基于注意力机制的深度学习网络自动识别垂直位下颌第三磨牙牙根数目的研究

doi:10.13591/j.cnki.kqyx.2024.11.006

口腔医学 ›› 2024, Vol. 44 ›› Issue (11): 831-836.doi: 10.13591/j.cnki.kqyx.2024.11.006

基于注意力机制的深度学习网络自动识别垂直位下颌第三磨牙牙根数目的研究

孙春生^1,^2,³,戴修斌⁴,周曼婷⁵,景秋平^1,^2,³,张驰^1,^2,³,阳胜军^1,^2,³,王东苗^1,^2,³()

1 南京医科大学附属口腔医院口腔颌面外科,江苏南京(210029)
2 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),江苏南京(210029)
3 江苏省口腔转化医学工程研究中心,江苏南京(210029)
4 南京邮电大学自动化学院,江苏南京(210023)
5 南京邮电大学通信与信息工程学院,江苏南京(210023)

收稿日期:2024-01-22 出版日期:2024-11-28 发布日期:2024-11-18
通讯作者: 王东苗 Tel:(025) 69593091 E-mail:wdm9921@njmu.edu.cn
基金资助:
江苏省卫健委科研项目(M2020021);江苏高校优势学科建设工程(PAPD);江苏高校优势学科建设工程(2018-87);江苏省科教能力提升工程——江苏省研究型医院(YJXYYJSDW4);,江苏省医学创新中心(CXZX202227);江苏省口腔转化医学工程研究中心开放课题(GCZX2023-04);中国牙病防治基金会科研项目(JKLKMNLOL);江苏省重点研发计划(BE2023833)

Automatic assessment of root numbers of vertical mandibular third molar using a deep learning model based on attention mechanism

SUN Chunsheng^1,^2,³,DAI Xiubin⁴,ZHOU Manting⁵,JING Qiuping^1,^2,³,ZHANG Chi^1,^2,³,YANG Shengjun^1,^2,³,WANG Dongmiao^1,^2,³()

Department of Oral and Maxillofacial Surgery, The Affiliated Stomatological Hospital of Nanjing Medical University, Nanjing 210029, China

Received:2024-01-22 Online:2024-11-28 Published:2024-11-18

摘要/Abstract

摘要：

目的构建基于注意力机制的深度学习网络,利用曲面断层片实现垂直位下颌第三磨牙牙根数目(单根或双根)的自动识别。方法将1 045例1 642颗成人垂直位下颌第三磨牙病例同时拍摄曲面断层片和CBCT,随机分成训练组(80%)、验证组(10%)和测试组(10%)。分别使用曲面断层片和CBCT评估其牙根数目,分为单根和双根。构建基于注意力机制的深度学习网络模型(RN-MTMnet),以曲面断层片作为输入,以配对CBCT诊断作为金标准,自动定位垂直位下颌第三磨牙所在区域并判断其牙根数目(单根或双根)。RN-MTMnet模型性能使用正确率、灵敏度、特异度、阳性预测值,以及ROC曲线及对应AUC面积值衡量,并与Faster-RCNN、CenterNet及SSD等网络模型以及人工诊断相比较。结果经CBCT诊断,纳入病例单根336(20.46%)侧,双根1 306(79.54%)侧。RN-MTMnet获得结果的正确率为0.888,灵敏度为0.885,特异度为0.903,阳性预测值为0.976,受试者工作曲线下面积为0.90。结论 RN-MTMnet模型提升了曲面断层片诊断垂直位下颌第三磨牙牙根数目的性能。

关键词: 下颌第三磨牙, 曲面断层片, CBCT, 深度学习, 注意力机制

Abstract:

Objective To develop a deep learning network based on attention mechanism to identify the number of the vertical mandibular third molar(MTM) roots(single or double) on panoramic radiographs in an automatic way. Methods The sample consisted of 1 045 patients with 1 642 MTMs on paired panoramic radiographs and Cone-beam computed tomography(CBCT) and were randomly grouped into the training(80%), the validation(10%), and the test(10%). The evaluation of CBCT was defined as the ground truth. A deep learning network based on attention mechanism, which was named as RN-MTMnet, was trained to judge if the MTM on panoramic radiographs had one or two roots. Diagnostic performance was evaluated by accuracy, sensitivity, specificity, and positive predict value(PPV), and the receiver operating characteristic(ROC)curve with the area under the ROC curve(AUC). Its diagnostic performance was compared with dentists’ diagnosis, Faster-RCNN, CenterNet, and SSD using evaluation metrics. Results On CBCT images, single-rooted MTM was observed on 336(20.46%) sides, while two-rooted MTM was 1 306(79.54%). The RN-MTMnet achieved an accuracy of 0.888, a sensitivity of 0.885, a specificity of 0.903, a PPV of 0.976, and the AUC value of 0.90. Conclusion RN-MTMnet is developed as a novel, robust and accurate method for detecting the numberof MTM roots on panoramic radiographs.

Key words: mandibular third molar, panoramic radiography, CBCT, deep learning, attention mechanism

中图分类号:

R782.1

孙春生, 戴修斌, 周曼婷, 景秋平, 张驰, 阳胜军, 王东苗. 基于注意力机制的深度学习网络自动识别垂直位下颌第三磨牙牙根数目的研究[J]. 口腔医学, 2024, 44(11): 831-836.

SUN Chunsheng, DAI Xiubin, ZHOU Manting, JING Qiuping, ZHANG Chi, YANG Shengjun, WANG Dongmiao. Automatic assessment of root numbers of vertical mandibular third molar using a deep learning model based on attention mechanism[J]. Stomatology, 2024, 44(11): 831-836.

图/表 5

图1

图2

表1

图3

图4

参考文献 23

[1]	Carvalho RWF, do Egito Vasconcelos BC. Assessment of factors associated with surgical difficulty during removal of impacted lower third molars[J]. J Oral Maxillofac Surg, 2011, 69(11): 2714-2721.
[2]	Jing QP, Song HC, Huang H, et al. Characterizations of three-dimensional root morphology and topological location of mandibular third molars by cone-beam computed tomography[J]. Surg Radiol Anat, 2023, 45(5): 527-536. doi: 10.1007/s00276-023-03111-0 pmid: 36884060
[3]	da Silva Sampieri MB, Viana FLP, Cardoso CL, et al. Radiographic study of mandibular third molars: Evaluation of the position and root anatomy in Brazilian population[J]. Oral Maxillofac Surg, 2018, 22(2): 163-168. doi: 10.1007/s10006-018-0685-y pmid: 29450664
[4]	Carter K, Worthington S. Predictors of third molar impaction: A systematic review and meta-analysis[J]. J Dent Res, 2016, 95(3): 267-276. doi: 10.1177/0022034515615857 pmid: 26561441
[5]	Bailey E, Kashbour W, Shah N, et al. Surgical techniques for the removal of mandibular wisdom teeth[J]. Cochrane Database Syst Rev, 2020, 7(7): CD004345.
[6]	Hauge Matzen L, Christensen J, Hintze H, et al. Diagnostic accuracy of panoramic radiography, stereo-scanography and cone beam CT for assessment of mandibular third molars before surgery[J]. Acta Odontol Scand, 2013, 71(6): 1391-1398. doi: 10.3109/00016357.2013.764574 pmid: 23356838
[7]	Matzen LH, Berkhout E. Cone beam CT imaging of the mandibular third molar: A position paper prepared by the European Academy of DentoMaxilloFacial Radiology (EADMFR)[J]. Dentomaxillofac Radiol, 2019, 48(5): 20190039.
[8]	Shan T, Tay FR, Gu L. Application of artificial intelligence in dentistry[J]. J Dent Res, 2021, 100(3): 232-244. doi: 10.1177/0022034520969115 pmid: 33118431
[9]	Hung K, Montalvao C, Tanaka R, et al. The use and performance of artificial intelligence applications in dental and maxillofacial radiology: A systematic review[J]. Dentomaxillofac Radiol, 2020, 49(1): 20190107.
[10]	Hiraiwa T, Ariji Y, Fukuda M, et al. A deep-learning artificial intelligence system for assessment of root morphology of the mandibular first molar on panoramic radiography[J]. Dentomaxillofac Radiol, 2019, 48(3): 20180218.
[11]	Fukuda M, Inamoto K, Shibata N, et al. Evaluation of an artificial intelligence system for detecting vertical root fracture on panoramic radiography[J]. Oral Radiol, 2020, 36(4): 337-343. doi: 10.1007/s11282-019-00409-x pmid: 31535278
[12]	Kuwana R, Ariji Y, Fukuda M, et al. Performance of deep learning object detection technology in the detection and diagnosis of maxillary sinus lesions on panoramic radiographs[J]. Dentomaxillofac Radiol, 2021, 50(1): 20200171.
[13]	Lee JH, Kim DH, Jeong SN. Diagnosis of cystic lesions usingpanoramic and cone beam computed tomographic images based on deep learning neural network[J]. Oral Dis, 2020, 26(1): 152-158.
[14]	Wang ZP, Jin LY, Wang S, et al. Apple stem/calyx real-time recognition using YOLO-v5 algorithm for fruit automatic loading system[J]. Postharvest Biol Technol, 2022, 185: 111808.
[15]	Ren SQ, He KM, Girshick R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[J]. IEEE Trans Pattern Anal Mach Intell, 2017, 39(6): 1137-1149.
[16]	Duan KW, Bai S, Xie LX, et al. CenterNet: Keypoint triplets for object detection[C]// 2019 IEEE/CVF International Conference on Computer Vision (ICCV).Seoul, Korea (South). IEEE, 2019: 6568-6577.
[17]	Liu W, Anguelov D, Erhan D, et al. SSD: single shot MultiBox detector[C]// European Conference on Computer Vision. Cham: Springer, 2016: 21-37.
[18]	Yuasa H, Kawai T, Sugiura M. Classification of surgical difficulty in extracting impacted third molars[J]. Br J Oral Maxillofac Surg, 2002, 40(1): 26-31.
[19]	Akadiri OA, Obiechina AE. Assessment of difficulty in third molar surgery: A systematic review[J]. J Oral Maxillofac Surg, 2009, 67(4): 771-774.
[20]	Zhang WY, Tang Y, Liu C, et al. Root and root canal variations of the human maxillary and mandibular third molars in a Chinese population: A micro-computed tomographic study[J]. Arch Oral Biol, 2018, 95: 134-140. doi: S0003-9969(18)30435-7 pmid: 30107301
[21]	Carvalho JS, Lotz M, Rubi L, et al. Preinterventional third-molar assessment using robust machine learning[J]. J Dent Res, 2023, 102(13): 1452-1459.
[22]	Choi E, Lee S, Jeong E, et al. Artificial intelligence in positioning between mandibular third molar and inferior alveolar nerve on panoramic radiography[J]. Sci Rep, 2022, 12(1): 2456. doi: 10.1038/s41598-022-06483-2 pmid: 35165342
[23]	Fukuda M, Ariji Y, Kise Y, et al. Comparison of 3 deep learning neural networks for classifying the relationship between the mandibular third molar and the mandibular canal on panoramic radiographs[J]. Oral Surg Oral Med Oral Pathol Oral Radiol, 2020, 130(3): 336-343.

