Mesh based segmentation for automated margin line generation on incisors receiving crown treatment

  • 2025-07-30 17:34:45
  • Ammar Alsheghri, Ying Zhang, Farnoosh Ghadiri, Julia Keren, Farida Cheriet, Francois Guibault
  • 0

Abstract

Dental crowns are essential dental treatments for restoring damaged ormissing teeth of patients. Recent design approaches of dental crowns arecarried out using commercial dental design software. Once a scan of apreparation is uploaded to the software, a dental technician needs to manuallydefine a precise margin line on the preparation surface, which constitutes anon-repeatable and inconsistent procedure. This work proposes a new frameworkto determine margin lines automatically and accurately using deep learning. Adataset of incisor teeth was provided by a collaborating dental laboratory totrain a deep learning segmentation model. A mesh-based neural network wasmodified by changing its input channels and used to segment the prepared toothinto two regions such that the margin line is contained within the boundaryfaces separating the two regions. Next, k-fold cross-validation was used totrain 5 models, and a voting classifier technique was used to combine theirresults to enhance the segmentation. After that, boundary smoothing andoptimization using the graph cut method were applied to refine the segmentationresults. Then, boundary faces separating the two regions were selected torepresent the margin line faces. A spline was approximated to best fit thecenters of the boundary faces to predict the margin line. Our results show thatan ensemble model combined with maximum probability predicted the highestnumber of successful test cases (7 out of 13) based on a maximum distancethreshold of 200 m (representing human error) between the predicted and groundtruth point clouds. It was also demonstrated that the better the quality of thepreparation, the smaller the divergence between the predicted and ground truthmargin lines (Spearman's rank correlation coefficient of -0.683). We providethe train and test datasets for the community.

 

Quick Read (beta)

loading the full paper ...