{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T11:19:35Z","timestamp":1788261575749,"version":"build-2803163510"},"reference-count":27,"publisher":"IOP Publishing","issue":"17","license":[{"start":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T00:00:00Z","timestamp":1788220800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"},{"start":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T00:00:00Z","timestamp":1788220800000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/iopscience.iop.org\/info\/page\/text-and-data-mining"},{"start":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T00:00:00Z","timestamp":1788220800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/publishingsupport.iopscience.iop.org\/questions\/alternative-author-rights-policies\/"},{"start":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T00:00:00Z","timestamp":1788220800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/publishingsupport.iopscience.iop.org\/questions\/alternative-author-rights-policies\/"},{"start":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T00:00:00Z","timestamp":1788220800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/publishingsupport.iopscience.iop.org\/questions\/alternative-author-rights-policies\/"},{"start":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T00:00:00Z","timestamp":1788220800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/publishingsupport.iopscience.iop.org\/questions\/alternative-author-rights-policies\/"},{"start":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T00:00:00Z","timestamp":1788220800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/publishingsupport.iopscience.iop.org\/questions\/alternative-author-rights-policies\/"},{"start":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T00:00:00Z","timestamp":1788220800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/publishingsupport.iopscience.iop.org\/questions\/alternative-author-rights-policies\/"},{"start":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T00:00:00Z","timestamp":1788220800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/publishingsupport.iopscience.iop.org\/questions\/alternative-author-rights-policies\/"},{"start":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T00:00:00Z","timestamp":1788220800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/publishingsupport.iopscience.iop.org\/questions\/alternative-author-rights-policies\/"},{"start":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T00:00:00Z","timestamp":1788220800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/publishingsupport.iopscience.iop.org\/questions\/alternative-author-rights-policies\/"},{"start":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T00:00:00Z","timestamp":1788220800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/publishingsupport.iopscience.iop.org\/questions\/alternative-author-rights-policies\/"},{"start":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T00:00:00Z","timestamp":1788220800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/publishingsupport.iopscience.iop.org\/questions\/alternative-author-rights-policies\/"},{"start":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T00:00:00Z","timestamp":1788220800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/publishingsupport.iopscience.iop.org\/questions\/alternative-author-rights-policies\/"},{"start":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T00:00:00Z","timestamp":1788220800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/publishingsupport.iopscience.iop.org\/questions\/alternative-author-rights-policies\/"},{"start":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T00:00:00Z","timestamp":1788220800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/publishingsupport.iopscience.iop.org\/questions\/alternative-author-rights-policies\/"}],"funder":[{"DOI":"10.13039\/501100000780","name":"European Union","doi-asserted-by":"crossref","award":["101088594"],"award-info":[{"award-number":["101088594"]}],"id":[{"id":"10.13039\/501100000780","id-type":"DOI","asserted-by":"crossref"}]}],"content-domain":{"domain":["iopscience.iop.org"],"crossmark-restriction":false},"short-container-title":["Phys. Med. Biol."],"published-print":{"date-parts":[[2026,9,14]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>\n                    <jats:italic>Objective.<\/jats:italic>\n                    This study systematically investigates the robustness of radiomics-based machine learning models to distribution shifts caused by variations in MRI acquisition protocols and image segmentation strategies, and evaluates strategies for maintaining predictive performance and uncertainty calibration under such shifts.