Improving the efficiency of spine region segmentation using an ensemble of pre-trained neural networks
Abstract
Background: The accuracy of segmentation of vertebrae in X-ray images is critical for clinical decisions as the manual method is laborious. The use of deep learning is complicated by low contrast, noise, and patient position artifacts. These negative factors make a single neural network unreliable. Thus, to improve the accuracy and efficiency of segmentation, regardless of the quality of X-ray images, there is a need for an ensemble of neural networks that compensates for the individual shortcomings of the models by aggregating their results.
Objectives: Increasing the accuracy and efficiency of segmentation of a spinal region consisting of four vertebrae (Th8, Th9, Th10, Th11) in X-ray images by using an ensemble of pre-trained neural networks.
Materials and methods: Two datasets were used for the experiments: the first set with 183 images was distributed in the ratio of 70% / 10% / 20% for training, validation, and testing, in turn, the second set of 58 images was used exclusively for the final assessment of the generalization ability of the ensemble on new data. In the process of research, segmentation accuracy with and without augmentation was first compared, after which the 10 best from the initial 20 neural networks were selected for further use, and five ensemble algorithms were used for mask aggregation.
Results: For the ensemble of pre-trained neural networks, the best result was shown by soft voting. Comparing the obtained result with the results presented by Koniukhov et al. (2024), the improvement was 3.06%. This indicator clearly confirms the effectiveness of using pre-trained networks for segmentation of the spine area.
Conclusions: Soft voting for an ensemble of pre-trained neural networks demonstrated the greatest improvement in segmentation accuracy compared to other methods. Aggregating knowledge from 10 models successfully eliminated the limitations of individual models. The use of an ensemble of pre-trained neural networks improved segmentation accuracy for both the test data from the first dataset and the data from the second dataset. Such results confirm the feasibility of applying the proposed ensemble-based approach to chest X-ray radiographs for vertebrae segmentation in medical imaging tasks.
Downloads
References
GBD 2021 Low Back Pain Collaborators. Global, regional, and national burden of low back pain, 1990–2020, its attributable risk factors, and projections to 2050: a systematic analysis of the Global Burden of Disease Study 2021. Lancet Rheumatol. 2023;5(6):e316–e329. https://doi.org/10.1016/S2665-9913(23)00098-X
Ou X, Chen X, Xu X, Xie L, Chen X, Hong Z, et al. Recent development in X-ray imaging technology: future and challenges. Research. 2021;2021:9892152. https://doi.org/10.34133/2021/9892152
Goodwin ML, Buchowski JM, Sciubba DM. Why X-rays? The importance of radiographs in spine surgery. Spine J. 2022;22(11):1759–67. https://doi.org/10.1016/j.spinee.2022.07.102
Jin C, Wang S, Yang G, Li E, Liang Z. A review of the methods on Cobb angle measurements for spinal curvature. Sensors. 2022;22(9):3258. https://doi.org/10.3390/s22093258
Wills BP, Auerbach JD, Zhu X, Caird MS, Horn BD, Flynn JM, et al. Comparison of Cobb angle measurement of scoliosis radiographs with preselected end vertebrae: traditional versus digital acquisition. Spine. 2007;32(1):98–105. https://doi.org/10.1097/01.brs.0000251086.84420.d1
Minaee S, Boykov Y, Porikli F, Plaza A, Kehtarnavaz N, Terzopoulos D. Image segmentation using deep learning: a survey. IEEE Trans. Pattern Anal Mach Intell. 2022;44(7):3523–42. https://doi.org/10.1109/TPAMI.2021.3059968
Çallı E, Sogancioglu E, van Ginneken B, van Leeuwen KG, Murphy K. Deep learning for chest X-ray analysis: a survey. Med Image Anal. 2021;72:102125. https://doi.org/10.1016/j.media.2021.102125
Baltruschat IM, Nickisch H, Grass M, Knopp T, Saalbach A. Comparison of deep learning approaches for multi-label chest X-ray classification. Sci Rep. 2019;9(1):6381. https://doi.org/10.1038/s41598-019-42294-8
