SpineParseNet: Spine Parsing for Volumetric MR Image by a Two-Stage Segmentation Framework With Semantic Image Representation
The DOI matches its Crossref record; 35 of the 35 references it cites were found in a registry or an index.
Abstract
Spine parsing (i.e., multi-class segmentation of vertebrae and intervertebral discs (IVDs)) for volumetric magnetic resonance (MR) image plays a significant role in various spinal disease diagnoses and treatments of spine disorders, yet is still a challenge due to the inter-class similarity and intra-class variation of spine images. Existing fully convolutional network based methods failed to explicitly exploit the dependencies between different spinal structures. In this article, we propose a novel two-stage framework named SpineParseNet to achieve automated spine parsing for volumetric MR images. The SpineParseNet consists of a 3D graph convolutional segmentation network (GCSN) for 3D coarse segmentation and a 2D residual U-Net (ResUNet) for 2D segmentation refinement. In 3D GCSN, region pooling is employed to project the image representation to graph representation, in which each node representation denotes a specific spinal structure. The adjacency matrix of the graph is designed according to the connection of spinal structures. The graph representation is evolved by graph convolutions. Subsequently, the proposed region unpooling module re-projects the evolved graph representation to a semantic image representation, which facilitates the 3D GCSN to generate reliable coarse segmentation. Finally, the 2D ResUNet refines the segmentation. Experiments on T2-weighted volumetric MR images of 215 subjects show that SpineParseNet achieves impressive performance with mean Dice similarity coefficients of 87.32 ± 4.75%, 87.78 ± 4.64%, and 87.49 ± 3.81% for the segmentations of 10 vertebrae, 9 IVDs, and all 19 spinal structures respectively. The proposed method has great potential in clinical spinal disease diagnoses and treatments.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 1×
“With the depth of CNN increasing, degradation of the deep network appears, which can be avoided by the residual network (ResNet) [23].”
BackgroundHochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.8.1735 · Cited by 102,318 · cited 1×
“Introducing long short-term memory (LSTM) [26] into the segmentation networks is another approach for extracting semantic image representation.”
BackgroundU-Net: Convolutional Networks for Biomedical Image Segmentation
Ronneberger, O., Fischer, P., & Brox, T. (2015) · Lecture notes in computer science · doi:10.1007/978-3-319-24574-4_28 · Cited by 99,404
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946 · cited 1×
“The 3D U-Net, 3D ResUNet, and 3D DeepLabv3 + were trained with a batch size of 2 for 100 epochs using Adam optimizer [34].”
MethodSemi-supervised classification with graph convolutional networks
Kipf, T. N., & Welling, M. (2016) · arXiv · doi:10.48550/arxiv.1609.02907 · Cited by 24,853 · cited 12×
“The primary advantage of the proposed method is employing GCN to improve the discrimination of different spinal structures by capturing the dependencies between the neigh-TABLE boring spinal structures.”
BackgroundInfluentialXception: Deep Learning with Depthwise Separable Convolutions
Chollet, F. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.195 · Cited by 20,380 · cited 1×
“In contrast to DeepLabv3 + , the deep convNet removes the middle flow of the Xception [30] for simplifying the network, and generates the low-level image representation for obtaining the details of the spine.”
BackgroundEncoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation
Chen, L.-C., Zhu, Y., Papandreou, G., Schroff, F., & Adam, H. (2018) · arXiv · doi:10.48550/arxiv.1802.02611 · Cited by 20,063 · cited 3×
“Atrous spatial pyramid pooling (ASPP) were employed in DeepLabv3 + [25] for producing multi-scale semantic image representation without detailed information loss.”
BackgroundMethodInfluentialEncoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation
Chen, L.-C., Zhu, Y., Papandreou, G., Schroff, F., & Adam, H. (2018) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-01234-2_49 · Cited by 20,054
Yushkevich, P. A., Piven, J., Hazlett, H. C., Smith, R. G., Ho, S., Gee, J. C., & Gerig, G. (2006) · NeuroImage · doi:10.1016/j.neuroimage.2006.01.015 · Cited by 8,868 · cited 1×
“The vertebrae and IVDs were manually delineated by a junior expert, and then corrected by a senior expert using the ITK-SNAP 1 [31].”
