Research On Lung Tumor Segmentation Based On Three-Dimensional Reconstruction

Lung cancer is a disease with high morbidity and mortality.Computed tomography(CT)scans appear to be one of the main instruments for diagnose diseases and pathological features extraction.Accurate segmentation of lung tumors based on CT images is of great significance for radiotherapy,disease diagnosis and efficient evaluation of therapeutic response,which is a hot topic in the field of computer-aided diagnosis.At present,researches on the segmentation of solitary nodules and tumors are more prevalent.However,one continuing challenge of CT images is the segmentation with large adherent interfaces between tumors and anatomical structures around lungs.This paper mainly studies the segmentation of boundary adhesion tumors at different adhesion degrees and different locations.Study the methods of lung parenchyma extraction,lung parenchyma repair and lung tumor segmentation,and then the lung and tumor are reconstructed in three dimensions.To rapidly extract the lung parenchyma sequential images from the lung sequential CT images,a lung parenchyma sequential images extraction method based on 3D region growing algorithm is studied.And then design a technical framework for the segmentation of lung tumors.First,study the segmentation method of lung with solitary tumors.To accurately and efficiently extract the lung parenchyma sequential images from the lung sequential CT images,the Hessian matrix vascular enhancement image filter method combine with the random walk algorithm that with a designed way of seed points adaptive generation are used to segment tumors with complex borderline details and adherent vessels.Hessian matrix vascular enhanced image filter is used to preprocess the lung parenchyma image to reduce the effect of lung tumors adhere to blood vessels on tumor segmentation.And the random walk algorithm which has a strong processing ability to local borderline details is used to accurately segment the lung tumors.To repair the missing lung parenchyma boundary during the segmentation of lungs with boundary adhesive tumors,a modified method based on convex hull algorithm and a 3D surface reconstruction algorithm are proposed.The convex hull algorithm is used to repair the lungs with small and smooth adhesion boundaries,which has higher repair efficiency.And this paper improved the convex hull algorithm to reduce manual interaction and has a better repair effect.The 3D surface reconstruction algorithm is designed to solves the problem that the repair effect is poor due to the lack of characteristic information of the boundary of lung parenchyma in twodimensional CT images.Firstly,surface rendering is used to reconstruct the lung parenchyma with missing boundaries in 3D,and then poisson surface reconstruction algorithm is used to repair the 3d lung parenchyma by extracting the 3d point cloud information of missing location.Finally,the tumor is segmented from the repaired lung parenchyma,three-dimensional reconstruction of lung and tumor is realized.Through the experiment comparison,the method obtained good results.