Research On Method Of 3D Segmentation Of Liver Based On Conditional Generative Adversarial Neural Networks Using Feature Reduction

At present,liver cancer is seriously threatening the health of our citizens and becoming one of the incidence rate of malignant tumors.Every year there are a large number of new cases and showing a continuous growth trend.In recent years,with the rapid popularization of medical equipment,medical imaging has become the key basis of medical diagnosis.In the field of liver cancer treatment,computed tomography has become an important basis for doctors to make early diagnosis and preoperative planning.In the process of computer-aided diagnosis and treatment of liver cancer,accurate segmentation of liver region from abdominal CT image is an extremely important step.However,due to the defects of liver tissue distribution and CT imaging process,the gray level of liver region in CT image is uneven and the boundary with adjacent tissues and organs is blurred,which makes liver segmentation a very difficult challenge.In the traditional medical diagnosis,professional doctors need to manually segment the liver from abdominal CT images.This method is not only time-consuming and labor-consuming,but also low efficiency.In the process of manual segmentation,it is easy to add the subjective factors of doctors,lack of unified segmentation standards,resulting in uneven segmentation quality.Therefore,the computer image processing technology arises at the historic moment,which can use the well-trained segmentation model for automatic liver segmentation of abdominal CT image.This technology not only liberates the doctors,but also greatly improves the quality and accuracy of liver segmentation results,improves the efficiency of diagnosis,and gradually becomes an important means of liver CT image segmentation.With the rapid development of deep learning in recent years,the application of deep convolution neural network in the field of computer vision is becoming more and more mature.At the same time,medical image segmentation as an important research direction of computer vision has also been rapidly updated.The main research contents of this paper include:(1)Aiming at the problem of low segmentation accuracy of the original 3D U-Net network,an improved network based on 3D U-Net is proposed to segment the liver.In order to make the network pay more attention to the characteristic information of the liver and enhance its importance,and reduce the role of irrelevant information such as background,SE module is introduced;In order to enable the network to obtain multi-scale feature information of the feature map and expand the receptive field of the network,the pyramid pooling module is introduced.(2)In order to solve the problem of insufficient training data due to the difficulty of acquiring labeled 3D data,an improved 3D U-Net network is embedded into the framework of generative countermeasure network,and a semi-supervised3 D liver segmentation optimization method based on generative countermeasure network is proposed,in which the improved 3D U-Net network is used as the discrimination network of generative countermeasure network.The limited labeled image training set is used to train the depth model,and the unlabeled image training generation network is used to generate pseudo images.The better segmentation results can be obtained by expanding the data set.(3)In order to solve the problem of poor quality of 3D abdominal pseudo image generated by using random noise as input,a deep convolutional neural network based on feature reduction method is designed to generate more realistic pseudo image.The network is embedded into the framework of generation countermeasure network,which is used as the generation network of generation countermeasure network.The experimental results show that the proposed semi-supervised 3D liver segmentation method based on feature reduction generation antagonism network can greatly improve the segmentation effect of liver.