Improved RA-U-Net for Dental Image Segmentation
DOI:
https://doi.org/10.12974/2311-8695.2025.13.06Keywords:
RA-U-Net, Residual Module, Attention Mechanism, Dental Image SegmentationAbstract
The diversity of tooth morphology, irregular arrangement, and blurred grayscale characteristics of the alveolar bone present significant challenges in image recognition and segmentation. While deep learning-based U-Net networks have achieved considerable progress in medical image segmentation, this paper proposes the RA-U-Net architecture, an enhanced version of U-Net, achieving high-precision segmentation of multimodal dental images. To address challenges in dental image segmentation, the U-Net architecture is enhanced by incorporating residual modules and attention mechanisms. A group normalization strategy is applied for feature map channel segmentation, improving deep network learning capabilities and model generalization. The network was tested on a dental CBCT image dataset from the Shanghai Ninth People's Hospital. Results demonstrated an average Dice coefficient of 0.839, surpassing other image segmentation networks, with a precision of 0.961 and a recall of 0.719. These findings indicate that the RA-U-Net model delivers superior image segmentation performance, providing reliable support for dental diagnosis and treatment.
References
Ronneberger O, Fischer P, Brox T. U-Net: Convolutional Networks for Biomedical Image Segmentation[C]// International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer. 2015: 234-241. https://doi.org/10.1007/978-3-319-24574-4_28
DU Wenjie, Wang Yuanjun. Improved RA-U-Net network for liver tumor segmentation of CT images [J]. Progress in Biomedical Engineering, 2025, (5).
Kaiming He, Xiangyu Zhang, Shaoqing Ren, etc. Deep Residual Learning for Image Recognition[C]. CVPR IEEE Xplore, 2015,12,5.
Dong X Y, Chen P. Segmentation of liver tumors based on bottleneck residual attention mechanism U-net [J]. CT Theory and Applications, 2021, 30(6): 661-670.
Xu P Q, Liang Y X, LI Y. Medical image segmentation fusing multi-scale semantic and residual bottleneck attention[J]. Computer Engineering, 2023, 49(1 0): 1 62-1 70.
Oktay O, Schlemper J, Folgoc LL, et al. Attention U-Net: Learning where to look for the pancreas. arXiv: 1804. 03999, 2018.
HE Bin, JIN Yongjie, LI Yulan. Defining nuclear medical file formal based on DICOM standard [J]. Nuclear Electronics & Detection Technology, 2001, 21(6): 440-443.
Reza A M. Realization of the Contrast Limited Adaptive Histogram Equalization (CLAHE) for Real-Time Image Enhancement [J]. Journal of Vlsi Signal Processing, 2004, 38(1): 35-44. https://doi.org/10.1023/B:VLSI.0000028532.53893.82
PAN Zhenkuan, WEI Weibo, ZHANG Haitao. Variational models for image diffusion based on gradient and Laplacian [J]. Journal of Shandong University (Natural Science), 2008, 43(11): 11-16.
