Towards Automated Human Identification Using Dental X-ray Images

Abstract

Systems for automated human identification from dental X-ray images can be used to greatly reduce the necessary effort spent today by dental forensics experts. In this work a new methodology is proposed to create a system for automated dental X-ray identification. The methodology includes both state-of-the-art methods and a novel method for separating a dental X-ray image into individual teeth. The novel method is based on lowest cost pathfinding and is shown to achieve comparable results to the state-of-the-art. In experiments it is able to separate 88.7% of the teeth in the test images correctly. The identification system extracts tooth and dental work contours from the dental X-ray images and uses the Hausdorff-distance measure for ranking persons. The results of testing the system on a new data set show that the new method for dental X-ray separation functions well as a component in a functional identification system and that the methodology on the whole can be used to identify persons with comparable accuracy to related work. In 86% of cases, the correct person is ranked highest. This accuracy increases to 94% when the five highest ranked images are considered. Due to small distances in similarity between highest ranked individuals, doubts are raised concerning the scalability of the method. This is seen as a matter of expansion, such as refining features, rather than redesign. The conclusion is that the proposed methodology, including the path-based method of separation, performs well enough to be worth consideration when designing an automated dental identification system.