Pericardium Segmentation Based On Convolutional Neural Network In CT Image

Medical image segmentation is the basis of computer-aided diagnosis and treatment.It is of unique value for improving the efficiency and accuracy of doctors' diagnosis of diseases such as pericardial effusion and pericardial tumor.Pericardial segmentation based on CT images can effectively improve the efficiency and accuracy of doctor diagnosis.At present,the results of pericardial segmentation based on traditional methods cannot meet the needs in practical applications,and the basic deep learning method is widely used and has good segmentation results.Therefore,the goal of this article is to propose a pericardium segmentation method based on convolutional neural networks.On this basis,the main work of this article is as follows:Firstly,for the problem that there is no public data set for pericardial segmentation,under the guidance of professional doctors,using the medical image annotation tool ITK-SNAP,a pericardial segmentation data set was produced,and the result of the data set was approved by the doctor.Secondly,according to the characteristics of the pericardium in the image,the three-dimensional convolutional neural network is taken as the research focus,and the pericardial segmentation algorithm based on the SDV-Net of the three-dimensional convolutional neural network is designed and implemented.The pericardium segmentation algorithm mainly includes data preprocessing and enhancement,network model training and testing,and data postprocessing.Among them,SDV-Net is based on V-Net,which leads to the optimization of some existing problems.Finally,by using the same training and test data as other models for training and testing,the evaluation indicators of the test results are analyzed to prove that the improved SDV-Net has a better effect on pericardial segmentation,thus proving that this is based onSDV-Net.The pericardial segmentation algorithm can perform automatic and accurate pericardial segmentation tasks.