国际口腔医学杂志

国际口腔医学杂志 ›› 2026, Vol. 53 ›› Issue (5): 615-623.doi: 10.7518/gjkq.2026628

• 专家笔谈 •    

三维颅颌面软硬组织智能头影测量分析的研究进展

刘璐玮(),严斌()   

  1. 南京医科大学附属口腔医院正畸科 口腔疾病研究与防治国家级重点实验室培育建设点（南京医科 大学） 江苏省口腔转化医学工程研究中心（南京医科大学） 南京 210029
  • 收稿日期:2026-03-01 修回日期:2026-06-03 出版日期:2026-09-01 发布日期:2026-08-28
  • 通讯作者: 严斌
  • 作者简介:刘璐玮，博士，副教授，副主任医师，硕士生导师。现任南京医科大学附属口腔医院正畸科副主任，兼任中华口腔医学会正畸专业委员会青年委员、南京市口腔医学会正畸专业委员会委员。曾赴美国凯斯西储大学牙学院、美国TWEED正畸培训中心等院校交流学习。主持国家自然科学基金等科研、教学项目10余项。获第五届全国高校混合式教学设计创新大赛一等奖、江苏省医学引进新技术一等奖等奖项20余项，获国家专利4项、软件著作权8项，参编专著7部。在国内外学术期刊上以独立第一作者/通信作者发表论著20余篇。|刘璐玮，副教授，博士，Email：liuluwei_orth@njmu.edu.cn|严斌，博士，二级教授，主任医师，博士生导师。现任南京医科大学附属口腔医院党委书记，国际牙医师学院中国区院士、“国家高层次人才特殊支持计划”教学名师、国家级一流课程负责人、江苏省“333工程”二层次人才，兼任中华口腔医学会数字口腔医学专业委员会副主任委员、口腔正畸专业委员会常务委员、口腔医学教育专业委员会委员，江苏省口腔医学会正畸专业委员会主任委员、江苏省整形美容协会副会长和江苏省医学会数字医学分会副主任委员。曾赴美国俄亥俄州立大学和荷兰格罗宁根大学牙学院访学研修。主持国家自然科学基金、国家重点研发计划课题等省级以上课题10余项，主编国家“十四五”规划教材《数字化口腔医学》等教材及专著4部，以第一作者/通信作者发表学术论文80余篇，获发明专利授权24项（国际专利1项，已转化4项），获国家教学成果二等奖、江苏省科学技术二等奖等省级以上成果奖励4项。
  • 基金资助:
    国家重点研发计划课题(2023YFC2413605);江苏省科技重点研发计划国际合作项目(BZ2023042);全国医药学研究生在线课程建设与教学研究课题(2024-28)

Research progress in the intelligent cephalometric analysis of three-dimensional craniomaxillofacial soft and hard tissues

Luwei Liu(),Bin Yan()   

  1. Dept. of Orthodontics, Affiliated Stomatological Hospital of Nanjing Medical University; State Key Laboratory Cultivation Base of Research, Prevention and Treatment for Oral Diseases (Nanjing Medical University); Jiangsu Province Engineering Research Center of Stomatological Translational Medicine (Nanjing Medical University), Nanjing 210029，China
  • Received:2026-03-01 Revised:2026-06-03 Online:2026-09-01 Published:2026-08-28
  • Contact: Bin Yan
  • Supported by:
    National Key Research and Development Program of China(2023YFC2413605);International Cooperation Project of Jiangsu Science and Technology Key R&D Plan(BZ2023042);National Online Course Research Project for Graduate Students in Medicine and Pharmacy(2024-28)

RichHTML

6

PDF (PC)

