Automatic Segmentation Of The MRI Sequences Using Level Set Method

Medical image segmentation is the basic issue of medical image processing technologies, as well as one of the problems. Medical image segmentation provides important supplementary data for medical pathology, clinical treatment, rehabilitation assessment, which helps doctors obtain useful information from the medical imaging quickly and accurately. All along, many segmentation methods have been applied to the medical field. Among them, level set segmentation method as a more active method has made better application. This article uses the level set method for the automatic segmentation of skull and brain tissue extraction of brain magnetic resonance(MRI) image sequences.Firstly, our article analyzes that the method Li C(Minimization of Region-Scalable Fitting Energy for Image Segmentation,2008) will cause the defect of the local minimum points. The method proposed by our article removes the energy constraints containing the local information in article 32, and increases the area of the region surrounded the curve as constraints. Meanwhile it introduces edge indicator function to the length constraint and the area constraint, driving the level set curve to the edge of the skull. And the result of the single brain MRI segmentation is proved effective.Secondly, after that single segmentation result is effective, in order to achieve the automatic segmentation of image sequences, we present two methods of initialing curves automatically—based on image expansion and circumscribed polygon. Experimental verification shows that the circumscribed polygon method is more ideal in the segmentation results, so we choose it for segmentation. According to the curve evolution result of the former image, we draw its circumscribed polygon as the initial curve outline of the latter image, so as to implement the automatic segmentation of brain MR image. And the accuracy and speed are perfect.Thirdly, we propose a method based on combining regional information and edge information for brain tissue segmentation of magnetic resonance images. Regional information ensures accurate tracking with complex groove back of brain tissue; while the edge information ensures the curve evolution successfully reach the edge of brain tissue.At last, the automatic segmentation of the skull has implemented on the VC++ software platform, including software interface design, algorithms transplantation, debugging and running. Experiments show that the software runs stably, less memory consumption.