Development Of Segmentation Software For Intelligent Radiotherapy System Guided By Mutil-model Medical Image

At present,clinically,tumor radiotherapy doctors need to manually delineate the GTV(gross target volume)according to the image information of PET(positron emission tomography),CT(computed tomography)and MRI(magnetic resonance imaging).It is not only time-consuming,but also a certain difference between the segmentation results of the same tumor medical image by different operators.Therefore,the design and development of the automatic segmentation software for GTV have important clinical application value.This paper mainly discusses the development technology,system design and function realization of the target segmentation software.The main research results are as follows:(1)According to the requirement of clinical radiotherapy and using the oriented object programming design method via the integrated development environment of VS2017,we use Qt as the interface design.We call the ITK class library to process the medical image.We call the medical image three-dimensional display VTK class library which displays the three-dimensional data of the image on three orthogonal slices to reflect the internal structure level of the tissue and organ.We use mouse to realize the interactive processing of medical images such as rotation,translation and so on.The whole software has good scalability and portability.(2)The image fusion algorithm based on spatial domain is designed to process the gray value of the source image directly.By 3D linear interpolation,the origin,spacing and direction of the fusion image are aligned with the space of the source image.Then the weight coefficient of the weighted average method is adjusted to realize the visual display of PET/MRI fusion image by VTK,which can make the GTV delineation more accurate.(3)We have designed and implemented the automatic delineation of PET and MRI target areas for tumor radiotherapy,including threshold segmentation and region growth segmentation methods.And analyze the results.In the threshold segmentation module,we have two methods for the intelligent delineation of the GTV,which set arbitrary SUV value or the maximum SUV value ratio as threshold.In the region growth segmentation module,we have three methods for the intelligent delineation of the GTV,namely,confidence region growth,adaptive region growth and two-stage region growth algorithms.