Deep Learning-based Point Cloud Completion Network for Restoration of Calcified Root Canals in Maxillary Molars

doi:10.13701/j.cnki.kqyxyj.2023.05.015

Abstract

Abstract: Objective: To train a deep learning-based point cloud completion network by simulated calcified root canal samples for restoration of calcified canals in upper and middle thirds in maxillary molars. Methods: Two hundred maxillary molars (with complete pulp cavities and three root canals) were collected and scanned by Micro-CT to obtain 200 samples of 3D model of the pulp. The pulp morphology was edited using Mimics software to simulate the calcified root canals in upper and middle thirds of maxillary molars. The paired data of complete pulp and simulated calcified pulp were used as training samples for the point cloud completion network. A dataset containing 200 samples was produced, which was divided into 180 sets as the training set and the remaining 20 sets as test set. In the testing stage, the output results were grid processed through the reverse engineering software, and then compared with the ground truth for morphology evaluation. Results: The DSC between the prediction model and the ground truth was (86.80±1.32)%, the ASSD was (0.04±0.02) mm, and the HD was (0.28±0.05) mm. Conclusion: The predicted results are highly similar with the ground truth and the deviation of shape difference is small. The deep learning-based point cloud completion network can restore the calcified canals in upper and middle thirds in maxillary molars with good completion effect. It can provide a pre-theoretical basis for subsequent artificial intelligence-assisted calcified root canal treatment.

Key words: dental pulp calcification, point cloud completion, deep learning, the maxillary molar

WEN Jiahuan, FU Yujie, REN Genqiang, CHEN Yufei, ZHANG Qi. Deep Learning-based Point Cloud Completion Network for Restoration of Calcified Root Canals in Maxillary Molars[J]. Journal of Oral Science Research, 2023, 39(5): 455-459.

References

[1] 凌均棨.显微牙髓治疗学[M].北京:人民卫生出版社,2014.91-92.
[2] Fleig S, Attin T, Jungbluth H. Narrowing of the radicular pulp space in coronally restored teeth [J]. Clin Oral Investig, 2017, 21(4):1251-1257.
[3] Carvalho TS, Lussi A. Age-related morphological, histological and functional changes in teeth [J]. J Oral Rehabil, 2017, 44(4):291-298.
[4] Hargreaves KM, Berman LH, Rotstein I, et al. Cohen's pathways of the pulp [M], 11th ed. St. Louis: Elsevier, 2015. 49-50.
[5] Wu D, Shi W, Wu J, et al. The clinical treatment of complicated root canal therapy with the aid of a dental operating microscope [J]. Int Dent J, 2011, 61(5): 261-266.
[6] 史瑞棠,侯本祥.牙髓钙化的病因、诊断和治疗策略[J].中华口腔医学杂志,2022,57(3):220-226.
[7] McCabe PS, Dummer PM. Pulp canal obliteration: an endodontic diagnosis and treatment challenge [J]. Int Endod J, 2012, 45(2):177-197.
[8] Yang Y, Guo B, Guo L, et al. CBCT-aided microscopic and ultrasonic treatment for upper or middle thirds calcified root canals [J]. Biomed Res Int, 2016, 2016:4793146.
[9] Matherne RP, Angelopoulos C, Kulild JC, et al. Use of cone-beam computed tomography to identify root canal systems in vitro [J]. J Endod, 2008, 34(1): 87-89.
[10] van der Meer WJ, Vissink A, Ng YL, et al. 3D computer aided treatment planning in endodontics [J]. J Dent, 2016, 45:67-72.
[11] Lin X, Fu Y, Ren G, et al. Micro-computed tomography-guided artificial intelligence for pulp cavity and tooth segmentation on cone-beam computed tomography [J]. J Endod, 2021, 47(12):1933-1941.
[12] Fei B, Yang W, Chen W, et al. Comprehensive review of deep learning-based 3D point cloud completion processing and analysis [J]. IEEE trans Intell Transp Syst, 2022, 23(12):22862-22883.
[13] Yuan W, Khot T, Held D, et al. PCN: Point completion network[C]//2018 International Conference on 3D Vision (3DV), Verona, Italy:IEEE, 2018: 728-737.
[14] Huang Z, Yu Y, Xu J, et al. PF-Net: Point fractal network for 3D point cloud completion [C]//IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Seattle, USA:IEEE, 2020:7659-7667.
[15] 李美佳,于泽宽,刘晓,等.点云算法在医学领域的研究进展[J].中国图象图形学报,2020,25(10): 2013-2023.
[16] Qi CR, Su H, Mo K, et al. PointNet: Deep learning on point sets for 3D classification and segmentation [C]//IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Honolulu, USA:IEEE, 2017:77-85.
[17] Xi L, Zhao Y, Chen L, et al. Recovering dense 3D point clouds from single endoscopic image [J]. Comput Methods Programs Biomed, 2021, 205:106077.
[18] Ma W, Yang X, Wang Q, et al. Multi-scope feature extraction for intracranial aneurysm 3D point cloud completion [J]. Cells, 2022, 11(24):4107.
[19] Valentini V, Boldrini L, Damiani A, et al. Recommendations on how to establish evidence from auto-segmentation software in radiotherapy [J]. Radiother Oncol, 2014, 112(3):317-320.
[20] Buchgreitz J, Buchgreitz M, Mortensen D, et al. Guided access cavity preparation using cone beam computed tomography and optical surface scans-an ex vivo study [J]. Int Endod J, 2016, 49(8):790-795.
[21] Nayak A, Jain PK, Kankar PK, et al. Computer-aided design-based guided endodontic: A novel approach for root canal access cavity preparation [J]. Proc Inst Mech Eng H, 2018, 232(8):787-795.
[22] Fonseca Tavares WL, de Oliveira Murta Pedrosa N, Moreira RA, et al. Limitations and management of static-guided endodontics failure [J]. J Endod, 2022, 48(2):273-279.