Abstract
Magnetic resonance imaging (MRI) is a cornerstone in the evaluation and monitoring of axial spondyloarthritis (axSpA), a chronic inflammatory condition primarily affecting the sacroiliac joints (SIJs), spine, entheses, and peripheral joints. Accurate quantification of axSpA-related changes on MRI is critical for effective research and patient management; however, current lesion detection and grading approaches suffer from substantial intra- and inter-reader variability, limiting their consistency and reliability. To address these challenges, we propose a fully automated machine learning system for SIJ delineation and lesion classification on coronal MRI. The end-to-end pipeline automatically extracts SIJ contours using a vector-field—based open-contour model and classifies the presence or absence of five lesion types (bone marrow oedema, ankylosis, sclerosis, erosions, and fatty lesions) using both T1-weighted and STIR sequences. A multi-reader learning framework is employed to explicitly model inter- and intra-reader variability by leveraging multiple readings and consensus labels. Model performance was evaluated using patient-wise cross-validation on data from the MEASURE-1 clinical trial and further validated on other clinical datasets (PREVENT, SURPASS). Lesion classification performance was assessed using area under the receiver operating characteristic curve (AUC), balanced accuracy, sensitivity, and specificity, while contouring accuracy was quantified using root-mean-square error, where we found that 95% of the whole test set had errors below 2.76mm. The proposed approach achieved AUCs ranging from 0.85 to 0.99 across the five lesion types, with the highest performance observed when using consensus-based labels, and results were comparable to expert inter-reader agreement. These findings demonstrate that fully automated SIJ delineation and lesion scoring can achieve expert-level performance and have the potential to reduce reader burden and variability in large-scale axSpA MRI studies.
Data availability
The MEASURE 1, PREVENT, and SURPASS datasets were obtained through the Big Data Institute and Novartis research alliance and are owned by Novartis Pharmaceuticals. Access to these datasets can be requested directly from Novartis Pharmaceuticals. The OSIJ scans were extracted and anonymised from local hospital systems under a contract that explicitly prohibits data sharing with third parties for patient privacy reasons; therefore, they cannot be made publicly available.
References
Baraliakos, X. et al. MRI lesions of the spine in patients with axial spondyloarthritis: An update of lesion definitions and validation by the ASAS MRI Working Group. Ann. Rheum. Dis. 81, 1243–1251. https://doi.org/10.1136/annrheumdis-2021-222081. https://ard.bmj.com/content/81/9/1243.full.pdf (2022).
Maksymowych, W. P. et al. Data-driven definitions for active and structural MRI lesions in the sacroiliac joint in spondyloarthritis and their predictive utility. Rheumatology 60, 4778–4789. https://doi.org/10.1093/rheumatology/keab099. https://academic.oup.com/rheumatology/article-pdf/60/10/4778/40500312/keab099.pdf (2021).
Østergaard, M. & Lambert, R. G. W. Imaging in ankylosing spondylitis. Ther. Adv. Musculoskelet. Dis. 4, 301–311. https://doi.org/10.1177/1759720X11436240 (2012).
Caetano, A. P., Mascarenhas, V. V. & Machado, P. M. Axial spondyloarthritis: Mimics and pitfalls of imaging assessment. Front. Med. https://doi.org/10.3389/fmed.2021.658538 (2021).
Lukas, C. et al. Scoring inflammatory activity of the spine by magnetic resonance imaging in ankylosing spondylitis: A multireader experiment. J. Rheumatol. 34, 862–870 (2007).
Song, I. et al. Relationship between active inflammatory lesions in the spine and sacroiliac joints and new development of chronic lesions on whole-body MRI in early axial spondyloarthritis: Results of the Esther trial at week 48. Ann. Rheum. Dis. 70, 1257–1263. https://doi.org/10.1136/ard.2010.147033 (2011) (epub 2011 May 8).
Jamaludin, A., Lootus, M., Kadir, T. & Zisserman, A. Automatic intervertebral discs localization and segmentation: A vertebral approach. In Computational Methods and Clinical Applications for Spine Imaging (2015).
Windsor, R., Jamaludin, A., Kadir, T. & Zisserman, A. A convolutional approach to vertebrae detection and labelling in whole spine MRI. In Medical Image Computing and Computer Aided Intervention (2020).
