Improved segmentation of teeth in dental models

The identification of different teeth in a human dental model is an essential part of virtual treatment planning in digital orthodontics. An algorithm that is almost fully automatic is presented to segment a 3-dimensional dental model (a triangle mesh) into individual tooth objects, along with its implementation in a software tool that can be deployed in clinical settings. The proposed technique avoids the limitations of previous approaches by first separating gums from teeth and then separating individual teeth. The only user intervention needed in most cases is the one-time setting of a certain curvature threshold via an intuitive slider, coupled with real-time visual feedback. The method is found to be robust across a range of real-world models, even in the presence of malocclusions and noise.