Medical Image Segmentation Based On Deep Learning And Distance Regularization Level Set Evolution

With the continuous development of computer science,image technology began to be integrated with clinical medicine,resulting in modern medical imaging technology.By extracting and analyzing the biometric information of various tissues of the human body,modern medical imaging technology can obtain information on the physiological function and metabolic state of the tissue at the examination site of the patient,and reflect the health status of the patient in real time.The development of medical image imaging technology has transformed the traditional anatomical observation of human tissue into the selection and processing of organ images with the help of computer tomography instruments.Clinicians can diagnose patients more intuitively and efficiently and give corresponding treatment plans,reducing the need for The pain and trauma of the patients are saved,and precious treatment time is obtained for the patients.By deepening the research of medical image segmentation technology,the accuracy of medical images can be improved and the corresponding diagnostic operations can be performed,which promotes the progress of modern medicine and has a profound impact on the future medical field.Therefore,the popularity of medical image segmentation technology is in China Outside has been high.With the development of digital technology and artificial intelligence technology,traditional tissue slices are replaced by digital slices.Digital slices can be saved and transmitted,and can also be processed and analyzed to assist doctors in making professional diagnosis of patients.The traditional level set evolution method has been very popular in image segmentation.The traditional level set model evolves according to the gradient information of the image.However,when there are multiple boundaries and strong boundaries,the gradient information cannot distinguish the target boundary and the background boundary.So when there are multiple unwanted boundaries in the image,it is often impossible to converge to the desired boundary.In addition,the traditional level set model is very sensitive to the initial position,and the initial contours of different shapes and positions have a great impact on the segmentation results,and even lead to the failure of segmentation.In order to solve the problem of multiple boundaries and strong boundaries,we improve the traditional distance regularization level set evolution method,introduce a new edge indicator function,and propose an improved distance regularization level set evolution method for image segmentation.At the same time,we added a new regular term to make the curve converge smoothly.In order to solve the problem of initial point selection in level set segmentation,this paper combines deep learning with Unet network to presegment the original image,and proposes a new automatic detection method of initial contour,which can make the initial contour in the center of the image From the beginning,it lays the foundation for the second level set fine segmentation.To sum up,our new model combined with deep learning has good results in medical image segmentation.