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

doi:10.13591/j.cnki.kqyx.2023.06.010

摘要/Abstract

摘要：

目的构建轻量型单步深度学习网络利用曲面断层片自动检测下颌智齿牙根与下颌管位置关系。方法将1 570例2 543颗同时拍摄曲面断层片和CBCT的成人下颌智齿病例,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评估下颌智齿牙根与下颌管的关系,分为非接触和接触。构建基于YOLO(You only look once)模型改良的轻量型单步深度学习网络算法模型(IAC-MTMnet),以曲面断层片作为输入端,以配对CBCT诊断作为金标准,评估下颌智齿牙根与下颌管的关系。诊断性能使用正确率、灵敏度、特异度、阳性预测值以及受试者工作曲线及曲线下面积表示。结果经CBCT诊断,下颌智齿牙根与下颌管接触的发生率为31.38%。IAC-MTMnet的诊断正确率为0.885,灵敏度为0.747,特异度为0.956,阳性预测值为0.899,受试者工作曲线下面积为0.95,测试运行时间为0.059 s。结论 IAC-MTMnet模型提升了曲面断层片诊断下颌智齿牙根与下颌管关系的性能。

关键词: 下颌智齿, 下颌管, 曲面断层片, 锥形束CT, 深度学习

Abstract:

Objective To develop a lite one-step deep learning network to detect the topographic proximity of the mandibular third molar(MTM)root to the inferior alveolar canal(IAC)on panoramic radiographs. Methods The samples, which consisted of 1 570 patients with 2 543 MTMs on paired panoramic radiographs and cone-beam computed tomography(CBCT), were randomly divided into the training group (80%), the validation group (10%), and the test group (10%). The evaluation of CBCT was defined as the ground truth. An extension of YOLO(You only look once)network, named as IAC-MTMnet, was trained to detect the proximity of MTM root to IAC on panoramic radiographs. Diagnostic performance analysis used accuracy, sensitivity, specificity, and positive predict value(PPV), and the area under the curve(AUC)was calculated based on the receiver operating characteristic(ROC)curve. Results On CBCT images, direct contact between MTM and IAC was observed on 798(31.38%)sides. The IAC-MTMnet achieved an accuracy of 0.885, a sensitivity of 0.747, a specificity of 0.956, and a PPV of 0.899. The AUC value achieved 0.95 and the test time was 0.059 s. Conclusion IAC-MTMnet is developed as a novel, robust and accurate method for detecting the proximity of MTM/IAC on panoramic radiographs.

Key words: mandibular third molar, inferior alveolar canal, panoramic radiography, CBCT, deep learning

中图分类号:

R814

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.

WANG Zhifan, DAI Xiubin, ZHOU Yanqi, MAO Tianyi, HUANG Hong, SONG Hongcheng, WANG Dongmiao. Automated detection of mandibular third molar root contacting with inferior alveolar canal on panoramic radiographs using a lite one-stage deep learning model[J]. Stomatology, 2023, 43(6): 534-539.