18

摘要：

近年来，深度学习技术在医学影像分析领域取得了显著进展，颅颌面软硬组织自动定点及头影测量作为三维测量与分析的核心环节，在提高测量效率、降低观察者差异及支持个体化诊疗方面展现出重要潜力。本文回顾二维自动头影测量的发展历程，重点叙述基于深度学习的三维颅颌面软硬组织自动定点及头影测量方法的研究进展，分析三维自动定点在测量规范、数据依赖、泛化能力与临床可解释性等方面面临的主要挑战，对未来临床数据库建设、测量标准体系构建以及多模态多任务模型在临床决策支持中的应用前景进行展望，为三维颅颌面自动定点及头影测量技术的进一步研究与临床转化提供参考。

关键词: 三维头影测量, 颅颌面结构, 自动定点, 深度学习, 口腔正畸学

Abstract:

In recent years, deep learning has achieved remarkable advances in medical image analysis. Automated landmark identification and cephalometric analysis of craniofacial hard and soft tissues are key components of three-dimensional (3D) craniofacial assessment. These techniques have shown great potential for improving measurement efficiency, reducing interobserver variability, and supporting personalized diagnosis and treatment planning. This article reviews the development of two-dimensional automated cephalometric analysis and summarizes recent progress in deep learning-based methods for 3D automated landmark identification and cephalometric analysis of craniofacial hard and soft tissues. Major challenges are discussed, including the lack of standardized measurement protocols, dependence on large annotated datasets, limited model generalizability, and insufficient clinical interpretability. Future perspectives are also presented, with a focus on the establishment of large-scale clinical databases, the development of standardized measurement systems, and the application of multimodal and multitask models for clinical decision support. This review aims to provide a reference for future research and clinical translation of 3D automated craniofacial landmark identification and cephalometric analysis.

Key words: three-dimensional cephalometry, craniomaxillofacial structure, automatic landmark detection, deep learning, orthodontics

中图分类号: 

  • R783.5

图1

三维自动定点深度学习模型的总体框架"

表 1

2023?2025年间颅颌面三维数据自动定点测量的相关研究"

学者模型构架数据样本量

定位

对象

训练集样本量测试集样本量关键点数量定点准确性

Liu等［18］

改进3D U-Net＋体素热图回归

1 190例（800例SCT+390例CBCT；内部测试720例，外部测试/推理470例）

SCT＋CBCT

504例（SCT 336例+CBCT 168例；另设验证集72例）144例内部测试（SCT 96例+CBCT 48例）+ 470例外部测试/推理（SCT 320例+CBCT 150例）

SCT 41个点；CBCT 14个点

总体 MRE为1.22 mm±0.72 mm

Jiang等［25］3D CNN注意力回归网络＋直接回归3D坐标

498例

CBCT

-

-

43个点（颅骨5，上颌8，下颌9，牙槽及牙列10，软组织11）总体 MRE为1.76 mm±1.13 mm
Tang等［21］Mean-Teacher半监督框架+3D V-Net211例CBCT192例19例18个骨性标志点平均误差为1.91 mm±1.17 mm

