Example-oriented full mandible reconstruction based on principal component analysis

The human mandible reconstruction is widely used in medical cosmetology, criminal investigation, archaeology and anthropology. This study is devoted to rapidly reconstructing 3D personalized mandibles, especially when the mandible is severely damaged or totally lost. An example-oriented method based on principal component analysis (PCA) is used to reconstruct the 3D shapes of mandibles. The predictor variable is an input cranium, while the response variables is a full mandible. The linear relationship between the cranium and mandible is obtained based on a number of skull samples, which is used as prior knowledge. The PCA method reduces the dimensionality of the skull data to improve the accuracy and speed of the reconstruction process. Experiments are conducted based on 215 skull models in the Chinese craniofacial database established in the study of craniofacial morphology informatics at Beijing Normal University. In our experiments, parameters are modified to help improve the reconstruction results. The proposed method exhibits higher accuracy than state-of-the-art methods according to comparison experiments results. Moreover, this method is robust to various types of input data and noise. Results indicate the proposed method is a fast and feasible tool for reconstructing 3D customized full mandibles.