The Research Of Intracranial Brainstem Images Segmentation Method Based On Active Contour Model

There are a large number of digital images (such as angiography images, remote sensing images, etc)in the field of various engineering (such as military, medical, etc) with the rapid development of information technology and science. How to extract useful information from a large number of images has become An important research focus in the image processing technology.The human brain is an important organ of the human body, And the function of intracranial brainstem is to maintain the individual life, including heart rate, respiration, digestion and other important physiological functions. This shows that the study of the intracranial brainstem region segmentation technique are of great significance for three-dimensional reconstruction of the intracranial brainstem and the study of the brainstem adjacent to blood vessels and the clinical diagnosis of brain diseases.Level set method is an important method of image segmentation based on partial differential equations, whose main idea is the usage of higher-dimensional space curve evolution. It can make the curve to split and merge well, and get the level set function evolution partial differential equations by minimizing the energy functional, this method is also known as variational level set method. Based on the level set theory, this article introduced the relevant classical methods of image segmentation and the mathematical theory of the level set firstly, besides focusing on a few more classic active contour models, and analyzed the advantages and disadvantages and the applicable occasions of various models.This paper improves the boundary indicator function of model Li, which enhance the velocity of the evolution of the model. Through a lot of experimental study on the boundary level set model and the regional level set model, finding that images can not pinpoint the target boundary by the only use of regional information, also there will be a certain lack of anti-noise performance and the speed of the model-by the only use of the boundary information. Therefore, taking advantage of boundary information and regional information can not only be more accurate positioning boundaries, but also improving the speed and noise performance of the model. So this paper combines the advantages of boundary-based model and region-based model and proposes a hybrid model based on boundary and region (referred to as DR-CV model). DR-CV model takes good use of the boundary and regional information of the image, which has many advantages such as insensitive to the initial contour position, strong anti-noise capability, more fast split, more accurate positioning boundary and so on. In addition, the new model is applied to intracranial brainstem split of the brain MRI and blood vessels to enhance CT image, we found that DR-CV segmentation is the best by comparing with other methods, it has certain advantages in the speed of model, the accuracy of the positioning border, noise immunity. By obtaining eight cases of brainstem DSI(split similarity coefficient) randomly of the intracranial MRI cases, the experimental results are more than90%, the average of DSI is95.32%, it proves that the segmentation is better, and provides a solid foundation for the intracranial brainstem dimensional reconstruction.