| The complex organ structure makes human medical image unique characteristicscompared with natural images. In recent years, medical image processing algorithmresearches have made great progress. Medical image segmentation is an importantpart of image processing in clinical application. According to the difference of clinicalimaging principle, the medical images can be divided by two categories, one reflectsthe organ metabolism and function information, such as PET, SPECT, fMRI, MEGimages, the other reflects the human organ anatomy structure information, such as CT,MRI, MRA, DSA image etc. Of course, every category images has their advantagesand disadvantages. The medical image processing algorithm can reflect the patient’scondition accurately and reasonably and make treatment plan by combining withvarious modes of advantages.Image guided radiation therapy has been widely applied to clinical diagnosis andadjuvant therapy, one of the important parts is the image segmentation andregistration. This paper mainly focuses on the liver segmentation using differentscheme. The objective of this study is to improve the real-time and adaptableperformance of radiotherapy in IGRT system based on CBCT and control irradiationdose of important risking organs of liver region for thoracic and abdominal tumor. Atpresent, the liver segmentation mainly depends on the clinician’s manual sketch, butthe process is time-consuming and can’t be copied, and it is easy to lead in manualerror. In view of the current situation of manually outline liver organs, for solving theexisting problems in image guided radiation therapy system and improving theperformance of the system, we study systematically the methods of imagesegmentation, image guided radiation therapy, adaptive radiation therapy. At the sametime, we propose the corresponding solution schemes. The contents of this paper are as follows:(1)Liver segmentation method combined probabilistic atlas with level setWe can obtain the data of various liver patients in clinical, these cases are classifiedand screened according to age, gender, etc. We construct the liver probabilistic atlas,and we segment liver tissue using the method combined probabilistic atlas with levelset when patients receive treatment. Because of inherent image characteristics inCBCT, the process of it will generate the part of motion artifacts, so that it makes theimage anatomy unclear. However, the images of different modality can reflect thedifferent characteristics for the same organ. CT image can reflect anatomical structureclearly, so that we automatically segment liver using this method, then combining ofdeformable registration method based on orthogonal wavelet basis is to transfer theliver contour from the reference image (CT image) to the target image (CBCT image)in deformation domain, thus the liver contour automatically access to CBCT imagewhen patients are treated. Deformable registration method on orthogonal waveletbasis is made use of Navier partial differential equation, and the normalized mutualinformation works as similarity measure, finally the registration be succeeded whenthe inside and outside two kinds of forces reach to balance. In order to reduce thecomplexity energy function, we use three dimensional wavelet decompose whendeformation domain is recovered. According to the required precision, waveletdecomposition converge to a certain scale, finally the deformation field is representedby the wavelet coefficients.(2)Synchronous segmentation and registration method based on the region ofinterest narrow bandIn current clinical, liver segmentation mainly rely on manual delineation of doctors,and it is rigid registration between CBCT and CT image, and no relationship betweensegmentation and registration. A synchronous segmentation and registration jointmethod based on the region of interest narrow band is proposed in this paper toachieve the segmentation and registration for radiotherapy treatment planning system.The key issue in our method is to construct an ASOR synchronization model byintegrating narrow band model with active contour model to accomplish initial segmentation，and then we combined optical flow based deformable registrationmethod to optimize the process iteratively. At first，both of nonlinear diffusion modeland narrow band active contour are used to get the liver position information of CTimage to provide reasonable initial contour for synchronization model. Secondly, theliver contour and corresponding narrow band was mapped from planning CT toCBCT by affine transformation. Finally, ASOR synchronization model are used tosegment and register simultaneously in evolution process by optical flow fordetermining the active contour level set movements. The experiment results show thatwhen the proposed method is applied in CBCT based IGRT system, it canautomatically segment liver to implement the real time calculation for followingradiation therapy plan，and also the deformation field which is obtained during thesegmentation process can transfer the radiation plan from Planning CT to CBCTadaptively.(3)Adaptive radiation therapy combined with image segmentation andregistrationThis part of the paper is applied (1),(2) to clinical practice. Segmentation andregistration advantages largest apply to the image guided radiation therapy system.Automatic segmentation of liver tissue replaces the manual segmentation, and thedeformable registration replaces the rigid registration to improve the matchingprecision. At the same time, deformation domain transferred the information (organand tumor target contour line and isodose line) from CT to CBCT automatically forcontrolling the radiation dose in normal tissue. Segmentation and registration is thekey technology of IGRT system, we can improve the performance of the system if weadopt more accurate algorithm.In a word, the research contents of this paper can improve the segmentationaccuracy and efficiency of liver tissue, improve the performance of IGRT systembased on CBCT, automatically delineate liver and target quickly and update thetreatment plan when patients receive radiation treatment, finally improve the accuracyand real-time of radiotherapy. |
