Mandibular Canal Segmentation Based on Shape-Driven Level-set Algorithm Restrained by Local Information
1 School of Life Science and Technology, University of Electronic Science and Technology, Chengdu 610054,China
2 College of Electronic Engineering, Chengdu University of Information Technology, Chengdu 610225,China
3 College of Network Engineering, Chengdu University of Information Technology, Chengdu 610225,China
Abstract：In CT images, the mandibular canal is difficult to distinguish from other surrounding tissues due to its tubular structure. This paper proposed a shapedriven levelset algorithm restrained by local information to segment the mandibular canal with high accuracy. Firstly, according to the location of the distribution of the mandibular canal, we reconstructed many cross sectional images of corresponding parts of the mandibular canal, and extracted the mandibular canal from the cross sectional images, and then applied principal component analysis (PCA) to carry out the shape priori statistics of the mandibular canal. Finally, based on shape-driven level-set algorithm, restrain the evolution of the level-set energy function to improve the segment result in the fuzzy region by introducing the local information of the mandibular canal’s surrounding tissue. By applying the criterion of segmentation accuracy in the region of interest, the segmentation accuracy of shape-driven level-set algorithm was 4.19%, while the segmentation accuracy of our method is 1.82%. Experimental results showed that this method could effectively segment patients’ mandibular canal and provide an effective method for locating the position of the mandibular canal.
