Research On Deep Learning And Gray Model Of Thalamus Segmentation

The thalamus is a very important sensory center in human body,which plays an important role in the regulation of movement and nerve.Certain neurological diseases can cause changes in the thalamus,such as Parkinson's disease and schizophrenia.Doctors can target thalamic tissue to perform the surgery.Therefore,it is the key step of quickly and accurately segment thalamus regions from the images for the treatment of diseases.Because the gray value of the thalamus tissue is similar to the gray value of surrounding tissues,and there is noise interference,the results of some current thalamic segmentation methods are poor.Therefore,the accuracy of thalamus segmentation is low and the time-consuming.In this paper,an algorithm for thalamic tissue segmentation based on deep learning and gray scale model is proposed.The algorithm consists of two steps:(1)Preprocessing and rough thalamic segmentation:the image size is reduced by cutting method,and the cropped thalamic images are expanded by translation transformation and other extension methods.Then,the convolutional neural network model is used to obtain the rough thalamic segmentation results.(2)Grayscale model:the initial contour of the level set model is obtained through the result image of the first step.The gray constraint term is established,and the constraint term is added to the locally fitted level set model.Then,the accurate thalamus segmentation can be achieved by the evolution of the level set contour.In this paper,deep learning and gray model were combined to segment the thalamus,and the initial contour of the level set was obtained through rough segmentation results to avoid the problem of initial contour selection by the traditional level set.In addition,the gray constraint term obtained from the coarse segmentation results can ensure that the evolution curve evolves in the thalamus region and improve the segmentation accuracy of thalamus.This project designed the thalamus assisted segmentation system,which can be used for doctors to diagnose and treat specific nervous system diseases.By comparing a variety of level set methods,the experimental results show that the proposed method is better than other methods in the segmentation of thalamus.Our framework can also apply to other brain tissues segmentation.