Automatic teeth thresholding in cone beam CT with convolutional neural networks and tooth segmentation with the watershed transform

The segmentation of a tooth from a Cone Beam Computed Tomography (CBCT) scan is a time consuming process. This study focuses on the automatic segmentation of teeth from CBCT data. It proposes a method for automatic thresholding of teeth in a CT scan using convolutional neural networks, as well as a method for the automatic segmentation of a tooth using the watershed transform. The study shows that the models used in this study suffer from underfitting and are therefore not suitable for automatic thresholding in CBCT scans. This study also finds that the watershed transform is able to segment a tooth slice by slice with reasonable accuracy