\n                    <jats:italic>Approach<\/jats:italic>\n                    . Using a controlled phantom comprising 16 fruits of four types, we acquired images across five MRI sequences (T2-HASTE, T2-TSE, T2-MAP, T1-TSE, T2-FLAIR) with multiple scans, observers, and segmentation variants (full, partial, rotated). XGBoost classifiers were trained using three feature selection strategies: eight protocol-invariant features, sequence-specific robust features, and all 107 available features. Model performance and calibration were evaluated under in-domain, cross-protocol, segmentation-induced, and compound distribution shift scenarios using a nested, group-wise (leave-one-row-out) cross-validation scheme that prevents instance-level data leakage. Because all data derive from a controlled phantom comprising only 16 independent physical objects, reported variability reflects repeated technical evaluation conditions (scans, observers, segmentation variants) rather than uncertainty across independent biological samples.\n                    <jats:italic>Main results<\/jats:italic>\n                    . Under in-domain conditions, all three feature strategies achieved comparable performance. Under distribution shifts, protocol-invariant features retained substantially more predictive power than sequence-specific or all-feature models, and multi-protocol training further improved generalization, though residual degradation persisted. Calibration proved far more vulnerable than accuracy, with expected calibration error reaching 0.4 under compound shifts. Training-time feature-space augmentation yielded more consistent improvements in calibration than post-hoc temperature scaling, though its effect on predictive performance was mixed.\n                    <jats:italic>Significance<\/jats:italic>\n                    . These findings establish protocol-invariant feature selection, multi-protocol training, and feature-space augmentation as complementary strategies for improving radiomics model resilience, and identify calibration as a vulnerable dimension of model reliability under distribution shift. The controlled phantom framework isolates specific failure modes and enables quantitative benchmarking of robustness strategies prior to clinical validation. As this study is based on a controlled phantom, these findings are hypothesis-generating and require independent validation in clinical cohorts before any translation to patient imaging.\n                  <\/jats:p>","DOI":"10.1088\/1361-6560\/ae97ad","type":"journal-article","created":{"date-parts":[[2026,8,10]],"date-time":"2026-08-10T22:52:17Z","timestamp":1786402337000},"page":"175005","update-policy":"https:\/\/doi.org\/10.1088\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Robustness of radiomics-based machine learning to real-world MRI distribution shifts: a phantom study"],"prefix":"10.1088","volume":"71","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-6696-6391","authenticated-orcid":true,"given":"Sarmad Ahmad","family":"Khan","sequence":"first","affiliation":[{"name":"German Cancer Consortium (DKTK), partner site Frankfurt\/Mainz, a partnership between DKFZ and UCT Frankfurt- Marburg"},{"name":"German Cancer Research Center (DKFZ)"},{"name":"Goethe University Frankfurt"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Simon","family":"Bernatz","sequence":"additional","affiliation":[{"name":"Goethe University Frankfurt"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9022-7513","authenticated-orcid":true,"given":"Zahra","family":"Moslehi","sequence":"additional","affiliation":[{"name":"German Cancer Consortium (DKTK), partner site Frankfurt\/Mainz, a partnership between DKFZ and UCT Frankfurt- Marburg"},{"name":"Goethe University Frankfurt"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5587-6761","authenticated-orcid":true,"given":"Florian","family":"Buettner","sequence":"additional","affiliation":[{"name":"German Cancer Consortium (DKTK), partner site Frankfurt\/Mainz, a partnership between DKFZ and UCT Frankfurt- Marburg"},{"name":"German Cancer Research Center (DKFZ)"},{"name":"Goethe University Frankfurt"},{"name":"Frankfurt Cancer Institute (FCI)"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"266","published-online":{"date-parts":[[2026,9,1]]},"reference":[{"key":"pmbae97adbib1","doi-asserted-by":"publisher","first-page":"829","DOI":"10.1093\/jnci\/86.11.829","type":"journal-article","article-title":"Dangers of using \u201coptimal\u201d cutpoints in the evaluation of prognostic factors","volume":"86","author":"Altman","year":"1994","journal-title":"JNCI J. Natl Cancer Inst."