图/表 6

图1

图2

图3

表1

表2

图4

参考文献 29

[1]	Cheung LK, Leung YY, Chow LK, et al. Incidence of neurosensory deficits and recovery after lower third molar surgery: A prospective clinical study of 4 338 cases[J]. Int J Oral Maxillofac Surg, 2010, 39(4):320-326. doi: 10.1016/j.ijom.2009.11.010
[2]	Hasegawa T, Ri S, Shigeta T, et al. Risk factors associated with inferior alveolar nerve injury after extraction of the mandibular third molar—A comparative study of preoperative images by panoramic radiography and computed tomography[J]. Int J Oral Maxillofac Surg, 2013, 42(7):843-851.
[3]	Leung YY. Management and prevention of third molar surgery-related trigeminal nerve injury: Time for a rethink[J]. J Korean Assoc Oral Maxillofac Surg, 2019, 45(5):233-240. doi: 10.5125/jkaoms.2019.45.5.233 pmid: 31728330
[4]	Kang F, Sah MK, Fei G. Determining the risk relationship associated withinferior alveolar nerve injury following removal of mandibular third molar teeth: A systematic review[J]. J Stomatol Oral Maxillofac Surg, 2020, 121(1):63-69. doi: S2468-7855(19)30200-9 pmid: 31476533
[5]	Ghaeminia H, Meijer GJ, Soehardi A, et al. Position of the impacted third molar in relation to the mandibular canal. Diagnostic accuracy of cone beam computed tomography compared with panoramic radiography[J]. Int J Oral Maxillofac Surg, 2009, 38(9):964-971. doi: 10.1016/j.ijom.2009.06.007
[6]	Rood JP, Shehab BA. The radiological prediction of inferior alveolar nerve injury during third molar surgery[J]. Br J Oral Maxillofac Surg, 1990, 28(1):20-25. doi: 10.1016/0266-4356(90)90005-6
[7]	Su NC, van Wijk A, Berkhout E, et al. Predictive value of panoramic radiography for injury of inferior alveolar nerve after mandibular third molar surgery[J]. J Oral Maxillofac Surg, 2017, 75(4):663-679. doi: 10.1016/j.joms.2016.12.013
[8]	Baqain ZH, AlHadidi A, AbuKaraky A, et al. Does the use of cone-beam computed tomography before mandibular third molar surgery impact treatment planning?[J]. J Oral Maxillofac Surg, 2020, 78(7):1071-1077. doi: 10.1016/j.joms.2020.03.002
[9]	Matzen LH, Berkhout E. Cone beam CT imaging of the mandibular third molar: A position paper prepared by the European Academy of Dento Maxillo Facial Radiology (EADMFR)[J]. Dentomaxillofac Radiol, 2019, 48(5):20190039. doi: 10.1259/dmfr.20190039
[10]	Ohashi Y, Ariji Y, Katsumata A, et al. Utilization of computer-aided detection system in diagnosing unilateral maxillary sinusitis on panoramic radiographs[J]. Dentomaxillofac Radiol, 2016, 45(3):20150419. doi: 10.1259/dmfr.20150419
[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]	Lee JH, Kim DH, Jeong SN. Diagnosis of cystic lesions using panoramic and cone beam computed tomographic images based on deep learning neural network[J]. Oral Dis, 2020, 26(1):152-158. doi: 10.1111/odi.v26.1
[13]	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. doi: 10.1259/dmfr.20180218
[14]	Vinayahalingam S, Xi T, Bergé S, et al. Automated detection of third molars and mandibular nerve by deep learning[J]. Sci Rep, 2019, 9(1):9007. doi: 10.1038/s41598-019-45487-3 pmid: 31227772
[15]	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. doi: 10.1016/j.oooo.2020.04.005
[16]	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. doi: 10.1016/j.postharvbio.2021.111808
[17]	Bochkovskiy A, Wang CY, Liao HY M. YOLOv4: Optimal speed and accuracy of object detection[EB/OL]. 2020: arXiv: 2004. 10934[2023-03-13]. https://arxiv.org/abs/2004.10934.
[18]	Szalma J, Vajta L, Lovász BV, et al. Identification of specific panoramic high-risk signs in impacted third molar cases in which cone beam computed tomography changes the treatment decision[J]. J Oral Maxillofac Surg, 2020, 78(7):1061-1070. doi: 10.1016/j.joms.2020.03.012
[19]	谢理哲, 樊瑜波. 深度学习技术在口腔医学中的应用研究[J]. 口腔医学, 2022, 42(1):8-13.
[20]	Kuwada C, Ariji Y, Fukuda M, et al. Deep learning systems for detecting and classifying the presence of impacted supernumerary teeth in the maxillary incisor region on panoramic radiographs[J]. Oral Surg Oral Med Oral Pathol Oral Radiol, 2020, 130(4):464-469. doi: 10.1016/j.oooo.2020.04.813
[21]	Chang HJ, Lee SJ, Yong TH, et al. Deep learning hybrid method to automatically diagnose periodontal bone loss and stage periodontitis[J]. Sci Rep, 2020, 10(1):7531. doi: 10.1038/s41598-020-64509-z
[22]	Krois J, Ekert T, Meinhold L, et al. Deep learning for the radiographic detection of periodontal bone loss[J]. Sci Rep, 2019, 9(1):8495. doi: 10.1038/s41598-019-44839-3 pmid: 31186466
[23]	Ekert T, Krois J, Meinhold L, et al. Deep learning for the radiographic detection of apical lesions[J]. J Endod, 2019, 45(7):917-922.e5. doi: S0099-2399(19)30256-0 pmid: 31160078
[24]	Ariji Y, Yanashita Y, Kutsuna S, et al. Automatic detection and classification of radiolucent lesions in the mandible on panoramic radiographs using a deep learning object detection technique[J]. Oral Surg Oral Med Oral Pathol Oral Radiol, 2019, 128(4):424-430. doi: 10.1016/j.oooo.2019.05.014
[25]	Zhu TE, Chen DQ, Wu FL, et al. Artificial intelligence model to detect real contact relationship between mandibular third molars and inferior alveolar nerve based on panoramic radiographs[J]. Diagnostics (Basel), 2021, 11(9):1664.
[26]	Reia VCB, de Toledo Telles-Araujo G, Peralta-Mamani M, et al. Diagnostic accuracy of CBCT compared to panoramic radiography in predicting IAN exposure: A systematic review and meta-analysis[J]. Clin Oral Invest, 2021, 25(8):4721-4733. doi: 10.1007/s00784-021-03942-4
[27]	Wang DM, Lin TY, Wang YL, et al. Radiographic features of anatomic relationship between impacted third molar and inferior alveolar canal on coronal CBCT images: Risk factors for nerve injury after tooth extraction[J]. Arch Med Sci, 2018, 14(3):532-540. doi: 10.5114/aoms.2016.58842 pmid: 29765439
[28]	Ueda M, Nakamori K, Shiratori K, et al. Clinical significance of computed tomographic assessment and anatomic features of the inferior alveolar canal as risk factors for injury of the inferior alveolar nerve at third molar surgery[J]. J Oral Maxillofac Surg, 2012, 70(3):514-520. doi: 10.1016/j.joms.2011.08.021
[29]	Liu MQ, Xu ZN, Mao WY, et al. Deep learning-based evaluation of the relationship between mandibular third molar and mandibular canal on CBCT[J]. Clin Oral Invest, 2022, 26(1):981-991. doi: 10.1007/s00784-021-04082-5

信息项目	n(%)
性别
男性	667(42.48)
女性	903(57.52)
阻生侧
右侧	321(20.45)
左侧	276(17.58)
双侧	973(61.97)
CBCT诊断(金标准)
接触	798(31.38)
未接触	1 745(68.62)
曲面断层片诊断
接触	1 095(43.06)
未接触	1 448(56.94)
曲面断层片特征性影像
一个或多个	779(71.14)
无	316(28.86)

方法	正确率	灵敏度	特异度	阳性预测值	检测时间/s
IAC-MTMnet	0.885	0.747	0.956	0.899	0.059
YOLOv5	0.881	0.819	0.913	0.829	219.8
人工评估	0.845	0.741	0.892	0.759	-