Research On Medical Image Segmentation Algorithm Based On U-Net Network

With the development of medical imaging technology,the resolution of medical images has been continuously improved,and a variety of high-definition imaging methods such as CT,PET-CT,and MRI have appeared.At the same time,the segmentation of medical images also ushers in new challenges.Due to the complexity of the human organ structure,medical images have complex diversity and differences,making segmentation difficult.The traditional variational level set method has the advantages of flexible topological structure,convenient curve evolution energy allocation scheme,and simple and effective numerical solution method.U-Net networks have the advantages of simple network topology and small training set data requirements.These two methods are therefore widely used in the field of medical image segmentation.However,in the segmentation process,the variational level set is sensitive to the initial contour and the evolutionary control parameters,resulting in insufficient generalization ability,and the driving force of curve evolution can still be optimized.U-Net also has problems such as loss of edges in segmentation results,long network training time,and single application scenario.Although these two algorithms have been developed relatively well and many optimization schemes have been proposed,how to optimize the segmentation process so that the segmentation results can be obtained quickly and effectively is still a problem to be explored.Based on the classic variational level set algorithm and U-Net network,this paper studies the above problems.The specific work is as follows:(1)Aiming at the problem that U-Net network segmentation results are easy to lose edges,and the variational level set is sensitive to the initial contour position,a new model combining the advantages of the two models is proposed.U-Net network segmentation results can ensure that the initial contour of the variational level set fits the boundary as much as possible,and the variational level set algorithm based on edge information can further optimize the segmentation result.Experimental results show that the U-DRLSE model has high segmentation accuracy and the results are more robust.(2)Aiming at the single application scenario of the U-Net network,the U-Net network is applied to the FiCD(nerve fiber connection)process.Utilizing the advantages of accurate U-Net segmentation results and fast segmentation speed,it speeds up the process of comparing FiCD groups.