Back to Projects List
Baptiste Baquero Maxime Gillot Lucia Cevidanes Juan Carlos Prieto
The intraoral scanner is a new tool for clinicians that allows new perspectives of development in the dental field and more particularly in orthodontics. For clinicians, it is important to quantitatively evaluate and compare their results on a large number of data. They need information about distance and angle from the 3D coordinates of the dental landmarks. 42 dental landmarks are digitized manually for each patient by orthodontists and clinicians, but manual processing is really problematic because it’s time consuming and accuracy errors can regularly be found. Therefore, proposing a robust, fast and accurate method to automatically find landmark can assist clinicians in those important but time consuming tasks. In this article, we will explore techniques to automate the search for these landmarks from a 3D digital dental model and the 3D landmark coordinates. We are going to work on deep learning techniques to allows models to determine by itself image features to better capture the complex anatomical variation and find the perfect landmark position on the teeth.
The goal is to have a model that automatically finds accurate landmarks on the digital dental model.
Previous work :
Work to continue :
I began a new method for my project (automatic landmark identification on digital dental model) The previous method was based on the movement of one agent in the 3D space to reach the perfect postion of the landmark on the tooth, but this method was not really efficient and precise. We decided to work on an other method based on segmentation with PyTorch. The objective is to have different 2D views of the jaw in input of the model and in output the same model of the jaw with a segmentated area in the region of the landmark. With this segmentation we’ll be able to recover the coordinates of the points in this area and then the position of the landmark.