Statistical Elimination Based Approach to Jaw and Tooth Separation on Panoramic Radiographs for Dental Human Identification

Abstract

Dental biometrics is a type of biometrics that uses dental information to identify individuals. Well-known biometric features such as fingerprints and gait images have been successfully used to identify individuals. However, these features can be easily damaged. Teeth are more durable than other biometric features. Therefore, dental biometrics are used when other biometric features are not available. There are several types of dental radiographs. Panoramic radiographs are a type of x-ray that show the entire jaw. In these x-rays, all the teeth are viewed together. Panoramic x-rays contain more information about the tooth and jaw structures. However, they also contain unwanted elements such as the bite disc, mandible, nasal bone, etc. This makes them more difficult to process. All types of dental radiographs have difficulties in processing due to slight differences in brightness, overlapping or differently aligned teeth. Identifying individuals from dental radiographs often involves the following main steps: jaw separation, tooth segmentation, feature extraction, and feature matching. The accuracy of jaw and tooth segmentation influences the next steps. In this study, a new fully automatic method for separating mandible, maxilla, and teeth in panoramic radiographs is proposed. The proposed method achieved high accuracy in jaw separation. It also achieved better jaw separation performance than comparable studies. Although the proposed method is a fully automatic method, its performance in tooth separation is close to that of the compared semi-automatic method. In the proposed study, a jaw separation ratio of 0.99, based on the number of correctly aligned teeth, was achieved. The detection rate of the separators for the teeth in the mandibular jaw is 0.90 and the accuracy is 0.86. For maxillary teeth, these values are 0.92 and 0.90, respectively. The results are promising for the automatic segmentation of panoramic radiographs for human identification.