Automatic 3D segmentation of mandible for assessment of facial asymmetry

摘要

This paper presents an vision-based approach to support orthodontists in assessing facial asymmetry by means of: 1) a fully automatic 3D mandible segmentation method which works on CT (Computer Tomography) scans and 2) surface measurements between user-defined landmarks located on the reconstructed surface. The proposed method shows several advantages with respect to the conventional techniques: it is much faster since manual segmentation of the areas of interest on the CT 2D slices is not needed and perform surface measurements instead of simply computing linear distances as carried out by other 3D methods. The results showed that the measurements obtained by our method are compatible with the ones obtained by two experts using a digital caliper on 10 CT scans.