Method3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation
Çiçek, Ö., Abdulkadir, A., Lienkamp, S. S., Brox, T., & Ronneberger, O. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46723-8_49 · Cited by 7,945 · cited 2×
“The configurations of the 3D coarse segmentation methods except 3D GCSN are as follows: • 3D U-Net [32]: Four levels were used in the encoder path and the number of feature maps in the first encoder module was set to 32.”
MethodIterative fully convolutional neural networks for automatic vertebra segmentation and identification
Lessmann, N., van Ginneken, B., de Jong, P. A., & Išgum, I. (2019) · Medical Image Analysis · doi:10.1016/j.media.2019.02.005 · Cited by 266 · cited 2×
“To employ the dependencies between the neighboring spine structures, Lessmann et al. [19] proposed an iterative FCN to segment and label vertebrae from CT images and vertebral bodies from MR images one after the other.”
BackgroundMethodSpine-GAN: Semantic segmentation of multiple spinal structures
Han, Z., Wei, B., Mercado, A., Leung, S., & Li, S. (2018) · Medical Image Analysis · doi:10.1016/j.media.2018.08.005 · Cited by 211 · cited 1×
“Spine-GAN [27], a 2D semantic segmentation network for multiple spinal structures, utilized LSTM to generate semantic image representation by exploiting the long-range spatial correlation of the pixels in the feature maps.”
MethodSuperhuman Accuracy on the SNEMI3D Connectomics Challenge
Lee, K., Zung, J., Li, P. H., Jain, V., & Seung, H. S. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1706.00120 · Cited by 196 · cited 2×
“As shown in Table I and Table II, the 2D segmentation refinement stage improves the segmentation performance for all coarse segmentation results generated by the five 3D models including 3D U-Net [32], 3D ResUNet [33], 3D DeepLabv3 + , 3D Graphonomy, and 3D GCSN.”
MethodDeformations Incorporating Rigid Structures
Little, J. A., Hill, D. L. G., & Hawkes, D. J. (1997) · Computer Vision and Image Understanding · doi:10.1006/cviu.1997.0608 · Cited by 193 · cited 1×
“A practicable solution for spine registration is to achieve rigid registration for the vertebra while deformable registration for the remainder [2].”
BackgroundGraphonomy: Universal Human Parsing via Graph Transfer Learning
Gong, K., Gao, Y., Liang, X., Shen, X., Wang, M., & Lin, L. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00763 · Cited by 182 · cited 6×
“The matrix A e ∈ R n × n is a normalized symmetric adjacency matrix [29], in which the node adjacency weight a v → v ∈ A e is defined according to the edge connections in (cid:7) v, v (cid:8) ∈ E .”
BackgroundMethodInfluentialLi, X., Dou, Q., Chen, H., Fu, C., Qi, X., Belavý, D. L., Armbrecht, G., Felsenberg, D., Zheng, G., & Heng, P. (2018) · Medical Image Analysis · doi:10.1016/j.media.2018.01.004 · Cited by 169 · cited 2×
“Deep learning-based methods such as fully convolutional network (FCN) [12]–[15] and U-Net [16]–[18] have been popular in vertebrae segmentation for CT image and IVDs segmentation for MR image.”
MethodJanssens, R., Zeng, G., & Zheng, G. (2018) · IEEE International Symposium on Biomedical Imaging (ISBI) · doi:10.1109/isbi.2018.8363715 · Cited by 147
Neubert, A., Fripp, J., ENGSTROM, C. M., Schwarz, R., Lauer, L., Salvado, O., & Crozier, . (2012) · Physics in Medicine and Biology · doi:10.1088/0031-9155/57/24/8357 · Cited by 106 · cited 1×
“Most of these deformable models achieve promising accuracy but are sensitive to the initialization, which is done either manually [5], [8] or automatically [6], [9], [10].”
BackgroundSpine Segmentation in Medical Images Using Manifold Embeddings and Higher-Order MRFs
Kadoury, S., Labelle, H., & Paragios, N. (2013) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2013.2244903 · Cited by 89 · cited 2×
“Their surface-to-surface errors [7] for thoracic vertebrae and lumbar vertebrae segmentation from CT/MR images were 2.2 ± 1.5mm/2.9 ± 1.8mm and 2.8 ± 1.9mm/3.0 ± 1.9mm respectively.”
