深度学习在脊柱质心定位与分割的应用进展

doi:10.16180/j.cnki.issn1007-7820.2023.11.005

摘要/Abstract

摘要：

脊柱医学图像质心定位与椎骨分割在脊柱手术引导中具有重要意义,准确定位脊柱质心与分割椎骨已成为重要的研究课题。近年来,随着GPU算力的提高以及医学图像数据的累积,深度学习在脊柱图像中的应用取得突破。为研究深度学习在脊柱医学图像定位与分割任务中的应用现状与发展,文中对该领域近几年脊柱定位与分割模型进行整理与研究。收集脊柱常用数据集与评价指标并探讨深度学习模型在脊柱质心定位与分割的应用,分析模型实现过程及存在的不足之处。文中还针对目前深度学习在脊柱质心定位与分割应用面临的问题给出应对策略,并提出未来可行的发展方向。

关键词: 脊柱图像, 脊柱质心定位, UNet, 椎骨分割, 深度学习, 上下文信息, CT图像, MR图像

Abstract:

Spinal centroid localization and vertebral segmentation have great significance in the guidance of spinal surgery. Accurately locate the spinal centroid and segment vertebrae has become an important research topic.In recent years, with the improvement of GPU computing power and the accumulation of medical image data, the application of deep learning in spinal imaging has made a major breakthrough. In order to study the application status and development of deep learning in the task of spinal medical image localization and segmentation, this study storts out and studies the models of spinal localization and segmentation in this field in recent years. This study collects the commonly used data sets and evaluation indexes of the spine, discusses the application of a deep learning model in spinal centroid location and segmentation, and analyzes the realization process and shortcomings of the model. This study also outlines the countermeasures for the problems faced by the current application of deep learning in spinal centroid location and segmentation and outlines the feasible development direction in the future.

Key words: spinal images, spinal centroid localization, UNet, vertebra segmentation, deep learning, contextual information, CT image, MR image

中图分类号:

TP391

孙红,莫光萍,徐广辉,杨晨. 深度学习在脊柱质心定位与分割的应用进展[J]. 电子科技, 2023, 36(11): 28-34.

SUN Hong,MO Guangping,XU Guanghui,YANG Chen. Advances in Application of Deep Learning in Centroid Localization and Vertebrae Segmentation of Spine[J]. Electronic Science and Technology, 2023, 36(11): 28-34.

