Study On Segmentation And 3D Reconstruction Of Glioma

Due to the complexity of medical imaging technology and the high heterogeneity of glioma surface,image segmentation and 3D reconstruction of human glioma has always been one of the most challenging tasks in medical image analysis.In the process of pathological analysis of glioma,the structure of glioma is usually divided manually by doctors,which is not only time-consuming and labor-intensive,but also relies too much on the pathological experience of doctors.Traditional medical image segmentation algorithms are usually easily affected by external factors such as noise and manual intervention,so the segmentation accuracy cannot be guaranteed.Through the research of UNet++ medical image segmentation network and 3D reconstruction algorithm,its expressive power in MRI brain tumor images is explored,and the improvement of UNet++ network structure,MC algorithm optimization,and construction of brain tumor auxiliary diagnosis system are carried out respectively.research.The main work contents are as follows:(1)The application of the improved UNet++ medical image segmentation algorithm in the automatic segmentation of brain glioma is explored.On the basis of the original network structure,the decoder downsampling stage is fused across channels,and introduce a new deep supervision scheme,at this time,the improved network can fuse coarse-grained semantic and fine-grained semantic information at full scale.Segmentation experiments were carried out on 335 images in the publicly available Bra TS brain tumor segmentation dataset.2D and 3D contrastive segmentation experiments were used to comprehensively evaluate the segmentation performance of the improved network.The segmentation results were compared with the results of UNet,UNet++,and UNet3+ medical image segmentation networks.For comparison,the performance of the above four segmentation algorithms is comprehensively evaluated using the four indicators of DSC similarity coefficient,95%Hausdorff surface distance,Sensitivity,and PPV.Experiments show that the improved algorithm makes the segmentation results of glioma have more overlap with the gold standard in regions,which can better complete the segmentation of gliomas.In clinical applications,it can help neurosurgeons to efficiently separate brain tumors from surrounding tissues of the human brain,so as to achieve rapid computer diagnosis and treatment.(2)After completing the segmentation of the glioma,use the MC surface rendering method and the staggered deformation volume rendering algorithm to reconstruct the segmented image in 3D,so that the doctor can analyze the pathological structure of the brain tumor from multiple angles,using adaptive trilinear The interpolation algorithm and the asymptote method eliminate the approximate representation and connection ambiguity problems existing in the MC algorithm,making the reconstruction results more accurate.The above algorithms are implemented using C++ and Python programming languages respectively,and the MC surface drawing and staggered deformation are The reconstruction results of the volume rendering algorithm and the execution time of the algorithm are compared,and the advantages and disadvantages of the two algorithms are summarized.(3)Combining the improved segmentation algorithm and 3D reconstruction method with VTK,Simple ITK,Open GL and other development tools,using Python to build a brain tumor image segmentation and 3D reconstruction auxiliary diagnosis system on the QT platform,realizing medical image reading and preprocessing,segmentation and 3D reconstruction and other related functions.The system maintains good operability and expandability,and can help doctors play an important role in brain tumor simulation treatment plans,surgical planning and anatomical research.