Derin öğrenme ile panoramik radyografi görüntülerinden diş segmentasyonu: UNet, FPN ve PSPNet karşılaştırması

Year 2024, , 1347 - 1354, 15.10.2024

Salih Taha Alperen Özçelik , Hüseyin Üzen , Abdülkadir Şengür , Adalet Çelebi

https://doi.org/10.28948/ngumuh.1497540

Abstract

Panoramik radyografi, diş hastalıklarının erken teşhis ve tedavisinde yaygın olarak kullanılan bir araçtır. Ancak bu tekniklerin başarısı, diş hekiminin deneyimine ve doğru yorum yapma yeteneğine bağlıdır. Bu çalışmada, insan hatasını azaltmak ve hızlı çözümler sunmak amacıyla panoramik radyografi görüntülerinden derin öğrenme ile diş segmentasyonu yapılmıştır. 18-65 yaş arasındaki hastalardan alınan 313 panoramik radyografi görüntüsü, deneyimli bir ağız ve çene cerrahı tarafından piksel seviyesinde etiketlenmiştir. UNet, FPN ve PSPNet mimarileri ile segmentasyon yapılmış, VGG16 omurga ağı ile UNet modeli %93,74 F1 skoru, %88.22 KüB ve %98.25 doğruluk ile en iyi sonucu vermiştir. Ayrıca EfficientNet, ResNet50, InceptionV3, DenseNet121 ve MobileNet omurga ağları ile denemeler yapılmış, en yüksek performans EfficientNet omurga ağı ile %93.91 F1 skoru, %88.52 KüB ve %98.26 doğruluk olarak elde edilmiştir.

Keywords

Derin Öğrenme, Diş segmentasyonu, UNet

Tooth segmentation from panoramic radiography images with deep learning: comparison of UNet, FPN and PSPNet

Abstract

Panoramic radiography is a widely used tool for the early diagnosis and treatment of dental diseases. However, the success of these techniques depends on the dentist's experience and ability to interpret correctly. In this study, dental segmentation was performed using deep learning on panoramic radiography images to reduce human error and provide faster solutions. A total of 313 panoramic radiographs from patients aged 18-65 were pixel-wise labeled by an experienced oral and maxillofacial surgeon. Segmentation was performed using UNet, FPN, and PSPNet architectures, and the UNet model with a VGG16 backbone achieved the best result with a 93.74% F1 score, 88.22% IoU, and 98.25% accuracy. Additionally, tests were conducted with EfficientNet, ResNet50, InceptionV3, DenseNet121, and MobileNet backbones, with EfficientNet showing the highest performance with a 93.91% F1 score, 88.52% IoU, and 98.26% accuracy.

