上海口腔医学

上海口腔医学 ›› 2025, Vol. 34 ›› Issue (4): 433-439.doi: 10.19439/j.sjos.2025.04.016

• 论著 • 上一篇    下一篇

基于SAM模型进行牙槽窝CBCT测量的初步研究

樊林峰1, 宋忠臣2, 张楚南3#, 艾松涛1#   

  1. 1.上海交通大学医学院附属第九人民医院 放射科, 上海 200011;
    2.上海交通大学医学院附属第九人民医院 牙周病科, 3.口腔种植科, 上海交通大学口腔医学院, 国家口腔医学中心, 国家口腔疾病临床医学研究中心, 上海市口腔医学重点实验室, 上海 200011
  • 收稿日期:2025-02-24 修回日期:2025-04-12 出版日期:2025-08-25 发布日期:2025-08-26
  • 通讯作者: 艾松涛,E-mail：ai.songtao@qq.com;张楚南,E-mail：zcn1114@163.com。#共同通信作者
  • 作者简介:樊林峰（1980-）,男,硕士,副主任技师,E-mail： 847472535@qq.com

Preliminary study of alveolar socket measurement on CBCT based on SAM

Fan Linfeng1, Song Zhongchen2, Zhang Chunan3, Ai Songtao1   

  1. 1. Department of Radiology, Shanghai Ninth People's Hospital, Shanghai Jiao Tong University School of Medicine. Shanghai 200011;
    2. Department of Periodontology, 3. Department of Oral Implantology, Shanghai Ninth People's Hospital, Shanghai Jiao Tong University School of Medicine; College of Stomatology, Shanghai Jiao Tong University; National Center for Stomatology; National Clinical Research Center for Oral Diseases; Shanghai Key Laboratory of Stomatology. Shanghai 200011, China
  • Received:2025-02-24 Revised:2025-04-12 Online:2025-08-25 Published:2025-08-26

PDF (PC)

2

摘要： 目的：评估基于深度学习技术开发的锥形束CT（CBCT）图像上自动测量牙槽窝的工具,比较其与人工测量的准确性,验证其有效性和可行性。方法：纳入29例成年患者（男11例,女18例）,平均年龄（31.31±13.77）岁。入组时均行CBCT检查,提取（5-5）牙位共427个横断面进行牙槽窝颊舌向骨量测量。开发一种新的基于SAM模型（Segment Anything Model）的交互式分割和测量工具,将其用于CBCT的牙槽窝尺寸评估。通过建立测试集和验证集,分别进行牙槽窝横断面颊舌侧骨量的人工测量和自动测量,并进行数据比较。结果：CBCT自动测量方法与人工测量之间具有显著的相关性和一致性,测试集回归分析的决定系数（R2）为0.942,验证集测量误差主要集中在-0.43~0.47 mm之间,Pearson相关系数为0.9746（P<0.001）。结论：本研究通过深度学习,开发了一种基于SAM的CBCT的自动测量工具,准确性高,显著提高了牙槽窝测量的效率。

关键词: SAM, 锥形束CT, 横断面, 牙槽窝, 自动测量

Abstract: PURPOSE: To evaluate the tool of automatic measuring CBCT, developed based on deep learning technology, and to compare its accuracy with manual measurement and to verify its effectiveness and feasibility. METHODS: Twenty-nine adult patients (11 males, 18 females) with mean age of (31.31±13.77) years old were enrolled, CBCT of enrolled patients were collected, and 427 transverse sections of alveolar teeth were extracted(5-5 position). A novel Segment Anything Model(SAM) -based interactive segmentation and measurement tool was developed and applied to the assessment of alveolar socket dimensions in CBCT. Manual and automatic measurements of bone mass in the buccal and lingual transverse section of the alveolar socket were performed by establishing the test set and the validation set, respectively, and the data were compared. RESULTS: There was significant correlation and consistency between CBCT automatic measurement method and manual measurement. The coefficient of determination(R2) of regression analysis in test set was 0.942, the measurement error in validation set was mainly varing between -0.43~0.47 mm, Pearson correlation coefficient was 0.9746 (P<0.001). CONCLUSIONS: This study developed an automatic CBCT measurement tool based on SAM through deep learning, with high accuracy and significantly improved the efficiency of alveolar socket measurement.

Key words: SAM, CBCT, Cross-section, Alveolar socket, Automatic measurement

中图分类号: