Journal of the Korea Computer Graphics Society (한국컴퓨터그래픽스학회논문지)

Korea Computer Graphics Society (한국컴퓨터그래픽스학회)
권호 표지
DOI QR코드

DOI QR Code

Orbital Bone Segmentation using Improved Skip Connection of U-Net Structure in Facial CT Images

안면부 CT 영상에서 U-Net 구조의 개선된 스킵 연결을 사용한 안와 분할

  • Jinseo An (Dept. of Software Convergence, Seoul Women's University) ;
  • Min Jin Lee (Dept. of Software Convergence, Seoul Women's University) ;
  • Kyu Won Shim (Dept. of Pediatric Neurosurgery, Severance Children's Hospital) ;
  • Helen Hong (Dept. of Software Convergence, Seoul Women's University)
  • 안진서 (서울여자대학교 소프트웨어융합학과) ;
  • 이민진 (서울여자대학교 소프트웨어융합학과) ;
  • 심규원 (연세대학교 의과대학 신경외과학교실 소아신경외과) ;
  • 홍헬렌 (서울여자대학교 소프트웨어융합학과)
  • Received : 2022.12.21
  • Accepted : 2023.04.05
  • Published : 2023.06.01
Download PDF
⟨ Previous Next ⟩

Abstract

Orbital bones are composed of thick cortical bone with high intensity values and thin bones with low intensity values, making consistent segmentation difficult. In addition, the medial wall and the orbital floor are composed of thin bones, making it difficult to distinguish the intensity values from surrounding tissues due to the partial volume effect. In this paper, we propose MSDA-Net to improve segmentation performance by considering the anatomical structure of orbital bones with various thickness and the characteristics of thin bones with low intensity values and small areas in facial CT images. By applying a multi-scale module and a dual attention module that performs channel and spatial attention sequentially to the skip connection of U-Net, a feature map emphasizing the features to be paid attention to is delivered to the decoder. The paper presents the results of experiments that evaluate the effect of the multi-scale hierarchical module, single attention, and dual attention on segmentation performance. When using the proposed method, the Dice similarity coefficient (DSC) of the global and regional evaluation regions shows excellent performance with 92%, 86%, and 87%, respectively.

안와 뼈는 두껍고 높은 밝기값의 피질골과 매우 얇고 낮은 밝기값의 얇은 뼈로 이루어져 있어 일관된 분할이 어렵다. 또한 안와 내측벽과 하벽은 뼈의 두께가 얇아 부분용적효과로 인해 주변 연조직과 밝기값 구분이 어렵다는 한계가 있다. 본 논문은 안면부 CT 영상에서 다양한 두께를 갖는 안와 뼈의 해부학적 구조와 낮은 밝기값과 작은 영역을 갖는 얇은 뼈의 특성을 고려하여 분할 성능을 개선하기 위해 MSDA-Net을 제안한다. U-Net의 스킵 연결 부분에 다중 스케일 모듈과 채널 및 공간 어텐션을 순차적으로 수행하는 듀얼 어텐션 모듈을 함께 적용하여 주의 집중할 특징을 강조한 특징 맵을 디코더에 전달한다. 실험을 통해 다중 스케일 계층 모듈의 효과와 싱글 어텐션과 듀얼 어텐션 효과를 제시하였고, 제안 방법 사용 시 전역적 및 지역적 평가 영역의 DSC가 92%, 86%, 87%로 우수한 성능을 보였다.

Keywords

Acknowledgement

본 논문은 서울여자대학교 학술연구비의 지원(2023-0111) 및 보건복지부의 재원으로 한국보건산업진흥원의 보건의료기술 연구개발사업 지원에 의하여 이루어진 것임(과제고유번호 : HI22C1496).

