Research On Segmentation And Three-Dimensional Reconstruction Of Liver Tumor In CT Image

Liver tumor has a high mortality rate and a serious threat to human health in the world,and it mainly appears malignant tumor in clinical diagnosis.At present,it is also used to remove the tumor for clinical treatment,which is related to the accuracy of breast tumor segmentation.Therefore,how to effectively segment liver tumor has become an urgent problem,which makes a great impact on the effect of patient diagnose.In other words,how to improve the accuracy of liver tumor segmentation has been the focus of scholars' research.The study on this thesis is to segment liver tumor CT images.Firstly,the pre-processing is employed before liver tumor segmentation,which can remove the noises and irregular particulars;Secondly,liver tumor segmentation using multi-scale morphology of eliminating local minima is proposed based on morphology theory;Finally,liver tumor is reconstructed by using the moving cube,which is based on the result images of previous segmentation.So the three-dimensional reconstruction model of liver tumor is ultimately obtained.This thesis mainly focuses on the following aspects:(1)The pre-processing is necessary before segmenting image and its purpose is to remove noise and highlight contrast.According to the characteristic of low contrast and low ratio of signal to noise,a method based on area constraint and multi-scale morphology is used for liver CT image.Firstly,gradient transform is utilized to achieve geomorphology of liver image.In gradient relief,there exists different size of areas.For small areas,they are generally corrected with irregular particulars and have different level grays.For large areas,they usually correspond to interested areas;Secondly,top-hat transform is employed to mark peak values so as to distinguish the noises from other objects.Then area constraint is used to filter noises,which are smaller than area threshold;Finally,the function between gradient levels and morphological structure element is established,and viscous morphological closing operator is utilized to modify the relief of gradient image with different sizes of structuring elements.This method can eliminate noises effectively while preserves the location of object contours.(2)Some small areas of liver tumor image are removed after pre-processing,but it also exists amount irregular particulars and noise.In order to eliminating these local minima,a method using multi-scale morphology of eliminating local minima is proposed to segment liver tumor,which is purposed to remove the interference information,such as irregular details and noise.Firstly,gradient transform is used for liver tumor image;Secondly,local minima in gradient image are distinguished by the combined knowledge of statistic characteristics and morphological properties including depth and scale.After partition,the function relationship is established between multi-scale structure elements and local minima.Morphological close operation is then employed to adaptively modify the image so as to large size of structure elements filtering some minima corresponding to noises and major object being preserved;Finally,standard watershed transform is utilized to implement segmentation of liver tumor.(3)Liver tumor is reconstructed by using the moving cube,which is based on the previous segmentation method for a sequence of two-dimensional liver tumor CT images.The three-dimensional reconstruction model of liver tumor is ultimately obtained by OpenGL method.