{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,29]],"date-time":"2026-07-29T05:02:24Z","timestamp":1785301344951,"version":"3.55.0"},"reference-count":35,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,7,29]],"date-time":"2026-07-29T00:00:00Z","timestamp":1785283200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Physiol."],"abstract":"<jats:sec>\n                    <jats:title>Background<\/jats:title>\n                    <jats:p>Spinal tuberculosis is the most common extrapulmonary manifestation of tuberculosis. Magnetic resonance imaging (MRI), particularly fat-suppressed T2-weighted imaging (FS-T2WI), is the modality of choice for preoperative evaluation; yet indistinct lesion boundaries render manual delineation subjective and poorly reproducible. Although deep learning segmentation has advanced considerably, its \u201cblack-box\u201d nature and limited interpretability remain critical obstacles. This study aimed to develop an uncertainty-guided deep learning segmentation framework and evaluate its accuracy and clinical utility.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>\n                      We retrospectively enrolled 210 patients with spinal tuberculosis from an initial cohort of 300 screened at three centers, and acquired preoperative FS-T2WI scans. Data from Center 1 (\n                      <jats:italic>n<\/jats:italic>\n                      \u00a0=\u00a0160) were used for five-fold cross-validation, while the remaining 50 cases served as an external test set. We built an improved model on nnU-Net by integrating boundary-aware loss with Monte Carlo Dropout, and compared it against U-Net, Attention U-Net, and TransUNet. The uncertainty threshold was determined through internal cross-validation, and clinical validation followed a within-subject crossover design involving nine physicians.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>\n                      On the external test set, the improved model achieved a Dice similarity coefficient of 0.858 and an AUC of 0.912, outperforming all comparative models. Uncertainty correlated strongly and positively with pixel-level error rates (Spearman\n                      <jats:italic>\u03c1<\/jats:italic>\n                      \u00a0=\u00a00.74). At a threshold of 0.52, sensitivity for identifying unreliable segmentation was 84.3% and negative predictive value was 89.5%. In the physician validation, overlaying uncertainty maps significantly increased trust scores (3.8 \u00b1 0.7 vs. 3.2 \u00b1 0.8;\n                      <jats:italic>P<\/jats:italic>\n                      \u00a0&amp;lt;\u00a00.001, Cohen\u2019s\n                      <jats:italic>d<\/jats:italic>\n                      \u00a0=\u00a00.80) and reduced review time (\n                      <jats:italic>P<\/jats:italic>\n                      \u00a0=\u00a00.018), with the greatest benefit observed among resident physicians. In cases of obvious AI failure, uncertainty alerts accurately flagged erroneous regions in 80.0% of instances.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusion<\/jats:title>\n                    <jats:p>The proposed uncertainty-guided framework improved spinal tuberculosis lesion segmentation accuracy. Pixel-level uncertainty maps reliably identified unreliable predictions and enhanced clinician trust, offering a robust AI decision-support tool for precise surgical planning.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/fphys.2026.1885886","type":"journal-article","created":{"date-parts":[[2026,7,29]],"date-time":"2026-07-29T04:34:29Z","timestamp":1785299669000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Deep learning segmentation with uncertainty quantification for spinal tuberculosis on fat-suppressed T2-weighted MRI"],"prefix":"10.3389","volume":"17","author":[{"given":"Xingyu","family":"Duan","sequence":"first","affiliation":[{"name":"Department of Orthopedics, General Hospital of Ningxia Medical University","place":["Yinchuan, Ningxia, China"]},{"name":"First Clinical Medical College, Ningxia Medical