Research On Automatic Segmentation Algorithm Of CT Liver Image Based On The Improved FCN

Liver cancer is the second most common cancer in the world,and the number of liver cancer patients in China ranks first in the world.In the field of medical research,automatic segmentation of CT liver images is a key step in clinical treatment and an important step in computer-assisted diagnosis and liver surgery for the early diagnosis of liver diseases.However,since the size and shape of the liver vary from person to person,and the gray contrast between the liver and adjacent organs in the CT images is low,it is difficult to accurately judge the boundary information of the liver images.In this case,it is impossible to accurately segment the liver images.Aiming at the above problems,an improved segmentation method of RV-FCN network model is proposed.The main research work is as follows:(1)Aiming at the 3DIRCADb public dataset,in order to prevent the noise in the original image from interfering with the liver image segmentation,an improved BM3 D method combined with the adaptive threshold normalization strategy is proposed for denoising.Before denoising,the original image is firstly subjected to windowing processing.Then the image is denoised by BM3 D method.In the denoising process,the adaptive threshold normalization method is combined to solve the problem of uneven light distribution generated in the CT scanning process,and the image contrast is improved at the same time.The experimental results show that compared with other methods,the denoising effect of the proposed method is relatively good.(2)An improved RV-FCN network structure model is proposed to automatically segment CT liver images.This method mainly adds the optimal combination of Resnet and VGG-16 network based on the Fully Convolutional Neural Network(FCN)model.At the same time,in order to improve the generalization ability and convergence speed of the network,the Batch Normalization and PReLU activation functions are also added to improve the generalization ability and convergence speed of the network.In order to avoid overfitting in the training process,the loss function is normalized.Then,the improved model is trained and tested to obtain the initial segmentation results of liver images.Conditional Random Fields(CRF)method is used to further optimize the segmentation results of liver images to improve the segmentation accuracy.The practicality and accuracy of the improved network model are verified by the specific test results.(3)Based on the VTK platform,the two-dimensional liver images after segmentation is reconstructed into a three-dimensional structure by using the RayCasting algorithm,which is convenient for medical experts to visually measure the shape,size and lesion location of the liver to improve the accuracy of the diagnosis and clinical surgery.