Tao, R., Liu, W. & Zheng, G. Spine-transformers: Vertebra labeling and segmentation in arbitrary field-of-view spine CTS via 3D transformers. MedIA https://doi.org/10.1016/j.media.2021.102258 (2022).
Jamaludin, A., Kadir, T., Clark, E. & Zisserman, A. Predicting scoliosis in dxa scans using intermediate representations. In Computational Methods and Clinical Applications for Spine Imaging (2018).
Bourigault, E., Jamaludin, A., Kadir, T. & Zisserman, A. Scoliosis measurement on dxa scans using a combined deep learning and spinal geometry approach. In MIDL (2022).
Faleiros, M. C. et al. Machine learning techniques for computer-aided classification of active inflammatory sacroiliitis in magnetic resonance imaging. Adv. Rheumatol. https://doi.org/10.1186/s42358-020-00126-8 (2020).
Bressem, K. K. et al. Deep learning detects changes indicative of axial spondyloarthritis at MRI of sacroiliac joints. Radiology https://doi.org/10.1148/radiol.212526 (2022).
Mescheder, L., Oechsle, M., Niemeyer, M., Nowozin, S. & Geiger, A. Occupancy networks: Learning 3D reconstruction in function space. In CVPR (2019).
Park, J., Florence, P., Straub, J., Newcombe, R. & Lovegrove, S. Learning continuous signed distance functions for shape representation. In CVPR, DeepSDF (2019).
Chibane, J., Alldieck, T. & Pons-Moll, G. Implicit functions in feature space for 3D shape reconstruction and completion. In CVPR (2020).
Smith, H., Banerjee, A., Choudhury, R. & Grau, V. Automated torso contour extraction from clinical cardiac MR slices for 3D torso reconstruction. EMBC https://doi.org/10.1109/EMBC48229.2022.9871643 (2022).
Banerjee, A., Camps, J., Zacur, E., Andrews, C. M., Rudy, Y., Choudhury, R. P., Rodriguez, B. & Grau, V. A completely automated pipeline for 3D reconstruction of human heart from 2D cine magnetic resonance slices. Philos. Trans. R. Soc. A 379 (2021).
Kucybała, I., Tabor, Z., Polak, J., Urbanik, A. & Wojciechowski, W. The semi-automated algorithm for the detection of bone marrow oedema lesions in patients with axial spondyloarthritis. Rheumatol. Int. 40, 625–633 (2020).
Zarco, P., Almodóvar, R., Bueno, Á. & Molinero, L. M. SCAISS Study Group. Development and validation of SCAISS, a tool for semi-automated quantification of sacroilitis by magnetic resonance in spondyloarthritis. Rheumatol. Int. 38, 1919–1926 (2018).
Garrido-González, C. et al. Pos0958 responsiveness of conventional, semi-automatic and full-automatic methods to quantify marrow bone edema lesions in mri of axial spondyloarthritis patients: A pilot study. Ann. Rheum. Dis. 80, 743–744. https://doi.org/10.1136/annrheumdis-2021-eular.864. https://ard.bmj.com/content/80/Suppl_1/743.2.full.pdf (2021).
Ożga, J., Wyka, M., Raczko, A., Tabor, Z., Oleniacz, Z., Korman, M. & Wojciechowski, W. Performance of fully automated algorithm detecting bone marrow edema in sacroiliac joints. J. Clin. Med. 12 (2023).
Rzecki, K. et al. Fully automated algorithm for the detection of bone marrow oedema lesions in patients with axial spondyloarthritis – Feasibility study. Biocybern. Biomed. Eng. 41, 833–853. https://doi.org/10.1016/j.bbe.2021.05.005 (2021).
Bressem, K. K. et al. Deep learning detects changes indicative of axial spondyloarthritis at MRI of sacroiliac joints. Radiology 305, 655–665 (2022).
Lee, S. et al. Artificial intelligence for the detection of sacroiliitis on magnetic resonance imaging in patients with axial spondyloarthritis. Front. Immunol. 14, 1278247 (2023).
Nicolaes, J. et al. Performance analysis of a deep-learning algorithm to detect the presence of inflammation in MRI of sacroiliac joints in patients with axial spondyloarthritis. Ann. Rheum. Dis. (2024).