University","place":["Yinchuan, Ningxia, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jiaxing","family":"Wang","sequence":"additional","affiliation":[{"name":"Department of Orthopedics, General Hospital of Ningxia Medical University","place":["Yinchuan, Ningxia, China"]},{"name":"First Clinical Medical College, Ningxia Medical University","place":["Yinchuan, Ningxia, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Linan","family":"Wang","sequence":"additional","affiliation":[{"name":"Department of Orthopedics, General Hospital of Ningxia Medical University","place":["Yinchuan, Ningxia, China"]},{"name":"First Clinical Medical College, Ningxia Medical University","place":["Yinchuan, Ningxia, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yichao","family":"Fan","sequence":"additional","affiliation":[{"name":"Department of Orthopedics, Zhoukou Central Hospital","place":["Zhoukou, Henan, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jiong","family":"Wang","sequence":"additional","affiliation":[{"name":"Department of Orthopedics, General Hospital of Ningxia Medical University","place":["Yinchuan, Ningxia, China"]},{"name":"First Clinical Medical College, Ningxia Medical University","place":["Yinchuan, Ningxia, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Qianpeng","family":"Ma","sequence":"additional","affiliation":[{"name":"Department of Orthopedics, The Third Hospital of Shizuishan (Pingluo People\u2019s Hospital)","place":["Shizuishan, Ningxia, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ningkui","family":"Niu","sequence":"additional","affiliation":[{"name":"Department of Orthopedics, General Hospital of Ningxia Medical University","place":["Yinchuan, Ningxia, China"]},{"name":"Ningxia Key Laboratory of Clinical Pathogenic Microorganisms, General Hospital of Ningxia Medical University","place":["Yinchuan, Ningxia, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,7,29]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"206","DOI":"10.3390\/biomimetics11030206","article-title":"A trustable spine abnormalities classification system using ResNet50 and VGG16 supported by explainable artificial intelligence","volume":"11","author":"Ahmad","year":"2026","journal-title":"Biomimetics (Basel)"},{"key":"B2","doi-asserted-by":"publisher","DOI":"10.1002\/hbm.26803","article-title":"Segmentation of stroke lesions using transformers-augmented MRI analysis","volume":"45","author":"Ahmed","year":"2024","journal-title":"Hum. Brain Mapp."},{"key":"B3","doi-asserted-by":"publisher","first-page":"E72","DOI":"10.1007\/s00330-003-2064-8","article-title":"Spinal and spinal cord infection","volume":"14","author":"Bal\u00e9riaux","year":"2004","journal-title":"Eur. Radiol."},{"key":"B4","doi-asserted-by":"publisher","first-page":"3596","DOI":"10.3390\/s25123596","article-title":"Segment Anything Model (SAM) and Medical SAM (MedSAM) for lumbar spine MRI","volume":"25","author":"Chang","year":"2025","journal-title":"Sensors (Basel)"},{"key":"B5","doi-asserted-by":"publisher","DOI":"10.1016\/j.ejrad.2024.111655","article-title":"Differentiation of tuberculous and brucellar spondylitis using conventional MRI-based deep learning algorithms","volume":"178","author":"Chen","year":"2024","journal-title":"Eur. J. Radiol."},{"key":"B6","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2102.04306","article-title":"TransUNet: Transformers make strong encoders for medical image segmentation","author":"Chen","year":"2021","journal-title":"arXiv preprint arXiv:2102.04306"},{"key":"B7","doi-asserted-by":"publisher","first-page":"2737","DOI":"10.3390\/diagnostics13172737","article-title":"Diagnostic approach and differences between spinal infections and tumors","volume":"13","author":"Compagnone","year":"2023","journal-title":"Diagnostics (Basel)"},{"key":"B8","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2024.103287","article-title":"Image-level supervision and self-training for transformer-based cross-modality tumor segmentation","volume":"97","author":"de Boisredon d'Assier","year":"2024","journal-title":"Med. Image Anal."