Segmentation Algorithm Based On Brain MRI Images Finite Mixture Model

Medical imaging technology has become one of the essential tools for medical diagnosis. A full range of data of various imaging techniques provides convenience to the diagnosis of the disease, and also greatly improves the accuracy of disease diagnosis. In various imaging techniques, magnetic resonance imaging (MRI) is a kind of non-ionizing radiation imaging, a kind of no traumatic examination, providing multi-parameter and multiple-sequence imaging, and it provides the high resolution of the soft tissue, it is superior to CT in brain and spine imaging. In recent years, the incidence and disability of the brain disease has gradually increased. It is a serious threat to human life and health, so the brain MRI image processing and analysis has become a hot topic of research. Medical image segmentation is the foundation for further image analysis and processing, it provides support for three-dimensional reconstruction, surgical simulation and diagnosis. Because of the characteristics of itself and effects of electronic noise caused during the imaging process, segmentation of MRI images has become one of the difficult problems in image analysis.In this paper, the main work is as follows:(1) Studies and implements a Gaussian mixture model based on MRF. The model incorporates spatial information and improves the robustness to noise of the segmentation results. In comparison to other mixture models that are complex and computationally expensive, this method is fast and easy to implement. In mixture models based on MRF, the M-step of the expectation maximization (EM) algorithm can’t be directly applied to the prior distribution π(?) for maximization of the log-likelihood with respect to the corresponding parameters. Compared with these models, this method directly applies the EM algorithm to optimize the parameters, which makes it much simpler. Finally, experimental results demonstrate its effectiveness as compared with other methods such as FANTASM and HMRF-FCM.(2) Studies and implements a Finite mixture model based on the Student’s-t distribution. First, this model exploits Dirichlet distribution and Dirichlet law to incorporate the local spatial constrains in an image. Secondly, this model directly deal with the Student’s-t distribution in order to estimate the model parameters, whereas, the Student’s-t distributions in previous models are represented as an infinite mixture of scaled Gaussians that lead to an increase in complexity. Finally, instead of using expectation maximization (EM) algorithm, this method adopts the gradient method to optimize the parameters. Experiments show that it’s better than the previous method.(3) Propose a new mixture model based on the Student’s-t distribution. This method does not use MRF model or Dirichlet distribution. It incorporates the spatial relationships between pixels through prior distribution. Since it introduces no additional parameters, so the algorithm is fast and easy to implement. The algorithm also deals directly with the t-distribution, using a gradient method to estimate model parameters. By comparing the experimental results with other methods, it demonstrates the robustness, accuracy and effectiveness of this method.