Brain MR Image Segmentation Based On Measurement Error Mode

Due to the increasing aging of our population,brain diseases have become a major threat to human health.Accurate analysis and determination of the specific distribution of brain tissue is a basic prerequisite for doctors to design effective treatment plans,and the accuracy of brain image segmentation is the key to auxiliary diagnosis.Gaussian mixture models are widely used due to their low complexity and high accuracy.However,medical images often exhibit phenomena such as uneven grayscale,noise,and artifacts,while Gaussian mixture models are symmetrically distributed and rely on independent pixels without considering the relationship between adjacent pixels,so it is difficult to obtain ideal results.In order to solve these problems,this article conducts in-depth exploration from four aspects:model modeling,spatial information construction,bias field modeling,and parameter optimization.The specific content is as follows:(1)In view of the problem that the Gaussian mixture model cannot fit the asymmetric distribution data,this thesis proposes a skew normal mixture model,which is different from the hierarchical Gaussian mixture model.This method calculates the skewness through the improved expectation maximization algorithm rather than predetermined.The experimental results of distribution fitting show that the proposed method has better robustness.Aiming at the problem that the Gaussian mixture model is sensitive to noise,the skewed normal mixture model is used for noise modeling.A measurement error model based on the skewed normal mixture model is proposed.Aiming at the problem that the finite mixture model only considers the distribution information and does not consider any spatial information,the anisotropic spatial information is proposed and coupled to the measurement error model,so as to increase the robustness of the model.The experimental results and quantitative analysis show that the model proposed in this chapter has better performance than other algorithms based on finite mixture model.(2)According to issues such as uneven intensity in medical images,this thesis models the bias field and couples it into the measurement error model.It uses a linear combination of a set of orthogonal bases to fit the bias field,so as to retain more details.In addition,the initialization of this chapter uses the K-means algorithm,and optimizes the parameters of the model through the improved expectation maximization algorithm.A large number of experimental results show that the method proposed in this chapter has higher segmentation accuracy in medical brain MR images than other algorithms based on finite mixture model.