Dental CBCT single tooth segmentation method based on deep learning

• Main Authors: XU MIAO, WEI QIANG, MEN JINGRU
• Format: Patent
• Language: Chinese; English
• Published: 12.03.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention discloses a dental CBCT single tooth segmentation method based on deep learning. The method is mainly used for segmenting a single tooth in a cone-beam computed tomography image, and comprises three key steps of image preprocessing, semantic segmentation and single tooth segmentation. In the image preprocessing stage, a three-dimensional surrounding frame containing all teeth is successfully segmented from a CBCT image through maximum gray projection and threshold filtering. In the semantic segmentation stage, a new coding-decoding structure network is adopted, and a more accurate tooth semantic segmentation result is obtained by introducing mesh dense connection, a multi-scale feature module and depth supervision. And in a single tooth segmentation stage, processing the smoothed binary image through a marking watershed segmentation algorithm to obtain a single tooth segmentation image. The method provided by the invention provides an efficient and accurate segmentation means for the single toot