Keywords

Deep Learning, Tooth Segmentation, UNet

References

• S.A. Gill, Quinonez R.B., M. Deutchman, C.E. Conklin, D. Rizzolo, D. Rabago, H. Silk. Integrating Oral Health into Health Professions School Curricula. Medical Education Online, 27 (1). 2022. https://doi.org/10.1080/10872981.2022.2090308.
• Oral health: A window to your overall health - Mayo Clinic. (n.d.). Retrieved May 26, 2024, from https://www.mayoclinic.org/healthy-lifestyle/adult-he alth/in-depth/dental/art-20047475
• WHO. Oral health. Retrieved December 25, 2023, from https://www.who.int/news-room/fact-sheets/detail/ora l-health
• R. Izzetti, , M. Nisi, G. Aringhieri, L. Crocetti, F. Graziani and C. Nardi. Basic knowledge and new advances in panoramic radiography ımaging techniques: a narrative review on what dentists and radiologists should know. Applied Sciences, 11 (17), 7858. 2021. https://doi.org/10.3390/APP1117 7858
• H. Zhang, and Y. Qie. Applying deep learning to medical ımaging: a review. Applied Sciences 13 (18), 10521. 2023. https://doi.org/10.3390/APP131810521
• A.B. Oktay, and A. Gurses. Detection, segmentation, and numbering of teeth in dental panoramic images with mask regions with convolutional neural network features. State of the Art in Neural Networks and Their Applications: 1, 73–90. 2021. https://doi.org/10.101 6/B978-0-12-819740-0.00004-8.
• Y. Yang, R. Xie, W. Jia, Z. Chen, Y. Yang, L. Xie and B.X. Jiang. Accurate and automatic tooth image segmentation model with deep convolutional neural networks and level set method. Neurocomputing, 419, 108–125. 2021. https://doi.org/10.1016/J.NEUCO M.2020.07.110
• M.K. Alam, T. Haque, F. Akhter, H.N. Albagieh, A. Bin Nabhan, M.A. Alsenani, S. Islam. Teeth segmentation by optical radiographic images using VGG-16 deep learning convolution architecture with R-CNN network approach for biomedical sensing applications. Optical and Quantum Electronics, 55(9), 1–19. 2023. https://doi.org/10.1007/S11082-023-05096-X/TABLES/4.
• G. Rubiu, M. Bologna, M. Cellina, M. Cè, D. Sala, R. Pagani, … M. Alì. Teeth segmentation in panoramic dental x-ray using mask regional convolutional neural network. Applied Sciences 2023, Vol. 13, Page 7947, 13 (13), 7947. 2023. https://doi.org/10.3390 /APP13137947.
• M. Xu, Y. Wu, Z. Xu, P. Ding, H. Bai and X. Deng. Robust automated teeth identification from dental radiographs using deep learning. Journal of Dentistry, 136, 104607. 2023. https://doi.org/10.1016/J.JDEN T.2023.104607.
• Prados-Privado, M., J. García Villalón, A. Blázquez Torres, C.H. Martínez-Martínez and C. Ivorra. A convolutional neural network for automatic tooth numbering in panoramic ımages. BioMed Research International, 2021. https://doi.org/10.1155/2021/36 25386.
• A.E. Yüksel, S. Gültekin, E. Simsar, Ş.D. Özdemir, M. Gündoğar, S.B. Tokgöz and İ.E. Hamamcı. Dental enumeration and multiple treatment detection on panoramic X-rays using deep learning. Scientific Reports, 11(1), 1–10. 2021. https://doi.org/10.1038/s 41598-021-90386-1
• S.Y. Lin, and H.Y. Chang. Tooth numbering and condition recognition on dental panoramic radiograph ımages using CNNs. IEEE Access, 9, 166008–166026. 2021. https://doi.org/10.1109/ACCE SS.2021.3136026.
• S. Tian, N. Dai, B. Zhang, F. Yuan, Q. Yu and X. Cheng. Automatic classification and segmentation of teeth on 3D dental model using hierarchical deep learning networks. IEEE Access, 7, 84817–84828. 2019. https://doi.org/10.1109/ACCESS.2019.2924262.
• A. Jana, H.M. Subhash and D. Metaxas. Automatic tooth segmentation from 3D dental model using deep learning: a quantitative analysis of what can be learnt from a single 3D dental model, 12567, 42–51. 2023. https://doi.org/10.1117/12.2669716.
• H. Wang, J. Minnema, K.J. Batenburg, T. Forouzanfar, F.J. Hu and G. Wu. Multiclass CBCT image segmentation for orthodontics with deep learning. Journal of Dental Research, 100 (9), 943–949. 2021. https://doi.org/10.1177/00220345211005338.
• C. Szegedy, V. Vanhoucke, S. Ioffe, J. Shlens and Z. Wojna. Rethinking the inception architecture for Computer Vision. 2016.
• K. He, X. Zhang, S. Ren and J. Sun. Deep residual learning for ımage recognition. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 770–778. 2015. https://doi.org/10.1109/CVPR.2016.90.
• C. Szegedy, V. Vanhoucke, S. Ioffe, J. Shlens and Z. Wojna. Rethinking the inception architecture for computer vision. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2818–2826. 2016. https://doi.org/10. 1109/CVPR.2016.308.
• G. Huang, Z. Liu, L. Van Der Maaten and K.Q. Weinberger. Densely connected convolutional networks. Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2261–2269. 2017. https://doi.org/10.1109/CVPR.2017 .243
• M. Sandler, A. Howard, M. Zhu, A. Zhmoginov and L.C. Chen. MobileNetV2: Inverted residuals and linear bottlenecks. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 4510–4520. 2018. https://doi.org/10 .1109/CVPR.2018.00474
• T.Y. Lin, P. Dollár, R. Girshick, K. He, B. Hariharan and S. Belongie. Feature pyramid networks for object detection. 2016. Retrieved from https://arxiv. org/abs/1612.03144v2
• H. Zhao, J. Shi, X. Qi, X. Wang and J. Jia. Pyramid scene parsing network. proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR, 6230–6239. 2017. https://doi.org/ 10.1109/CVPR.2017.660.
• O. Ronneberger, P. Fischer and T. Brox. U-net: Convolutional networks for biomedical image segmentation. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 9351, 234–241. 2015. https://doi.org/10.1007/978-3-319-24574-4_28/COVER.
• K. Simonyan, and A. Zisserman. Very deep convolutional networks for large-scale image recognition. 3rd International Conference on Learning Representations, ICLR 2015 - Conference Track Proceedings. Retrieved from, 2014. https://arxiv. org/abs/1409.1556v6
• K. Becker, É. Da, S. Rocha and P.T. Endo. A comparative study of deep learning models for dental segmentation in panoramic radiograph. Applied Sciences 12 (6), 3103. 2022. https://doi.org /10.3390/APP12063103