Tanikawa等［26］

基于CBCT表面重建的3D网格+AI识别标志点

185例

CBCT152 例33 例64个点（颅骨19 个点+下颌骨45个点）上颌骨、下颌骨标志点平均定位误差分别为0.80 mm±0.57 mm、1.45 mm±0.34 mm
Sahlsten等［27］轻量级基于图像块的3D深度网络+热图回归309例（芬兰 199 例，泰国 110 例；训练/验证/测试：178/28/103例）CBCT178例103例46个点（骨性+牙性）芬兰、泰国队列平均定位误差分别为1.99 mm±1.55 mm、1.96 mm±1.25 mm
Deng等［28］深度学习61 例--61 例68个点（骨性+牙性+部分软组织点）MRE为2.3~2.4 mm
Tao等［29］两阶段3D卷积神经网络/U-Net80例（颌面畸形患者）SCT58例22例77个点（覆盖颅底、上颌、下颌及牙列）总体 MRE为1.81 mm± 0.89 mm
Nishimoto等［30］多阶段Deep learning networks公共数据库120例CT90例30例16个骨性标志点平均定位误差为2.88像素（约2.81 mm）
Blum等［19］三阶段的3D U-Net algorithm1 045例CBCT931例114例35个骨性标志点总体MRE为2.73 mm
Gillot等［31］深度强化学习143例中视野和大视野的CBCTCBCT--32个骨性标志点总体MRE为1.54 mm± 0.87 mm
Berends等［32］DiffusionNet算法2 897例三维颜面扫描2 463例434例10个软组织标志点总体MRE为1.69 mm± 1.15 mm
Al-Baker等［33］卷积神经网络408例三维图像生成10 200例PNG图像（训练/验证/测试：8 160/1 020/1 020例）三维颜面扫描8 160例PNG图像1 020例PNG图像37个软组织标志点总体MRE为0.83 mm±0.49 mm
[1] Durão AR, Pittayapat P, Rockenbach MIB, et al. Validity of 2D lateral cephalometry in orthodontics: a systematic review[J]. Prog Orthod, 2013, 14: 31.
[2] LeCun Y, Bengio Y, Hinton G. Deep learning[J]. Nature, 2015, 521(7553): 436-444.
[3] Broadbent BH. A new X-ray technique and its application to orthodontia[J]. Angle Orthod, 1931, 1(2): 45-66.
[4] Proffit WR, Fields HW, Sarver DM. Contemporary orthodontics[M]. 4th ed. New York: Elsevier Heal-th Sciences, 2007.
[5] Fracchia DE, Bignotti D, Lai S, et al. Reproducibility of AI in cephalometric landmark detection: a preliminary study[J]. Diagnostics, 2025, 15(19): 2521.
[6] Scarfe WC, Farman AG, Sukovic P. Clinical applications of cone-beam computed tomography in dental practice[J]. J Can Dent Assoc, 2006, 72(1): 75-80.
[7] Kapila S, Conley RS, Harrell WE Jr. The current status of cone beam computed tomography imaging in orthodontics[J]. Dentomaxillofac Radiol, 2011, 40(1): 24-34.
[8] Weinberg SM, Naidoo S, Govier DP, et al. Anthropometric precision and accuracy of digital three-dimensional photogrammetry[J]. J Craniofacial Surg, 2006, 17(3): 477-483.
[9] Nahm KY, Kim Y, Choi YS, et al. Accurate registration of cone-beam computed tomography scans to 3-dimensional facial photographs[J]. Am J Orthod Dentofacial Orthop, 2014, 145(2): 256-264.
[10] Bollen AM. Cephalometry in orthodontics: 2D and 3D[J]. Am J Orthod Dentofacial Orthop, 2019, 156(1): 161.
[11] Dot G, Schouman T, Chang S, et al. Automatic 3-dimensional cephalometric landmarking via deep lear-ning[J]. J Dent Res, 2022, 101(11): 1380-1387.
[12] Gwilliam JR, Cunningham SJ, Hutton T. Reprodu-cibility of soft tissue landmarks on three-dimensional facial scans[J]. Eur J Orthod, 2006, 28(5): 408-415.
[13] Perrotti G, Baccaglione G, Clauser T, et al. Total face approach (TFA) 3D cephalometry and superimposition in orthognathic surgery: evaluation of the vertical dimensions in a consecutive series[J]. Me-thods Protoc, 2021, 4(2): 36.
[14] Hsu SS, Gateno J, Bell RB, et al. Accuracy of a computer-aided surgical simulation protocol for orthognathic surgery: a prospective multicenter study[J]. J Oral Maxillofac Surg, 2013, 71(1): 128-142.