},{"key":"B9","doi-asserted-by":"publisher","first-page":"1809","DOI":"10.1007\/s00330-002-1609-6","article-title":"Imaging features of musculoskeletal tuberculosis","volume":"13","author":"De Vuyst","year":"2003","journal-title":"Eur. Radiol."},{"key":"B10","doi-asserted-by":"publisher","first-page":"4325","DOI":"10.2147\/idr.S417663","article-title":"Accurate differentiation of spinal tuberculosis and spinal metastases using MR-based deep learning algorithms","volume":"16","author":"Duan","year":"2023","journal-title":"Infect. Drug Resist."},{"key":"B11","first-page":"1050","article-title":"Dropout as a Bayesian approximation: Representing model uncertainty in deep learning","author":"Gal","year":"2015"},{"key":"B12","doi-asserted-by":"publisher","first-page":"113","DOI":"10.1097\/brs.0000000000001697","article-title":"Comparison of tuberculous and brucellar spondylitis on magnetic resonance images","volume":"42","author":"Gao","year":"2017","journal-title":"Spine (Phila Pa 1976)"},{"key":"B13","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pone.0272529","article-title":"Experimental validation of a subject-specific finite element model of lumbar spine segment using digital image correlation","volume":"17","author":"Garavelli","year":"2022","journal-title":"PloS One"},{"key":"B14","doi-asserted-by":"publisher","first-page":"4175","DOI":"10.1002\/mp.17742","article-title":"S(2)Net: Self-adaptive weighted fusion and self-adaptive aligned network for multi-modal MRI segmentation","volume":"52","author":"Gui","year":"2025","journal-title":"Med. Phys."},{"key":"B15","doi-asserted-by":"publisher","first-page":"9861","DOI":"10.1038\/s41598-026-37169-8","article-title":"TrustNet: A lightweight network with integrated uncertainty quantification and quantitative explainable AI for ischemic stroke detection in CT images","volume":"16","author":"Inamdar","year":"2026","journal-title":"Sci. Rep."},{"key":"B16","doi-asserted-by":"publisher","first-page":"203","DOI":"10.1038\/s41592-020-01008-z","article-title":"nnU-Net: A self-configuring method for deep learning-based biomedical image segmentation","volume":"18","author":"Isensee","year":"2021","journal-title":"Nat. Methods"},{"key":"B17","doi-asserted-by":"publisher","first-page":"617","DOI":"10.2106\/jbjs.19.00001","article-title":"Tuberculosis of the spine","volume":"102","author":"Jain","year":"2020","journal-title":"J. Bone Joint Surg. Am."},{"key":"B18","doi-asserted-by":"publisher","first-page":"E43","DOI":"10.1007\/s00330-003-2046-x","article-title":"Vertebral infection","volume":"14","author":"Jevtic","year":"2004","journal-title":"Eur. Radiol."},{"key":"B19","doi-asserted-by":"publisher","first-page":"1942","DOI":"10.1016\/j.spinee.2025.03.009","article-title":"Using deep learning to enhance reporting efficiency and accuracy in degenerative cervical spine MRI","volume":"25","author":"Lee","year":"2025","journal-title":"Spine J."},{"key":"B20","doi-asserted-by":"publisher","DOI":"10.3389\/fmicb.2022.823324","article-title":"Computer-aided diagnosis of spinal tuberculosis from CT images based on deep learning with multimodal feature fusion","volume":"13","author":"Li","year":"2022","journal-title":"Front. Microbiol."},{"key":"B21","doi-asserted-by":"publisher","first-page":"24306","DOI":"10.1038\/s41598-025-09351-x","article-title":"AG-MS3D-CNN multiscale attention guided 3D convolutional neural network for robust brain tumor segmentation across MRI protocols","volume":"15","author":"Lilhore","year":"2025","journal-title":"Sci. Rep."},{"key":"B22","doi-asserted-by":"publisher","first-page":"31","DOI":"10.2147\/idr.S482584","article-title":"Deep learning for discrimination of early spinal tuberculosis from acute osteoporotic vertebral fracture on CT","volume":"18","author":"Liu","year":"2025","journal-title":"Infect. Drug Resist."