BackgroundMethodInfluentialIntervertebral disc segmentation in MR images using anisotropic oriented flux
Law, M. W. K., Tay, K. Y., Leung, A., Garvin, G. J., & Li, S. (2012) · Medical Image Analysis · doi:10.1016/j.media.2012.06.006 · Cited by 67 · cited 1×
“Most of these deformable models achieve promising accuracy but are sensitive to the initialization, which is done either manually [5], [8] or automatically [6], [9], [10].”
BackgroundLocalization and Segmentation of 3D Intervertebral Discs in MR Images by Data Driven Estimation
Chen, C., Belavý, D. L., Yu, W., Chu, C., Armbrecht, G., Bansmann, M. P., Felsenberg, D., & Guoyan Zheng, A. (2015) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2015.2403285 · Cited by 66 · cited 2×
“Chen et al. [11] proposed a machine learning-based method to achieve localization and segmentation of IVDs on volumetric MR images.”
MethodTrends in lumbar radiofrequency ablation utilization from 2007 to 2016
Starr, J. B., Gold, L., McCormick, Z. L., et al. (2019) · The Spine Journal · doi:10.1016/j.spinee.2019.01.001 · Cited by 64 · cited 1×
“For example, lumbar radiofrequency ablation (RFA) is a commonly used minimally invasive intervention for the treatment of low back pain and functional limitations caused by the lumbar disc herniation [1].”
BackgroundAttention-Driven Deep Learning for Pathological Spine Segmentation
Sekuboyina, A., Kukačka, J., Kirschke, J. S., Menze, B. H., & Valentinitsch, A. (2018) · Lecture notes in computer science · doi:10.1007/978-3-319-74113-0_10 · Cited by 60
3D Fully Convolutional Networks for Intervertebral Disc Localization and Segmentation
Chen, H., Dou, Q., Wang, X., Qin, J., Cheng, J. C. Y., & Heng, P. (2016) · Lecture notes in computer science · doi:10.1007/978-3-319-43775-0_34 · Cited by 59 · cited 1×
“Deep learning-based methods such as fully convolutional network (FCN) [12]–[15] and U-Net [16]–[18] have been popular in vertebrae segmentation for CT image and IVDs segmentation for MR image.”
MethodA Localisation-Segmentation Approach for Multi-label Annotation of Lumbar Vertebrae using Deep NetsPreprint
Sekuboyina, A., Valentinitsch, A., Kirschke, J. S., et al. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1703.04347 · Cited by 51 · cited 1×
“Deep learning-based methods such as fully convolutional network (FCN) [12]–[15] and U-Net [16]–[18] have been popular in vertebrae segmentation for CT image and IVDs segmentation for MR image.”
MethodSegmentation of Pathological Structures by Landmark-Assisted Deformable Models
Ibragimov, B., Korez, R., Likar, B., Pernuš, F., Xing, L., & Vrtovec, T. (2017) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2017.2667578 · Cited by 51 · cited 3×
“Most of these deformable models achieve promising accuracy but are sensitive to the initialization, which is done either manually [5], [8] or automatically [6], [9], [10].”
BackgroundMethodZhang, R., Xiao, X., Liu, Z., et al. (2020) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2020.2969084 · Cited by 38 · cited 1×
“[20] presented a multi-task relational learning network (MRLN) that utilized the depen-”
MethodPang, S., Lu, Z., Jiang, J., Zhao, L., Lin, L., Li, X., Li, X., Huang, M., Yang, W., & Feng, Q. (2019) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2019.2906727 · Cited by 35 · cited 1×
“Dice similarity coefficient (DSC), Precision, and Recall [35] were used as the quantitative metrics to evaluate the segmentation performance.”
MethodSelf-citationFallah, F., Walter, S. S., Bamberg, F., & Yang, B. (2018) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2018.2872810 · Cited by 25 · cited 1×
“Fallah et al. [21] achieved simultaneous volumetric segmentation of vertebral bodies (VBs) and IVDs on fat-water MR images by a hierarchical random forest classifier and a hierarchical conditional random field based on a set of multiscale local and contextual features.”
