Teeth-Simplification：一种面向牙齿分割的性能提升方案

打开文本图片集

中图分类号：TP301 文献标志码：A 文章编号：2097-3853（2025）04-0367-07

Teeth-Simplification： a performance improvement scheme for tooth segmentation

XIE Qiqin1，²，LIU Shijian1²，CHEN Jie 1，2 ， ZOU Zheng³ （1.School of Computer Science and Mathematics，Fujian University of Technology，Fuzhou 35O118， China; 2. Fujian Provincial Key Laboratory of Big Data Mining and Applications，Fuzhou 350118，China; 3. School of Computer and Cyber Security，Fujian Normal University，Fuzhou 350117，China）

Abstract：In the field of oral medicine，accurate tooth segmentation is crucial for computer-aided diagnosis and treatment planning.To address the geometric and labeling errors introduced by simplified algorithms，a strategy called Tooth-Simplification is proposed to improve and evaluate the performance of tooth segmentation. It includes a simplified algorithm based on preserving boundaries，a label conversion strategy based on dimension mapping，and a performance evaluation system with asymmetric resolution.Experiments based on the public dataset 3DTeethSeg show that the Teth-Simplification strategy is adaptable to mesh segmentation methods such as MeshSegNet， TSGCNet，SGTNet，and GACNet，with boundary DCD error values reduced by 73.68% ， 64.63% ， 21.15% ，and 33.33% respectively，effectively improving the performance of tooth segmentation.

Keywords： triangular mesh ; deep learning; mesh simplification; tooth segmentation

在口腔医学领域，牙齿分割作为一项关键基础技术，对于计算机辅助诊断、设计和制造至关重要[1]。（剩余8604字）