网络模型	正确率	灵敏度	特异度	阳性预测值
人工诊断	0.792	0.789	0.793	0.495
Faster-RCNN	0.786	0.837	0.559	0.895
CenterNet	0.834	0.897	0.581	0.897
SSD	0.750	0.795	0.516	0.895
RN-MTMnet	0.888	0.885	0.903	0.976

基于注意力机制的深度学习网络自动识别垂直位下颌第三磨牙牙根数目的研究

Automatic assessment of root numbers of vertical mandibular third molar using a deep learning model based on attention mechanism

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 5

参考文献 23

相关文章 15

编辑推荐

Metrics

本文评价

[1]	孙皖苏, 吴辰一, 查楚萍, 洪礼琳, 张亨国. 基于CBCT骨微参数与骨阻力分析辅助下颌第三磨牙微创拔除[J]. 口腔医学, 2024, 44(9): 663-669.
[2]	李根, 郭松松, 蔡观慧, 孙莲, 孙雯, 王华. 骨性Ⅲ类患者双颌正颌手术后上气道形态变化与颌骨移动量的相关性研究[J]. 口腔医学, 2024, 44(7): 515-521.
[3]	吴玥, 唐雯, 张语嫣然, 袁玮煜, 潘逸菲, 陈新宇, 徐海洋, 吕云帆, IZADIKHAH Iman, 曹丹, 谢理哲, 严斌. 基于CBCT三维形态分析的中国青少年颈椎骨成熟度定量方法研究[J]. 口腔医学, 2024, 44(5): 321-328.
[4]	翟家彬, 吴雯丽, 沈铭, 陈盈盈. 下颌切牙区种植相关影像学测量及分析研究[J]. 口腔医学, 2024, 44(4): 250-254.
[5]	熊再道,张燕萍,顾永春,周志浩. 无托槽隐形矫治器掩饰治疗成人骨性Ⅲ类错𬌗的牙根吸收研究[J]. 口腔医学, 2024, 44(11): 847-580.
[6]	李宗峰, 孙莲, 潘永初. 不同矢状骨面型成人下颌磨牙远中舌侧间隙的CBCT研究[J]. 口腔医学, 2024, 44(11): 820-823.
[7]	王振婷, 李北, 季意, 张娟, 杨晨, 单华, 刘锐. 成年人翼钩及其周围组织的CBCT影像学研究[J]. 口腔医学, 2024, 44(10): 728-733.
[8]	周举, 马俊青. 腭正中旁区支持组织厚度和腭穹窿形态相关性的三维评估[J]. 口腔医学, 2024, 44(1): 36-40.
[9]	陈云, 曹灵, 王雅欣, 于剑南. 815颗儿童上颌多生牙的回顾性分析[J]. 口腔医学, 2023, 43(8): 717-721.
[10]	黄虹, 李媛, 景秋平, 石逸瑾, 阳胜军, 王东苗. 青少年正畸患者下颌第三磨牙萌出的影响因素研究[J]. 口腔医学, 2023, 43(8): 697-701.
[11]	王芷凡, 戴修斌, 周炎锜, 冒添逸, 黄虹, 宋洪丞, 王东苗. 轻量型单步深度学习网络自动识别下颌智齿牙根与下颌管位置关系的研究[J]. 口腔医学, 2023, 43(6): 534-539.
[12]	刘金明,吴雯丽,杜令仪,沈铭. 跨下牙槽神经种植相关解剖结构的影像学测量及分析研究[J]. 口腔医学, 2023, 43(2): 130-134.
[13]	雷巧玲,何文丹,雷期音,熊培颖,王忠东. 微型骨钉辅助上颌快速扩弓对下颌位置的影响[J]. 口腔医学, 2023, 43(2): 135-140.
[14]	宋洪丞, 黄虹, 王芷凡, 景秋平, 王东苗. 下颌阻生智齿相关第二磨牙牙根外吸收:曲面断层片与CBCT对比研究[J]. 口腔医学, 2023, 43(1): 70-74.
[15]	唐诗雨, 艾力麦尔旦·艾尼瓦尔, 马廷林, 苏庆玲, 王珊, 王玲. 基于CBCT的上颌第三磨牙拔除术与上颌窦穿孔的相关性分析[J]. 口腔医学, 2022, 42(4): 327-331.