References

  1. C. Rene, "Update on orbital anatomy," Eye, 20(10), pp. 1119-1129, 2006. https://doi.org/10.1038/sj.eye.6702376
  2. R. Schreurs, C. Klop, and T.J. Maal, "Advanced Diagnostics and Three-dimensional Virtual Surgical Planning in Orbital Reconstruction," Atlas of the oral and maxillofacial surgery clinics of North America, 29(1), pp. 79-96, 2021. https://doi.org/10.1016/j.cxom.2020.11.003
  3. H. Kim, T.G. Son, J. Lee, H.A. Kim, H. Cho, W.S. Jeong, et al, "Three-dimensional orbital wall modeling using paranasal sinus segmentation," Journal of Cranio-Maxillofacial Surgery, 47(6), pp. 959-967, 2019. https://doi.org/10.1016/j.jcms.2019.03.028
  4. G. Choi, J. Shin, J. Kyung, M. Kyung, and Y. Lee, "Data Augmentation Method for Deep Learning based Medical Image Segmentation Model," Journal of the Korea Computer Graphics Society, 25(3), pp. 123-131, 2019. https://doi.org/10.15701/kcgs.2019.25.3.123
  5. H. Pak, D. Kim, and Y.G. Shin, "Automatic Sagittal Plane Detection for the Identification of the Mandibular Canal," Journal of the Korea Computer Graphics Society, 26(3), pp. 31-37, 2020. https://doi.org/10.15701/kcgs.2020.26.3.31
  6. S. Ma, G. Ahn, and H. Hong, "Chest CT Image Patch-Based CNN Classification and Visualization for Predicting Recurrence of Non-Small Cell Lung Cancer Patients," Journal of the Korea Computer Graphics Society, 28(1), pp. 1-9, 2022. https://doi.org/10.15701/kcgs.2022.28.1.1
  7. Y. Jang, J. Yoo, and H. Hong, "Assessment and Analysis of Fidelity and Diversity for GAN-based Medical Image Generative Model," Journal of the Korea Computer Graphics Society, 28(2), pp. 11-19, 2022. https://doi.org/10.15701/kcgs.2022.28.2.11
  8. M.J. Lee, H. Hong, K.W. Shim, and S. Park, "MGB-NET: Orbital Bone Segmentation from Head and Neck CT Images Using Multi-Graylevel-Bone Convolutional Networks," IEEE 16th International Symposium on Biomedical Imaging (ISBI), 2019.
  9. M.J. Lee, H. Hong, K.W. Shim, and S. Park, "Orbital bone segmentation in head and neck CT images using multi-gray level fully convolutional networks," Medical Imaging 2019: Image Processing, SPIE, pp. 869-874, 2019.
  10. J. An, M.J. Lee, H. Hong, and K.W. Shim, "Effect of inter-slice resolution improvement for orbital bone segmentation of facial 3D CT images with different slice thicknesses," Poster presented at the annual meeting for the Radiological Society of North America(RSNA), 2022.
  11. J. Hamwood, B. Schmutz, M.J. Collins, M.C. Allenby, and D. Alongso-Caneiro, "A deep learning method for automatic segmentation of the bony orbit in MRI and CT images," Scientific Reports, 11(1), pp.1-12, 2021. https://doi.org/10.1038/s41598-020-79139-8
  12. O. Ronneberger, P. Fischer, and T. Borx, "U-net: Convolutional networks for biomedical image segmentation," International Conference on Medical image computing and computer-assisted intervention, pp. 234-241, Springer, Cham, 2015.
  13. S.H. Gao, M.M. Cheng, K. Zhao, X.Y. Zhang, M.H. Yang, and P. Torr, "Res2net: A new multi-scale backbone architecture," IEEE transactions on pattern analysis and machine intelligence, 43(2), pp. 652-662, 2019. https://doi.org/10.1109/TPAMI.2019.2938758
  14. S. Woo, J. Park, J.Y. Lee, and I.S. Kweon, "Cbam: Convolutional block attention module," Proceedings of the European conference on computer vision(ECCV), pp. 3-19, 2018.