In order to meet the clinical need for automatic calculation of Cobb angle in scoliosis disease, it is necessary to segment each vertebra of the spine. In this paper, a vertebra segmentation model combining HRNet and VP-UNet was proposed to achieve accurate vertebra segmentation and automatic calculation of Cobb angle. Firstly, the bone direction loss function (BD-Loss) was designed, and the spinal posture was used as a prior knowledge to guide the training of HRNet, so as to improve the adaptability of HRNet to the complex spinal morphology. Secondly, a position information perception module (PIPM) was proposed to integrate HRNet localization features into VP-UNet segmentation network to obtain multi-level information of spinal image. VP-UNet added Dropout layer and dense unit on the basis of VGG-Net to extract more global image information while reducing parameters. Finally, based on the results of vertebral localization and segmentation, a method of spine type judgment and Cobb angle calculation was proposed. The experimental results show that compared with VGG-Net, the improved segmentation model has improved Recall by 2.18 percentage points, Precision by 1.7 percentage points, Dice by 0.69 percentage points and IoU by 3.09 percentage points, which can accurately locate and segment each vertebra. The calculated results of Cobb angle meet the clinical needs and provide the technical support for the diagnosis of scoliosis.
HUANGRen, WANGXing, LIZhijun, et al. Research progress in diagnosis and treatment of adolescent idiopathic scoliosis[J]. Chinese Journal of Clinical Anatomy, 2016, 34(4): 472-475. (in Chinese)
[3]
MASOODR F, TAJI A, KHANM B, et al. Deep learning based vertebral body segmentation with extraction of spinal measurements and disorder disease classification[J]. Biomedical Signal Processing and Control, 2022, 71: 103230.
JIANGBaihao, LIUJing, QIUDawei, et al. Review of deep learning applications in spinal image segmentation[J]. Computer Engineering, 2024, 50(3): 1-15. (in Chinese)
[6]
LINY, ZHOUH Y, MAK, et al. Seg4Reg networks for automated spinal curvature estimation[C]//Computational Methods and Clinical Applications for Spine Imaging, 2020: 69-74.
[7]
CHENB, XUQ, WANGL, et al. An automated and accurate spine curve analysis system[J]. IEEE Access, 2019, 7: 124596-124605.
[8]
HUOL, CAIB, LIANGP, et al. Joint spinal centerline extraction and curvature estimation with row-wise classification and curve graph network[C]//Medical Image Computing and Computer Assisted Intervention, 2021: 377-386.
[9]
ZOUL, GUOL, ZHANGR, et al. VLTENet: A deep-learning-based vertebra localization and tilt estimation network for automatic Cobb angle estimation[J]. IEEE Journal of Biomedical and Health Informatics, 2023, 27(6): 3002-3013.
[10]
RAHMANIARW, SUZUKIK, LINT L. Auto-CA: Automated Cobb angle measurement based on vertebrae detection for assessment of spinal curvature deformity[J]. IEEE Transactions on Bio-Medical Engineering, 2024, 71(2): 640-649.
[11]
MAEDAY, NAGURAT, NAKAMURAM, et al. Automatic measurement of the Cobb angle for adolescent idiopathic scoliosis using convolutional neural network[J]. Scientific Reports, 2023, 13(1): 14576.
[12]
CAESARENDRAW, RAHMANIARW, MATHEWJ, et al. Automated Cobb angle measurement for adolescent idiopathic scoliosis using convolutional neural network[J]. Diagnostics, 2022, 12(2): 396.
[13]
SUNY, XINGY, ZHAOZ, et al. Comparison of manual versus automated measurement of Cobb angle in idiopathic scoliosis based on a deep learning keypoint detection technology[J]. European Spine Journal, 2022, 31(8): 1969-1978.
[14]
HORNGM H, KUOKC P, FUM J, et al. Cobb angle measurement of spine from x-ray images using convolutional neural network[J]. Computational and Mathematical Methods in Medicine, 2019, 2019(1): 1-18.
[15]
PANGS, PANGC, ZHAOL, et al. SpineParseNet: spine parsing for volumetric MR image by a two-stage segmentation framework with semantic image representation[J]. IEEE Transactions on Medical Imaging, 2021, 40(1): 262-273.
[16]
CHENGP, YANGY, YUH, et al. Automatic vertebrae localization and segmentation in CT with a two-stage Dense-U-Net[J]. Scientific Reports, 2021, 11(1): 22156.
ZHAOYang, ZHANGJunhua. Multi-scale feature fusion method for spinal X-ray lmage segmentation[J]. Computer Engineering and Applications, 2021, 57(8): 214-219. (in Chinese)
[22]
RAK M, STEFFENJ, MEYERA, et al. Combining convolutional neural networks and star convex cuts for fast whole spine vertebra segmentation in MRI[J]. Computer Methods and Programs in Biomedicine, 2019, 177: 47-56.
[23]
WUY, NAMDARK, CHENC, et al. Automated adolescence scoliosis detection using augmented U-net with non-square kernels[J]. Canadian Association of Radiologists Journal, 2023, 74(4): 667-675.
[24]
PANGS, PANGC, SUZ, et al. DGMSNet: Spine segmentation for MR image by a detection-guided mixed-supervised segmentation network[J]. Medical Image Analysis, 2022, 75: 102261.
[25]
ARIF S M M RAL, KNAPPK, SLABAUGHG. Fully automatic cervical vertebrae segmentation framework for X-ray images[J]. Computer Methods and Programs in Biomedicine, 2018, 157: 95-111.
[26]
SUNK, XIAOB, LIUD, et al. Deep high-resolution representation learning for human pose estimation[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 5693-5703.
[27]
RONNEBERGERO, FISCHERP, BROXT. U-net: Convolutional networks for biomedical image segmentation[C]//Medical Image Computing and Computer-Assisted Intervention, 2015: 234-241.
[28]
CHENL C, ZHUY, PAPANDREOUG, et al. Encoder-decoder with atrous separable convolution for semantic image segmentation[C]//Proceedings of the European Conference on Computer Vision (ECCV), 2018: 801-818.
[29]
ZHANGH, LUG, ZHANM, et al. Semi-supervised classification of graph convolutional networks with Laplacian rank constraints[J]. Neural Processing Letters, 2022, 54(4): 2645-2656.