图/表 4

图1

表1

表2

表3

参考文献 29

[1]	孔鸣, 何前锋, 李兰娟. 人工智能辅助诊疗发展现状与战略研究[J]. 中国工程科学, 2018, 20(2):86-91. doi: 10.15302/J-SSCAE-2018.02.013
	Kong Ming, He Qianfeng, Li Lanjuan. AI assisted clinical diagnosis & treatment and development strategy[J]. Strategic Study of CAE, 2018, 20(2):86-91. doi: 10.15302/J-SSCAE-2018.02.013
[2]	尹梓名, 孙大运, 胡晓晖, 等. 人工智能在骨质疏松症中的应用研究综述[J]. 小型微型计算机系统, 2019, 40(9):1839-1850.
	Yin Ziming, Sun Dayun, Hu Xiaohui, et al. Review on the application of artificial intelligence for osteoporosis[J]. Journal of Chinese Computer Systems, 2019, 40(9):1839-1850.
[3]	Löffler M T, Sekuboyina A, Jacob A, et al. A vertebral segmentation dataset with fracture grading[J]. Radiology:Artificial Intelligence, 2020, 2(4):138-146.
[4]	Hesamian M H, Jia W, He X, et al. Deep learning techniques for medical image segmentation: Achievements and challenges[J]. Journal of Digital Imaging, 2019, 32(4):582-596. doi: 10.1007/s10278-019-00227-x pmid: 31144149
[5]	马金林, 邓媛媛, 马自萍. 肝脏肿瘤CT图像深度学习分割方法综述[J]. 中国图象图形学报, 2020, 25(10):2024-2046.
	Ma Jinlin, Deng Yuanyuan, Ma Ziping. Review of deep learning segmentation methods for CT images of liver tumors[J]. Journal of Image and Graphics, 2020, 25(10):2024-2046.
[6]	闫超, 孙占全, 田恩刚, 等. 基于深度学习的医学图像分割技术研究进展[J]. 电子科技, 2021, 34(2):7-11.
	Yan Chao, Sun Zhanquan, Tian Engang, et al. Research progress of medical image segmentation based on deep learning[J]. Electronic Science and Technology, 2021, 34(2):7-11.
[7]	Sekuboyina A, Husseini M E, Bayat A, et al. VerSe:A vertebrae labelling and segmentation benchmark for multi-detector CT images[J]. Medical Image Analysis, 2021, 73(10):2166-2171.
[8]	Korez R, Ibragimov B, Likar B, et al. A framework for automated spine and vertebrae interpolation-based detection and model-based segmentation[J]. IEEE Transactions on Medical Imaging, 2015, 34(8):1649-1662. doi: 10.1109/TMI.2015.2389334 pmid: 25585415
[9]	Yao J, Burns J E, Munoz H, et al. Detection of vertebral body fractures based on cortical shell unwrapping[C]. Heidelberg: International Conference on Medical ImageComputing and Computer-Assisted Intervention, 2012:509-516.
[10]	Glocker B, Feulner J, Criminisi A, et al. Automatic localization and identification of vertebrae in arbitrary field-of-view CT scans[C]. Heidelberg: International Conference on Medical Image Computing and Computer-Assisted Intervention, 2012:590-598.
[11]	Jakubicek R, Chmelik J, Jan J, et al. Learning-based vertebra localization and labeling in 3D CT data of possibly incomplete and pathological spines[J]. Computer Methods and Programs in Biomedicine, 2020, 183(C):105081-105087.
[12]	Chen J, Wang Y, Guo R, et al. Lsrc:A long-short range context-fusing framework for automatic 3D vertebra localization[C]. Cham: International Conference on Medical Image Computing and Computer-Assisted Intervention, 2019:95-103.
[13]	Payer C, Stern D, Bischof H, et al. Coarse to fine vertebrae localization and segmentation with spatial Configuration-Net and U-Net[C]. Valletta: Proceedings of the Fifteenth International Joint Conference on Computer Vision,Imaging and Computer Graphics Theory and Applications, 2020:124-133.
[14]	Qin C, Yao D, Zhuang H, et al. Residual block-based multi-label classification and localization network with integral regression for vertebrae labeling[EB/OL].(2020-01-01)[2022-06-23]https://arxiv.org/abs/2001.0017.
[15]	Wang F, Zheng K, Lu L, et al. Automatic vertebra localization and identification in CT by spine rectification and anatomically-constrained optimization[C]. Nashville: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021:5280-5288.
[16]	Liao H, Mesfin A, Luo J. Joint vertebrae identification and localization in spinal CT images by combining short-and long-range contextual information[J]. IEEE Transactions on Medical Imaging, 2018, 37(5):1266-1275. doi: 10.1109/TMI.2018.2798293
[17]	McCouat J, Glocker B. Vertebrae detection and localization in CT with two-stage CNNS and dense annotateions[EB/OL].(201910-14)[2022-06-23]https://arxiv.org/abs/1910.05911.
[18]	Levine M, De Silva T, Ketcha M D, et al. Automatic vertebrae localization in spine CT:A deep-learning approach for image guidance and surgical data science[C]. San Diego: Image-Guided Procedures,Robotic Interventions and Modeling.International Society for Optics and Photonics, 2019:3651-3670.
[19]	Sekuboyina A, Rempfler M, Kukaĉka J, et al. Btrfly net:Vertebrae labelling with energy-based adversarial learning of local spine prior[C]. Cham: International Conference on Medical Image Computing and Computer-Assisted Intervention, 2018:649-657.
[20]	Yang D, Xiong T, Xu D. Automatic vertebra labeling in large-scale medical images using deep image-to-image network with message passing and sparsity regularization[C]. Cham: Deep Learning and Convolutional Neural Networks for Medical Imaging and Clinical Informatics, 2019:179-197.
[21]	Shi D, Pan Y, Liu C, et al. Automatic localization and segmentation of vertebral bodies in 3D CT volumes with deep learning[C]. New York: Proceedings of the Second International Symposium on Image Computingand Digital Medicine, 2018:42-46.
[22]	Rak M, Steffen J, Meyer A, et al. Combining convolutional neural networks and star convex cuts for fast whole spine vertebra segmentation in MRI[J]. Compute Methods and Programs in Biomedicine, 2019, 177(8):47-56.
[23]	Cheng P, Yang Y, Yu H, et al. Automatic vertebrae localization and segmentation in CT with a two-stage Dense-U-Net[J]. Scientific Reports, 2021, 11(1):1-13. doi: 10.1038/s41598-020-79139-8
[24]	Bae H J, Hyun H, Byeon Y, et al. Fully automated 3D segmentation and separation of multiple cervical vertebrae in CT images using a 2D convolutional neural networks[J]. Computer Methods and Programs in Biomedicine, 2020, 184(3):105119-105131. doi: 10.1016/j.cmpb.2019.105119
[25]	Lessmann N, Van G B, De P A, et al. Iterative fully convolutional neural networks for automatic vertebra segmentation and identification[J]. Medical Image Analysis, 2019, 53(4):142-155. doi: 10.1016/j.media.2019.02.005
[26]	Chuang C H, Lin C Y, Tsai Y Y, et al. Efficient triple output network for vertebral segmentation and identification[J]. IEEE Access, 2019, 7(2):117978-117985. doi: 10.1109/Access.6287639
[27]	Janssens R, Zeng G, Zheng G. Fully automatic segmentation of lumbar vertebrae from CT images using cascaded 3D fully convolutional networks[C]. Washington D.C.: IEEE the Fifteenth International Symposium on Biomedical Imaging, 2018:893-897.
[28]	Xia L, Xiao L, Quan G, et al. 3D cascaded convolutional networks for multi-vertebrae segmentation[J]. Current Medical Imaging, 2020, 16(3):231-240. doi: 10.2174/1573405615666181204151943 pmid: 32133953
[29]	Zareie M, Parsaei H, Amiri S, et al. Automatic segmentation of vertebrae in 3D CT images using adaptive fast 3D pulse coupled neural networks[J]. Australasian Physical & Engineering Sciences in Medicine, 2018, 41(4):1009-1020.