Akter S, Shamrat F, Chakraborty S, Karim A, Azam S. COVID-19 detection using deep learning algorithm on chest X-ray images. Biology. 2021;10(11):1174. https://doi.org/10.3390/biology10111174
Chen IDS, Yang C-M, Chen M-J, Chen M-C, Weng R-M, Yeh C-H. Deep learning-based recognition of periodontitis and dental caries in dental X-ray images. Bioengineering. 2023;10(8):911. https://doi.org/10.3390/bioengineering10080911
Hardalaç F, Uysal F, Peker O, Çiçeklidağ M, Tolunay T, Tokgöz N, et al. Fracture detection in wrist X-ray images using deep learning-based object detection models. Sensors. 2022;22(3):1285. https://doi.org/10.3390/s22031285
Abdullah SS, Rajasekaran MP. Automatic detection and classification of knee osteoarthritis using deep learning approach. Radiol med. 2022;127(3):398–406. https://doi.org/10.1007/s11547-022-01476-7
Hamdy S, Charrier A, Corre LL, Rasti P, Rousseau D. Toward robust and high-throughput detection of seed defects in X-ray images via deep learning. Plant Methods. 2024;20(1):63. https://doi.org/10.1186/s13007-024-01195-2
Alzubaidi L, Salhi A, Fadhel MA, Bai J, Hollman F, Italia K, et al. Trustworthy deep learning framework for the detection of abnormalities in X-ray shoulder images. PLoS One. 2024;19(3):e0299545. https://doi.org/10.1371/journal.pone.0299545
Cheng L-W, Chou H-H, Cai Y-X, Huang K-Y, Hsieh C-C, Chu P-L, et al. Automated detection of vertebral fractures from X-ray images: A novel machine learning model and survey of the field. Neurocomputing. 2024;566:126946. https://doi.org/10.1016/j.neucom.2023.126946
Koniukhov V. Improving the segmentation of the vertebrae using a multi-stage machine learning algorithm. Radioelectronic and Computer Systems. 2024;(4):83–90. https://doi.org/10.32620/reks.2024.4.07
Koniukhov VD, Morgun OM, Nemchenko KE. Impact of preprocessing and comparison of neural network ensemble methods for segmentation of the thoracic spine in X-ray images. Radio Electronics, Computer Science, Control. 2024;4:102–12. https://doi.org/10.15588/1607-3274-2024-4-10
Ronneberger O, Fischer P, Brox T. U-net: convolutional networks for biomedical image segmentation. In: Navab N, Hornegger J, Wells W, Frangi A, editors. Medical image computing and computer-assisted intervention – MICCAI 2015. MICCAI 2015. Lecture notes in computer science, vol 9351. Cham: Springer; 2015. p. 234-41. https://doi.org/10.1007/978-3-319-24574-4_28
Koniukhov VD, Morgun OM, Nemchenko KE. Comparative analysis of ensemble methods for X-ray images: study of the influence on trained and new data. In: Computer modeling in science-intensive technologies: proceedings of the International scientific and technical conference. Kharkiv; 2024. p. 100–103. (In Ukrainian). Available from: https://drive.google.com/file/d/1Vyvntnwud71YE21aGqyZMb483fAQL6pe/
Stanford AIMI. CheXpert Chest X-Rays [Internet]. AIMI Center. [cited 2025 Nov 16]. Available from: https://aimi.stanford.edu/datasets/chexpert-chest-x-rays
Vindr.ai. Vindr.ai datasets: SpineXR [Internet]. Vindr.ai. [cited 2025 Nov 16]. Available from: https://vindr.ai/spinexr
Qadri SF, Shen L, Ahmad M, Qadri S, Zareen SS, Akbar MA. SVseg: stacked sparse autoencoder-based patch classification modeling for vertebrae segmentation. Mathematics. 2022;10(5):796. https://doi.org/10.3390/math10050796
Lu H, Li M, Yu K, Zhang Y, Yu L. Lumbar spine segmentation method based on deep learning. J Appl Clin Med Phys. 2023;24(6):e13996. https://doi.org/10.1002/acm2.13996
van der Graaf JW, van Hooff ML, Buckens CFM, Rutten M, van Susante JLC, Kroeze RJ, et al. Lumbar spine segmentation in MR images: a dataset and a public benchmark. Scientific Data. 2024;11(1):264. https://doi.org/10.1038/s41597-024-03090-w
Copyright (c) 2026 V. D. Koniukhov, O. M. Morgun, K. E. Nemchenko
This work is licensed under a Creative Commons Attribution 4.0 International License.
Authors who publish with this journal agree to the following terms:
- Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See The Effect of Open Access).