MethodFully Automatic Intervertebral Disc Segmentation Using Multimodal 3D U-Net
Wang, C., Guo, Y., Chen, W., & Yu, Z. (2019) · IEEE Annual International Computer Software and Applications Conference (COMPSAC) · doi:10.1109/compsac.2019.00109 · Cited by 22 · cited 1×
“These FCN and U-Net approaches are computation-efficient (about 10s for a volumetric image [15], [17]) in the test phase compared to the traditional methods (e.g., 1.4min for IVDs detection and 30s per IVD segmentation in [9]), but failed to explicitly consider the dependencies between the adjacent…”
MethodAutomatic Segmentation of the Scoliotic Spine from Mr Images
Guerroumi, N., Playout, C., Laporte, C., & Cheriet, F. (2019) · IEEE International Symposium on Biomedical Imaging (ISBI) · doi:10.1109/isbi.2019.8759413 · Cited by 18 · cited 1×
“Deep learning-based methods such as fully convolutional network (FCN) [12]–[15] and U-Net [16]–[18] have been popular in vertebrae segmentation for CT image and IVDs segmentation for MR image.”
MethodKorez, R., Ibragimov, B., Likar, B., Pernuš, F., & Vrtovec, T. (2016) · Lecture notes in computer science · doi:10.1007/978-3-319-41827-8_11 · Cited by 17 · cited 3×
“Most of these deformable models achieve promising accuracy but are sensitive to the initialization, which is done either manually [5], [8] or automatically [6], [9], [10].”
BackgroundMethodS3egANet: 3D Spinal Structures Segmentation via Adversarial Nets
Li, T., Wei, B., Cong, J., Li, X., & Li, S. (2019) · IEEE Access · doi:10.1109/access.2019.2962608 · Cited by 16 · cited 1×
“Li et al. [22] achieved simultaneous volumetric segmentation of vertebrae, IVDs, and neural foramens on T1 weighted and T2 weighted MR images by multi-modality adversarial nets.”
BackgroundSegmentation of Intervertebral Discs in 3D MRI Data Using Multi-atlas Based Registration
Wang, C., & Forsberg, D. (2016) · Lecture notes in computer science · doi:10.1007/978-3-319-41827-8_10 · Cited by 9 · cited 2×
“Wang and Forsberg [4] presented a multi-atlas based approach to achieve IVDs segmentation for volumetric MR images by three steps including vertebral bodies detection and labeling for target volume initialization, registration, and label fusion.”
BackgroundMethodChen, T., Wang, Z., Li, L., Li, D., Li, Z., Chang, X., Bian, P., Liang, G., Wang, Q., & Xie, T. (2025) · Schloss Dagstuhl – Leibniz-Zentrum für Informatik · doi:10.4230/lipics.ecoop.2025.6
Cited by
Segment anything model for medical images?
Huang, Y., Yang, X., Liu, L., Zhou, H., Chang, A., Zhou, X., Chen, R., Yu, J., Chen, J., Chen, C., et al. (2024) · Medical Image Analysis · doi:10.1016/j.media.2023.103061 · Cited by 536
“Medical images cover a wide range of object types, such as brain organs and tumors [29], [35], [42], [47], [66], lungs and hearts [23], [61], [64], [70], abdomen [22], [24], [57], [63], [66], [80], spine [65], [76], [77], cells [32], and polyps [38], [59], etc.”
Metadata
- Authors
- Pang, S., Pang, C., Zhao, L., Chen, Y., Su, Z., Zhou, Y., Huang, M., Yang, W., Lu, H., & Feng, Q.
- Venue
- IEEE Transactions on Medical Imaging
- Published
- 2021-01
- Volume
- 40
- Issue
- 1
- Pages
- 262-273
- ISSN
- 0278-0062 (ISSN-L)
- 1558-254X
- OpenAlex
- W3087644100
- Cited by
- 183
- References
- 35
- FWCI
- 6.19
- Type
- journal article
- Language
- en
- Open access
- closed
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.