},{"key":"pmbae97adbib2","doi-asserted-by":"publisher","first-page":"221","DOI":"10.1097\/RLI.0000000000000530","type":"journal-article","article-title":"Robustness and reproducibility of radiomics in magnetic resonance imaging: a phantom study","volume":"54","author":"Bae\u00dfler","year":"2019","journal-title":"Invest. Radiol"},{"key":"pmbae97adbib3","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pone.0057105","type":"journal-article","article-title":"Predicting future morphological changes of lesions from radiotracer uptake in 18F-FDG-PET images","volume":"8","author":"Bagci","year":"2013","journal-title":"PLoS One"},{"key":"pmbae97adbib4","doi-asserted-by":"publisher","first-page":"407","DOI":"10.1148\/radiol.2018172361","type":"journal-article","article-title":"Radiomics of CT features may be nonreproducible and redundant: influence of CT acquisition parameters","volume":"288","author":"Berenguer","year":"2018","journal-title":"Radiology"},{"key":"pmbae97adbib5","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-021-93756-x","type":"journal-article","article-title":"Impact of rescanning and repositioning on radiomic features employing a multi-object phantom in magnetic resonance imaging","volume":"11","author":"Bernatz","year":"2021","journal-title":"Sci. Rep."},{"key":"pmbae97adbib6","doi-asserted-by":"crossref","DOI":"10.1145\/2939672.2939785","type":"preprint","article-title":"XGBoost: a scalable tree boosting system","author":"Chen","year":"2016"},{"key":"pmbae97adbib7","doi-asserted-by":"publisher","first-page":"1083","DOI":"10.1016\/j.ijrobp.2017.12.268","type":"journal-article","article-title":"Challenges and promises of PET radiomics","volume":"102","author":"Cook","year":"2018","journal-title":"Int. J. Radiat. Oncol. Biol. Phys."},{"key":"pmbae97adbib8","doi-asserted-by":"publisher","first-page":"1323","DOI":"10.1016\/j.mri.2012.05.001","type":"journal-article","article-title":"3D slicer as an image computing platform for the quantitative imaging network","volume":"30","author":"Fedorov","year":"2012","journal-title":"Magn. Reson. Imaging"},{"key":"pmbae97adbib9","doi-asserted-by":"publisher","first-page":"563","DOI":"10.1148\/radiol.2015151169","type":"journal-article","article-title":"Radiomics: images are more than pictures, they are data","volume":"278","author":"Gillies","year":"2016","journal-title":"Radiology"},{"key":"pmbae97adbib10","first-page":"1321","type":"conference-proceedings","article-title":"On calibration of modern neural networks","author":"Guo","year":"2017"},{"key":"pmbae97adbib11","doi-asserted-by":"publisher","first-page":"69","DOI":"10.1038\/s41571-022-00707-0","type":"journal-article","article-title":"Criteria for the translation of radiomics into clinically useful tests","volume":"20","author":"Huang","year":"2023","journal-title":"Nat. Rev. Clin. Oncol."},{"key":"pmbae97adbib12","doi-asserted-by":"publisher","first-page":"20","DOI":"10.1016\/j.ymeth.2020.05.022","type":"journal-article","article-title":"Radiomics for precision medicine: current challenges, future prospects, and the proposal of a new framework","volume":"188","author":"Ibrahim","year":"2021","journal-title":"Methods"},{"key":"pmbae97adbib13","doi-asserted-by":"publisher","first-page":"pp 12316","DOI":"10.5555\/3454287.3455390)","type":"conference-proceedings","article-title":"Beyond temperature scaling: Obtaining well-calibrated multi-class probabilities with Dirichlet calibration","volume":"vol 32","author":"Kull","year":"2019"},{"key":"pmbae97adbib14","doi-asserted-by":"publisher","first-page":"749","DOI":"10.1038\/nrclinonc.2017.141","type":"journal-article","article-title":"Radiomics: the bridge between medical imaging and personalized medicine","volume":"14","author":"Lambin","year":"2017","journal-title":"Nat. Rev. Clin. Oncol."},{"key":"pmbae97adbib15","doi-asserted-by":"publisher","first-page":"757","DOI":"10.1097\/RLI.0000000000000180","type":"journal-article","article-title":"Measuring computed tomography scanner variability of radiomics features","volume":"50","author":"Mackin","year":"2015","journal-title":"Invest. Radiol"},{"key":"pmbae97adbib16","first-page":"625","type":"conference-proceedings","article-title":"Predicting good probabilities with supervised learning","author":"Niculescu-Mizil","year":"2005"},{"key":"pmbae97adbib17","first-page":"632","type":"conference-proceedings","article-title":"3D Slicer","author":"Pieper","year":"2004"},{"key":"pmbae97adbib18","doi-asserted-by":"publisher","DOI":"10.1117\/1.jmi.9.4.044005","type":"journal-article","article-title":"Repeatability and reproducibility of magnetic resonance