tumor and organs can be reduced. We call it on-line adaptive radiotherapy (ART), and the cone beam CT (CBCT) has become a very useful tool for on-line ART. In the progress of on-line ART, the patient must lay in bed. The accurate registration algorithm is needed to accomplish the contours propagation from planning CT to CBCT, because manual contouring is highly time consuming. And the mapped contours needs to be confirmed by the doctor to deliver more dose to the tumor while reducing the dose received by the surrounding organs at risk.However, the CBCT image quality is not good enough. It will not only affects the accuracy of the registration but also affects the exact recognition of the tumor and organs at risk. Thus, on one hand, we need to develop the registration algorithm based on CBCT image to improve the accuracy of the registration and accomplish the auto-propagation of contours. On the other hand, we need to improve the CBCT image quality to allow the doctor to accurately identify the tumor and organs shapes. Therefore, the objective of this thesis is to study the registration algorithm between CBCT and CT image and improve the CBCT image quality.Research achievements and contributions are summarized as follows:(1) CBCT image registration algorithm based on the image gradientFirstly, take the Demons algorithm as an example, how the mismatch occurs was explained when the intensity-based deformable registration of CT and CBCT images was used; Then we proposed a gradient-based registration algorithm by analyzing the characteristics of the CBCT image and implemented the registration algorithm on GPU through OpenCL parallel programming in order to meet the real-time requirement of on-line ART application. Finally, the algorithm was applied to the on-line ART platform, and the applications of on-line adaptive image-guided radiation therapy, including auto-propagation of contours, aperture-optimization and dose volume histogram (DVH) in the course of radiation therapy were also studied.(2) The construction of the CBCT experimental platformIn order to obtain CBCT projection images, we have done the following work:1) The real projection images were acquired by the CBCT platform in the lab; 2) Monte Carlo method was employed to simulate a simplified CT system to obtain raw projection data sets.(3) Scatter correction for CBCT using rotating collimatorA CBCT scatter correction method based on the rotating collimator was proposed to reduce and control the x-ray scatter. Meanwhile, the parameters of the rotating collimator were studied tentatively, which depends on the penumbra edge of the image, the performances of the G-242 x-ray source and Thales flat panel detectors provided by the Shan Dong ShINVA company. In the end of this section, we analyzed the noise characters of the CBCT image corrected by the rotating collimator.(4) CBCT image deniosing based on fast nonlocal mean algorithmThe noise in the CBCT image corrected by the rotating collimator algorithm was studied, including:1) Two denoising types-one acting on the projection image and the other acting on the reconstruction image, were compared and proved their each excellences and flaws; 2) A fast non-local mean CBCT image denoising algorithm was proposed to satisfy the real-time’s demands of clinical application, and the validity of the algorithm was evaluated through phantom image.