Research On CT Image Organ Segmentation Algorithm Based On Cascaded Vnet-S Network

Computed Tomography(CT)is widely used in clinical diagnosis and is an important means to obtain internal information of the human body.With the rapid development of computer technology and artificial intelligence,computer aided diagnosis(CAD)system can automatically process CT images to provide doctors with a diagnosis basis,thereby improving the efficiency of doctors.In computer-aided diagnostic systems,organ segmentation technology is essential.Organ segmentation technology is the premise for qualitative and quantitative analysis of patients' organs,and is an important auxiliary means for doctors to diagnose and formulate treatment plans.Organ segmentation techniques are essential in techniques such as interventional ablation,image-guided surgery,and magnetic induction hyperthermia.Organ segmentation technology is mainly used to process CT image data,but it is a difficult task to achieve fast and accurate segmentation of organs in CT images.Because the CT image is more complicated,it not only lacks simple linear features,but also has problems such as uneven gray scale,artifacts,and similar gray scale between tissues.At the same time,CT images are three-dimensional images,and the amount of data is large.The organ segmentation algorithm is prone to the problem of excessive computation.These two reasons increase the difficulty of designing the algorithm.In order to solve the problem of fast and accurate segmentation of organs in CT images,a new three-dimensional segmentation network Vnet-S network is proposed.Then the Vnet-S network is used to propose an automatic segmentation algorithm based on cascaded Vnet-S network.The algorithm is characterized by high accuracy and low computational complexity.The organ segmentation algorithm was deployed using the Flask framework.Finally,a medical image visualization system was built based on VTK and ITK.Specifically,this thesis has done the following work on organ segmentation algorithms and system deployment:(1)Proposing a new three-dimensional segmentation network Vnet-S network.The Vnet-S network is a three-dimensional full convolutional neural network based on the Vnet network structure.The Vnet-S network optimizes the problems of the Vnet network structure.The comparison experiment proves that the Vnet-S network performance is better than the Vnet network,and the parameter quantity and calculation amount are much smaller than the Vnet network.The parameter quantity of the Vnet-S network is 15.58% of the Vnet network,and the calculation amount is 21.41% of the Vnet network..(2)Proposing a new organ segmentation algorithm.Based on the new three-dimensional segmentation network Vnet-S network,an organ segmentation algorithm based on cascaded Vnet-S network is proposed.The algorithm is built by two Vnet-S networks,the first Vnet-S network for organ localization and the second Vnet-S network for organ segmentation.In the liver segmentation experiment and lung segmentation experiment,the algorithm achieved Dice coefficients of 0.9600 and 0.9810 respectively,which proved that the algorithm can quickly and accurately segment the liver and lung.And compared with other organ segmentation algorithms,the high accuracy and low computational complexity of the algorithm are proved.(3)Building an organ segmentation system based on the Flask framework.Using the Flask framework,the organ segmentation algorithm is deployed on the server side to provide an organ segmentation algorithm service for the front-end visualization system.Through system testing,the system can quickly segment the CT image.(4)Building a medical image visualization system.Based on VTK and ITK to build a medical image visualization system,the system has image display function,image fusion function,measurement function and so on.Compared to the old version of the visualization system,the system adds a new visualization mode,a variety of interface layouts and functions,while the system needs less memory.