Jamaludin, A., Ather, S., Kadir, T. & Windsor, R. Contouring by unit vector field regression. In 2023 IEEE 20th International Symposium on Biomedical Imaging (ISBI). 1–5. https://doi.org/10.1109/ISBI53787.2023.10230540 (2023).
Krohn, M. et al. Erosions and fatty lesions of sacroiliac joints in patients with axial spondyloarthritis: Evaluation of different MRI techniques and two scoring methods. J. Rheumatol. 41, https://doi.org/10.3899/jrheum.130581. https://www.jrheum.org/content/41/3/473.full.pdf (2014).
Maksymowych, W. P. et al. Development and validation of web-based training modules for systematic evaluation of active inflammatory lesions in the spine and sacroiliac joints in spondyloarthritis. J. Rheumatol. Suppl. 84, 48–57 (2009).
Kass, M., Witkin, A. & Terzopoulos, D. Snakes: Active contour models. In IJCV. Vol. 1 (1988).
Xu, C. & Prince, J. Snakes, shapes, and gradient vector flow. IEEE TPAMI 7, 359–369. https://doi.org/10.1109/83.661186 (1998).
Payer, C., Štern, D., Bischof, H. & Urschler, M. Regressing heatmaps for multiple landmark localization using CNNs. In MICCAI (2016).
Zhou, X., Wang, D. & Krähenbühl, P. Objects as points. arXiv preprint arXiv:1904.07850 (2019).
Ronneberger, O., Fischer, P. & Brox, T. Convolutional networks for biomedical image segmentation. In MICCAI, U-net (2015).
Kingma, D. P. & Ba, J. A method for stochastic optimization. In ICLR, Adam (2015).
Blair, H. A. Secukinumab: A review in ankylosing spondylitis. Drugs 79, 433–443 (2019).
He, K., Zhang, X., Ren, S. & Sun, J. Deep residual learning for image recognition. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 770–778. https://doi.org/10.1109/CVPR.2016.90 (2016).
Windsor, R., Jamaludin, A., Kadir, T. & Zisserman, A. Spinenetv2: Automated detection, labelling and radiological grading of clinical MR scans. arxiv:2205.01683 (2022).
Tanno, R., Saeedi, A., Sankaranarayanan, S., Alexander, D. C. & Silberman, N. Learning from noisy labels by regularized estimation of annotator confusion. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2019).
Hu, E. J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L. & Chen, W. Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685 (2021).
Baeten, D. et al. Secukinumab, an interleukin-17A inhibitor, ankylosing spondylitis. N. Engl. J. Med. 373, 2534–2548 (2015).
Deodhar, A. et al. Improvement of signs and symptoms of nonradiographic axial spondyloarthritis in patients treated with secukinumab: Primary results of a randomized placebo-controlled phase III study. Arthritis Rheumatol. 73, 110–120 (2021).
Baraliakos, X. et al. Comparison of the effects of secukinumab and adalimumab biosimilar on radiographic progression in patients with ankylosing spondylitis: Design of a randomized, phase IIIb study (SURPASS). Clin. Drug Invest. 40, 269–278 (2020).
Lin, Y., Chan, S. C. W., Chung, H. Y., Lee, K. H. & Cao, P. A deep neural network for MRI spinal inflammation in axial spondyloarthritis. Eur. Spine J. 33, 4125–4134. https://doi.org/10.1007/s00586-023-08099-0 (2024).
Funding
A.R. and G.L. are supported by Novartis Pharmaceutical Corporation. P.M. received consulting fees from Novartis Pharmaceutical Corporation. A.J., S.A., R.W., and T.K. received funding from the Oxford BDI–Novartis Collaboration for AI in Medicine. A.J. is also supported by the Visual AI Programme Grant from the Engineering and Physical Sciences Research Council (EP/T028572/1).
Author information
Authors and Affiliations
Contributions
A.J. wrote the main manuscript and conducted all the experiments. S.A. collected and annotated the sacroiliac joint MRIs from the Oxford University Hospitals NHS Foundation Trust. R.W., G.L., A.R., P.M., and T.K. contributed to the development and discussion of the overall pipeline. All authors reviewed the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Jamaludin, A., Windsor, R., Ather, S. et al. Learning from multiple readings for axial spondyloarthritis classification of the sacroiliac joints. Sci Rep (2026). https://doi.org/10.1038/s41598-026-39417-3
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-026-39417-3
