Automatic Segmentation Of Teeth In Digital Dental Models

Authors

David Anton Mouritsen, University of Alabama at Birmingham

Advisory Committee Chair

Chung H Kau

Advisory Committee Members

Andre Ferreira

John K Johnstone

Christos Vlachos

Document Type

Thesis

Date of Award

2013

Degree Name by School

Master of Science in Dentistry (MScD) School of Dentistry

Abstract

Introduction: Within the last decade, there has been an increase in the use of digital records in orthodontics. An orthodontic diagnostic setup is a simulated treatment outcome prepared from dental models. To create a diagnostic setup from digital dental models, the first step is to segment the models, or identify the boundaries of each tooth. In the past, a human operator would manually segment the models using a computer program to draw the boundary that separates the teeth from each other and the gingiva. The ideal, fully automatic, segmentation algorithm would not require a user to interact with the digital model. The purpose of this research is to design, develop, and test a computer algorithm to automatically segment digital dental casts and to release the project as open-source software to stimulate future research. Methods: A tooth segmenting algorithm was conceptualized and implemented as a C++ plug-in for Meshlab, an open source system for the processing of 3D triangular meshes. The C++ source code for the project was written using Qt Creator 2.2.1. The only input required for the algorithm, other than the model file, is the number of teeth for which to search. As test cases for the algorithm, six maxillary and six mandibular OrthoCad models were converted from .3DM to .STL format. The accuracy of the segmentation, measured by visual inspection, and the runtimes for each test model were recorded. Results: The 12 models (6 mandibular and 6 maxillary) had 160 teeth total to segment. The segmentation algorithm correctly segmented 133 of the 160 teeth for an 83.13% success rate overall.The algorithm correctly segmented 70 of 80 maxillary teeth for an 87.5% maxillary success rate, and 63 of 80 mandibular teeth for a 78.8% mandibular success rate. The average runtime was 32 seconds on a 3.0GHz Intel® CoreTM2 Duo CPU with 4 GB RAM. Conclusions: With overall segmentation success at 83.13% and average runtime of 32 seconds, future research will focus on improving accuracy and speed, while maintaining usability. More research needs to be done before efficient, fully-automated segmentation of digital dental casts becomes a reality.

Recommended Citation

Mouritsen, David Anton, "Automatic Segmentation Of Teeth In Digital Dental Models" (2013). All ETDs from UAB. 2529.
https://digitalcommons.library.uab.edu/etd-collection/2529