Image Segmentation Algorithm Based On Prior Information And Its Application In Medical Imaging

In medical image segmentation,segmentation for specific anatomical structures is an important part of computer-aided medical treatment.CT has high accuracy for bone tissue imaging.The research on segmentation of CT data has important significance for preoperative planning and postoperative diagnosis of orthopedic patients.However,due to problems such as osteoporosis and bone lesions,the boundary of bone tissue is often blurred,and the gap between bones is small.These problems make it more difficult to accurately segment bone tissue from CT sequences.In this thesis,we propose a specific target level set segmentation model based on prior information for CT sequences.First,we preprocess the CT sequence.Then,the features of the image are extracted,and the discriminant function is constructed by learning these features through machine learning.After that,the intensity fitting term in the level set functional is constrained by discriminant function.And a new edge stop function is constructed by combining the discriminant function and the gradient features of the image.Furthermore,the energy functional is constructed by using the statistical information and gradient information of the image.The initial curve of the level set is designed according to the relationship between the slices and the intensity information of the current slice.Finally,the energy functional is solved by the variational method and the gradient descent flow,and the accurate segmentation of specific targets in the CT sequence is achieved.The main contributions of this thesis are as follows: 1.Constructing discriminant function based on automatic prior feature extraction and fusion 2.Curve evolution algorithm based on prior features.