[15] Yang GJ, Xie F, Chen FH, et al. A comparative study of traditional and computer-aided surgical simulation guides in orthognathic correction of bimaxillary protrusion[J]. J Craniofacial Surg, 2025, 36(2): 553-557.
[16] Song C, Jeong Y, Huh H, et al. Multi-scale 3D ce-phalometric landmark detection based on direct regression with 3D CNN architectures[J]. Diagnostics, 2024, 14(22): 2605.
[17] Serafin M, Baldini B, Cabitza F, et al. Accuracy of automated 3D cephalometric landmarks by deep learning algorithms: systematic review and Meta-analysis[J]. Radiol Med, 2023, 128(5): 544-555.
[18] Liu B, Liu C, Xiong Y, et al. Accuracy and reliability of 3D cephalometric landmark detection with deep learning[J]. Eur J Med Res, 2025, 30(1): 1000.
[19] Blum FMS, Möhlhenrich SC, Raith S, et al. Evaluation of an artificial intelligence-based algorithm for automated localization of craniofacial landmarks[J]. Clin Oral Invest, 2023, 27(5): 2255-2265.
[20] Lee JH, Yu HJ, Kim MJ, et al. Automated cephalometric landmark detection with confidence regions using Bayesian convolutional neural networks[J]. BMC Oral Health, 2020, 20: 270.
[21] Tang HM, Liu S, Shi YX, et al. Automatic segmentation and landmark detection of 3D CBCT images using semi supervised learning for assisting orthognathic surgery planning[J]. Sci Rep, 2025, 15: 8814.
[22] Tao LR, Zhang X, Yang Y, et al. Craniomaxillofacial landmarks detection in CT scans with limited labeled data via semi-supervised learning[J]. Heliyon, 2024, 10(14): e34583.
[23] Lu G, Shu HZ, Bao H, et al. CMF-Net: craniomaxillofacial landmark localization on CBCT images u-sing geometric constraint and transformer[J]. Phys Med Biol, 2023, 68(9): 095020.
[24] Bao H, Zhang KJ, Yu CH, et al. Evaluating the accuracy of automated cephalometric analysis based on artificial intelligence[J]. BMC Oral Heal, 2023, 23: 191.
[25] Jiang Y, Jiang CY, Shi B, et al. Automatic identification of hard and soft tissue landmarks in cone-beam computed tomography via deep learning with diversity datasets: a methodological study[J]. BMC Oral Health, 2025, 25: 505.
[26] Tanikawa C, Nakamura H, Mimura T, et al. Deve-lopment of artificial intelligence-supported automa-tic three-dimensional surface cephalometry[J]. Orthod Craniofacial Res, 2025, 28(4): 636-646.
[27] Sahlsten J, Järnstedt J, Jaskari J, et al. Deep learning for 3D cephalometric landmarking with heterogeneous multi-center CBCT dataset[J]. PLoS One, 2024, 19(6): e0305947.
[28] Deng HH, Liu Q, Chen A, et al. Clinical feasibility of deep learning-based automatic head CBCT image segmentation and landmark detection in computer-aided surgical simulation for orthognathic surgery[J]. Int J Oral Maxillofac Surg, 2023, 52(7): 793-800.
[29] Tao L, Li M, Zhang X, et al. Automatic craniomaxillofacial landmarks detection in CT images of individuals with dentomaxillofacial deformities by a two-stage deep learning model[J]. BMC Oral Heal-th, 2023, 23(1): 876.
