Robust Teeth Detection in 3D Dental Scans by Automated Multi-View Landmarking

• https://www.scitepress.org/PublicationsDetail.aspx?ID=6XIfWnl5LKU=&t=1

Landmark Detection in 3D, Polygonal Models, Multi-View Deep Neural Networks,
RANSAC, U-Net, Heatmap Regression, Teeth Detection, Dental Scans

Landmark detection is frequently an intermediate step in medical data analysis.
More and more often, these data are represented in the form of 3D models. An
example is a 3D intraoral scan of dentition used in orthodontics, where
landmarking is notably challenging due to malocclusion, teeth shift, and frequent
teeth missing. Whats more, in terms of 3D data, the DNN processing comes with
high requirements for memory and computational time, which do not meet the needs
of clinical applications. We present a robust method for tooth landmark detection
based on the multi-view approach, which transforms the task into a 2D domain,
where the suggested network detects landmarks by heatmap regression from several
viewpoints. Additionally, we propose a post-processing based on Multi-view
Confidence and Maximum Heatmap Activation Confidence, which can robustly
determine whether a tooth is missing or not. Experiments have shown that the
combination of Attention U-Net, 100 viewpoints, and RANSAC consensus method is
able to detect landmarks with an error of 0:75 0:96 mm. In addition to the
promising accuracies, our method is robust to missing teeth, as it can correctly
detect the presence of teeth in 97.68% cases.