Research On Medical Image Processing Method Based On Level Set And Convolutional Neural Network

With the advent of the era of big data and the continuous development of modern medical imaging technology,the quality and quantity of medical images are constantly growing,which makes it particularly difficult to process and analyze images by human.The images without processing have little value,so a method is naturally needed to automatically conduct preliminary process on a large number of image data.Compared with the traditional segmentation algorithm,the active contour model based on the level set method can deal with the complex changes of topology,and the numerical calculation of it is relatively simple.In recent years,the convolutional neural network has been developed rapidly and applied to the field of medical image processing,which can not only improve the accuracy of medical image classification,but also significantly improve the efficiency of image classification by saving the feature extraction steps of traditional classification methods.If the medical image can be automatically predicted and classified into normal and diseased classification by computer,and segment the image that is judged to be diseased,this will facilitate the subsequent diagnostic analysis of the doctor.This paper introduces the process of image segmentation and classification in two parts.In the segmentation part,the curve evolution theory and the principle of level set are introduced firstly.Then the recognition effect of level set model on the boundary of medical image is optimized by improving the traditional energy function,and the accuracy of segmentation is improved.In the classification part,the parameters and methods which affect the classification accuracy of the network are determined by comparative experiments.The network is established according to these parameters.Finally,train and test the network through medical images.In order to verify the effectiveness of the proposed method,this paper selects the MRI images of normal people and people who are ill for segmentation and classification.During the segmentation,the distance-regularization level set is used to remove the unrelated regions in the image.Then,the enhanced image is segmented by the improved energy function proposed in this paper which is the level set algorithm with the bilateral filter function being added to the energy function.Then the result is compared with the standard of manually segmented image.During the classification,the structure of the convolutional neural network is constructed and the parameters are determined first.Then the MRI images of the two groups of people are used to train and test the results.The results show that the classification algorithm proposed in this paper improves the classification accuracy,and the segmentation algorithm is more sensitive to the boundary compared with the traditional level set.Meanwhile,the algorithm proposed in this paper reduces the dependence on the initial contour and improves the accuracy,sensitivity and specificity of segmentation.