},{"key":"B23","doi-asserted-by":"publisher","DOI":"10.1080\/24699322.2025.2599329","article-title":"Deep learning diagnosis model of spinal tuberculosis based on CT bone window gradient attention mechanism: Multi-center study","volume":"31","author":"Mo","year":"2026","journal-title":"Comput. Assist. Surg. (Abingdon)"},{"key":"B24","doi-asserted-by":"publisher","DOI":"10.1002\/acm2.14586","article-title":"Performance of recurrent neural networks with Monte Carlo dropout for predicting pharmacokinetic parameters from dynamic contrast-enhanced magnetic resonance imaging data","volume":"26","author":"Murase","year":"2025","journal-title":"J. Appl. Clin. Med. Phys."},{"key":"B25","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1804.03999","article-title":"Attention U-Net: Learning where to look for the pancreas","author":"Oktay","year":"2018","journal-title":"arXiv preprint arXiv:1804.03999"},{"key":"B26","doi-asserted-by":"publisher","first-page":"67","DOI":"10.1097\/00003086-200205000-00010","article-title":"Atypical spinal tuberculosis","author":"Pande","year":"2002","journal-title":"Clin. Orthop. Relat. Res."},{"key":"B27","doi-asserted-by":"publisher","first-page":"3454","DOI":"10.21037\/qims-21-939","article-title":"Current development and prospects of deep learning in spine image analysis: A literature review","volume":"12","author":"Qu","year":"2022","journal-title":"Quant. Imaging Med. Surg."},{"key":"B28","doi-asserted-by":"publisher","DOI":"10.1016\/j.compbiomed.2023.107723","article-title":"Brain tumor segmentation based on optimized convolutional neural network and improved chimp optimization algorithm","volume":"168","author":"Ranjbarzadeh","year":"2024","journal-title":"Comput. Biol. Med."},{"key":"B29","first-page":"234","article-title":"U-Net: Convolutional networks for biomedical image segmentation","volume-title":"Medical Image Computing and Computer-Assisted Intervention (MICCAI)","author":"Ronneberger","year":"2015"},{"key":"B30","doi-asserted-by":"publisher","first-page":"1307","DOI":"10.3390\/diagnostics12061307","article-title":"Surgical management of spinal tuberculosis-the past, present, and future","volume":"12","author":"Ruparel","year":"2022","journal-title":"Diagnostics (Basel)"},{"key":"B31","doi-asserted-by":"publisher","DOI":"10.1016\/j.compbiomed.2023.107441","article-title":"Application of uncertainty quantification to artificial intelligence in healthcare: A review of last decade, (2013-2023)","volume":"165","author":"Seoni","year":"2023","journal-title":"Comput. Biol. Med."},{"key":"B32","doi-asserted-by":"publisher","first-page":"686","DOI":"10.31616\/asj.2018.12.4.686","article-title":"Role of whole-spine screening magnetic resonance imaging using short tau inversion recovery or fat-suppressed T2 fast spin echo sequences for detecting noncontiguous multiple-level spinal tuberculosis","volume":"12","author":"Siddiqui","year":"2018","journal-title":"Asian Spine J."},{"key":"B33","doi-asserted-by":"publisher","first-page":"264","DOI":"10.1038\/s41597-024-03090-w","article-title":"Lumbar spine segmentation in MR images: A dataset and a public benchmark","volume":"11","author":"van der Graaf","year":"2024","journal-title":"Sci. Data"},{"key":"B34","doi-asserted-by":"publisher","first-page":"4246","DOI":"10.1007\/s00586-023-07728-y","article-title":"Trends and prospects in spinal tuberculosis research: A future-oriented approach","volume":"32","author":"Zhang","year":"2023","journal-title":"Eur. Spine J."},{"key":"B35","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2025.103923","article-title":"DFuse-Net: Disentangled feature fusion with uncertainty-aware learning for reliable multi-modal brain tumor segmentation","volume":"109","author":"Zhou","year":"2026","journal-title":"Med. Image Anal."}],"container-title":["Frontiers in Physiology"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fphys.2026.1885886\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,29]],"date-time":"2026-07-29T04:34:34Z","timestamp":1785299674000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fphys.2026.1885886\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7,29]]},"references-count":35,"alternative-id":["10.3389\/fphys.2026.1885886"],"URL":"https:\/\/doi.org\/10.3389\/fphys.2026.1885886","relation":{},"ISSN":["1664-042X"],"issn-type":[{"value":"1664-042X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7,29]]},"article-number":"1885886"}}