A segmentation technique for maxillary sinus using the 3-D level set method

In this paper, we propose an improved model, called three-dimensional (3-D) level set method, for active contours to detect maxillary sinuses in cone beam computed tomography(CBCT) images. This method is based on the techniques of two-dimensional(2-D) level set method and active contour without edges (or Chan-Vese) method. This model can detect the maxillary sinus correctly not only in normal cases, but also in special cases such as low intensity regions, blurring regions, or multiple regions which normally occur in the CBCT images. Although, the energy decreasing speed of the 3-D level set method is faster than the 2-D level set method, this method takes more computation time than the 2-D level set method. The combination of the parameters depends on the input images. Although some of these parameters are quite sensitive to adjust, the effects of each parameter are discussed. Results show the advantages of the proposed method over the 2-D level set method.