AUTOMATED 3D BONE MODELLING BASED ON GEOMETRIC FEATURES FROM IMAGES

Abstract

Currently the computational modelling tools, expert algorithms for image segmentation and three-dimensional printing devices have improved the process of manufacture of customized pieces in the context of automation of prosthesis modelling. Among several strategies to obtain the correct shape of a missing bone part, we explore a modelling method based on geometric features defined by Bezier cubic curves. The data is obtained from sets of slices of tomographic scans as a reference. From the images, we know about data from the edges in the image but we do not have any information about a missing area in a specific bone region. Thus, the objective is to search patterns for features whose values are known from similar tomographic image which matches to fill a hole in a bone. Due the free form of a bone there are a lot of parameters to be evaluated. Thus, a Data Mining approach is applied for classification and to discover the best features as shape descriptors. In this way, the prosthesis manufacture can be automated in all stages among image scanning and printing.