目的：探索一种基于牙周膜生理影像特征快速提取活体牙根三维形态的新方法，并对该方法的提取精度进行评价。方法：收集口腔正畸临床减数拔除的完整离体牙齿15颗，其中单根牙11颗、双根牙4颗，使用牙颌模型扫描仪获得高精度（0.02 mm）牙根三维数字模型。在牙齿拔除前为患者拍摄大视野锥束CT（0.3 mm体素分辨率）， 应用Mimics 18.0医学影像学软件，基于减数牙齿的根骨和牙周膜阈值蒙版，运用形态学运算、布尔运算及智能扩展运算等功能获得牙根三维数字模型。在Geomagic Studio 2012软件中，对应用本研究方法提取的活体牙根三维模型与同名离体牙根三维扫描模型进行三维形貌偏差的比较和三维尺寸误差的测量，评价该提取方法的精度。结果：提取的15颗牙根三维形貌偏差平均为0.22 mm，近远中径误差为0.46 mm，颊舌径误差为0.36 mm，根长误差为-0.68 mm，单牙提取时间约为2~3 min，基本可以满足口腔临床对牙根重建的精度要求。结论： 建立了基于活体牙影像数据快速提取牙根三维形态的新方法，一定程度上简化了传统手工提取的步骤，提高了单牙牙根提取的效率和自动化程度，其针对全牙列提取的方法策略有待进一步研究。

Objective： To explore a new method for rapid extracting and rebuilding three-dimensional (3D) digital root model of vivo tooth from cone beam computed tomography (CBCT) data based on the anatomical characteristics of periodontal ligament, and to evaluate the extraction accuracy of the method. Methods: In the study, 15 extracted teeth (11 with single root, 4 with double roots) were collected from oral clinic and 3D digital root models of each tooth were obtained by 3D dental scanner with a high accuracy 0.02 mm in STL format. CBCT data for each patient were acquired before tooth extraction, DICOM data with a voxel size 0.3 mm were input to Mimics 18.0 software. Segmentation, Morphology operations, Boolean operations and Smart expanded function in Mimics software were used to edit teeth, bone and periodontal ligament threshold mask, and root threshold mask were automatically acquired after a series of mask operations. 3D digital root models were extracted in STL format finally. 3D morphology deviation between the extracted root models and corresponding vivo root models were compared in Geomagic Studio 2012 software. The 3D size errors in long axis, bucco-lingual direction and mesio-distal direction were also calculated. Results: The average value of the 3D morphology deviation for 15 roots by calcula-ting Root Mean Square (RMS) value was 0.22 mm, the average size errors in themesio-distal direction, the bucco-lingual direction and the long axis were 0.46 mm, 0.36 mm and -0.68 mm separately. The average time of this new method for extracting single root was about 2-3 min. It could meet the accuracy requirement of the root 3D reconstruction fororal clinical use. Conclusion: This study established a new method for rapid extracting 3D root model of vivo tooth from CBCT data. It could simplify the traditional manual operation and improve the efficiency and automation of single root extraction. The strategy of this method for complete dentition extraction needs further research.