Fully automated detection of the mandibular canal in cone beam CT images using Lie group based statistical shape models

Automatic detection of mandibular canal in cone beam CT data is an essential step for planning and guiding implant surgery. In this work, we present a new detection method based on combining statistical shape models and Lie group. The proposed methodology consists of three steps. Firstly, a method based on multi-scale low rank matrix decomposition is used for noise removal and image enhancement. Subsequently, a Lie group based statistical shape model is constructed to represent shape variation and fast marching is employed to localize the location of the mandibular canal more accurately. Quantitative results show that accurate and fully automatic detection of mandibular canal is feasible. Moreover, the proposed method based on Lie group based statistical shape model outperforms two previous methods based on statistical shape model in the literature, i.e. conventional and conditional statistical shape models. The average value of Dice similarity index and symmetric distance are 0.92 and 1.02 mm, respectively.