[30] Nishimoto S, Saito T, Ishise H, et al. Three-dimensional craniofacial landmark detection in series of CT slices using multi-phased regression networks[J]. Diagnostics, 2023, 13(11): 1930.
[31] Gillot M, Miranda F, Baquero B, et al. Automatic landmark identification in cone-beam computed tomography[J]. Orthod Craniofacial Res, 2023, 26(4): 560-567.
[32] Berends B, Bielevelt F, Schreurs R, et al. Fully automated landmarking and facial segmentation on 3D photographs[J]. Sci Rep, 2024, 14(1): 6463.
[33] Al-Baker B, Ayoub A, Ju XY, et al. Patch-based convolutional neural networks for automatic landmark detection of 3D facial images in clinical settings[J]. Eur J Orthod, 2024, 46(6): cjae056.
[34] Chong YM, Du FZ, Ma XD, et al. Automated anatomical landmark detection on 3D facial images u-sing U-NET-based deep learning algorithm[J]. Quant Imaging Med Surg, 2024, 14(3): 2466-2474.
[35] Wu YW, Zhao ZC, Lu JL, et al. A personalized automated system of 3D facial soft tissue landmarks annotation based on deep learning and computer vision[J]. BMC Oral Health, 2025, 26(1): 6.
[36] Baksi S, Freezer S, Matsumoto T, et al. Accuracy of an automated method of 3D soft tissue landmark detection[J]. Eur J Orthod, 2021, 43(6): 622-630.
[37] Weingart JV, Schlager S, Metzger MC, et al. Automated detection of cephalometric landmarks using deep neural patchworks[J]. Dentomaxillofac Radiol, 2023, 52(6): 20230059.
[38] Jindanil T, Burlacu-Vatamanu OE, Baldini B, et al. Automated orofacial virtual patient creation using two cohorts of MSCT vs. CBCT scans[J]. Head Face Med, 2025, 21(1): 21.
[39] Wang RH, Ho CT, Lin HH, et al. Three-dimensional cephalometry for orthognathic planning: normative data and analyses[J]. J Formos Med Assoc, 2020, 119(1 Pt 2): 191-203.
[40] Pittayapat P, Limchaichana-Bolstad N, Willems G, et al. Three-dimensional cephalometric analysis in orthodontics: a systematic review[J]. Orthod Craniofac Res, 2014, 17(2): 69-91.
[41] Muhammad D, Bendechache M. Unveiling the black box: a systematic review of Explainable Artificial Intelligence in medical image analysis[J]. Comput Struct Biotechnol J, 2024, 24: 542-560.
[42] Mathew MG, Jose S, Yadav S, et al. AI in dentistry: a legal minefield[J]. Br Dent J, 2025, 238(11): 838.
[1] 周文媛,范娟,熊再道,朱林,余泽正,王璐,金龙,张盼盼,顾永春. 卷积神经网络ResNet-101和Inception-v3对口腔曲面体层片中下颌第一磨牙远舌根变异的诊断[J]. 国际口腔医学杂志, 2026, 53(3): 328-334.
[2] 林超英,张岚,黄定明. 人工智能在根管治疗中的研究进展[J]. 国际口腔医学杂志, 2025, 52(5): 572-578.
[3] 周小洁,侯本祥. 基于深度学习技术诊断龋病方法的研究进展[J]. 国际口腔医学杂志, 2025, 52(5): 579-585.
[4] 汪云毅,朱珠,张峰. 人工智能在头影测量自动定点算法上的研究进展[J]. 国际口腔医学杂志, 2024, 51(5): 630-641.
[5] 陈新月,潘晓予,杨燕,贾媛媛,陈亮. 卷积神经网络在牙体牙髓病学中的应用进展[J]. 国际口腔医学杂志, 2024, 51(4): 483-488.
[6] 刘力嘉,毛婧,龙欢,蒲亚龙,王军. 二维与三维头影测量自动定点的研究进展[J]. 国际口腔医学杂志, 2022, 49(1): 100-108.
[7] 施丹妮,杨鑫,吴建勇. 锥形束CT三维头影测量参考坐标系的研究进展[J]. 国际口腔医学杂志, 2021, 48(4): 398-404.
[8] 王耀骏, 严斌, 王林. 牙周膜本构模型的研究进展[J]. 国际口腔医学杂志, 2017, 44(5): 538-543.
[9] 曹聪聪, 李精韬, 郑谦. 腭隐裂患者颅颌面生长发育的研究进展[J]. 国际口腔医学杂志, 2017, 44(4): 390-392.
[10] 宿洪丽. 口腔正畸学临床教学方法改革探讨[J]. 国际口腔医学杂志, 2017, 44(1): 41-44.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!