指标	名称	作用
id.rate	识别率	衡量模型预测椎骨坐标的准确率
d_mean	定位距离	衡量定位局部的准确性
d_std	定位误差	衡量定位误差
Dice	Dice系数	用于衡量两集合间的相似度
HD	Hausdorff距离	衡量分割任务中两组点集的相似性
ASD	平均表面距离	求集合中所有点的表面距离的平均

方法	应用数据集	d_mean/mm	d_st/mm	id.rate/%
文献[11]	CSI-Label 2014	5.08	3.95	90.90
文献[12]	CSI-Label 2014	7.10	7.70	88.00
文献[13]	VerSe2019	5.71	6.28	89.79
文献[14]	CSI-Label 2014	2.90	5.80	89.00
文献[15]	CSI-Label 2014	2.55	1.40	97.40
文献[16]	CSI-Label 2014	6.47	8.56	88.30
文献[17]	CSI-Label 2014	5.60	7.10	85.80
文献[19]	CSI-Label 2014	6.20	4.10	88.50
文献[20]	CSI-Label 2014	8.60	7.80	85.00

方法	数据集	分割段	距离/mm	Dice/%
文献[13]	VerSe2019	全脊柱	6.53^H	94.00
文献[22]	whole-spine	全脊柱	1.06^A(4.06^H)	93.80
文献[22]	whole-spine	胸腰椎	0.79^A(3.85^H)	96.00
文献[23]	xVertSeg	腰椎	NULL	87.70
文献[23]	CSI-Label2014	胸腰椎	NULL	95.30
文献[24]	SGSH	颈椎	20.89^A(0.33^H)	94.37
文献[24]	SNUBH	颈椎	16.23^A(0.39^H)	96.23
文献[26]	xVertSeg	腰椎	NULL	88.46
文献[27]	xVertSeg	腰椎	0.40^A	95.70
文献[28]	CSI-Label2014	胸腰椎	0.32^A	94.00