imaging-based radiomic features in rectal cancer","volume":"9","author":"Rai","year":"2022","journal-title":"J. Med. Imaging"},{"key":"pmbae97adbib19","doi-asserted-by":"publisher","DOI":"10.1259\/bjr.20220878","type":"journal-article","article-title":"Data drift in medical machine learning: implications and potential remedies","volume":"96","author":"Sahiner","year":"2023","journal-title":"Br. J. Radiol"},{"key":"pmbae97adbib20","doi-asserted-by":"publisher","first-page":"1143","DOI":"10.1016\/j.ijrobp.2018.05.053","type":"journal-article","article-title":"Repeatability and Reproducibility of Radiomic Features: a Systematic Review","volume":"102","author":"Traverso","year":"2018","journal-title":"Int. J. Radiat. Oncol. Biol. Phys."},{"key":"pmbae97adbib21","doi-asserted-by":"publisher","first-page":"e104","DOI":"10.1158\/0008-5472.CAN-17-0339","type":"journal-article","article-title":"Computational radiomics system to decode the radiographic phenotype","volume":"77","author":"van Griethuysen","year":"2017","journal-title":"Cancer Res"},{"key":"pmbae97adbib22","doi-asserted-by":"publisher","first-page":"R150","DOI":"10.1088\/0031-9155\/61\/13\/R150","type":"journal-article","article-title":"Applications and limitations of radiomics","volume":"61","author":"Yip","year":"2016","journal-title":"Phys. Med. Biol."},{"key":"pmbae97adbib23","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-023-45477-6","type":"journal-article","article-title":"Comparing effectiveness of image perturbation and test retest imaging in improving radiomic model reliability","volume":"13","author":"Zhang","year":"2023","journal-title":"Sci. Rep."},{"key":"pmbae97adbib24","first-page":"11117","type":"conference-proceedings","article-title":"Mix-n-match: ensemble and compositional methods for uncertainty calibration in deep learning","author":"Zhang","year":"2020"},{"key":"pmbae97adbib25","doi-asserted-by":"publisher","DOI":"10.3389\/fonc.2021.633176","type":"journal-article","article-title":"Understanding sources of variation to improve the reproducibility of radiomics","volume":"11","author":"Zhao","year":"2021","journal-title":"Front. Oncol."},{"key":"pmbae97adbib26","doi-asserted-by":"publisher","first-page":"171","DOI":"10.1097\/RLI.0000000000000827","type":"journal-article","article-title":"Feature robustness and diagnostic capabilities of convolutional neural networks against radiomics features in computed tomography imaging","volume":"57","author":"Ziegelmayer","year":"2022","journal-title":"Invest. Radiol"},{"key":"pmbae97adbib27","doi-asserted-by":"publisher","first-page":"614","DOI":"10.1038\/s41598-018-36938-4","type":"journal-article","article-title":"Assessing robustness of radiomic features by image perturbation","volume":"9","author":"Zwanenburg","year":"2019","journal-title":"Sci. Rep."}],"container-title":["Physics in Medicine &amp; Biology"],"original-title":[],"link":[{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/ae97ad","content-type":"text\/html","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/ae97ad\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/ae97ad","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/ae97ad\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/ae97ad\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/ae97ad\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/ae97ad\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"similarity-checking"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/ae97ad\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T11:01:52Z","timestamp":1788260512000},"score":1,"resource":{"primary":{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/ae97ad"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,1]]},"references-count":27,"journal-issue":{"issue":"17","published-online":{"date-parts":[[2026,9,1]]},"published-print":{"date-parts":[[2026,9,14]]}},"URL":"https:\/\/doi.org\/10.1088\/1361-6560\/ae97ad","relation":{},"ISSN":["0031-9155","1361-6560"],"issn-type":[{"value":"0031-9155","type":"print"},{"value":"1361-6560","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,1]]},"assertion":[{"value":"Robustness of radiomics-based machine learning to real-world MRI distribution shifts: a phantom study","name":"article_title","label":"Article Title"},{"value":"Physics in Medicine & Biology","name":"journal_title","label":"Journal Title"},{"value":"paper","name":"article_type","label":"Article Type"},{"value":"\u00a9 2026 The Author(s). Published on behalf of Institute of Physics and Engineering in Medicine by IOP Publishing Ltd","name":"copyright_information","label":"Copyright Information"},{"value":"2026-01-05","name":"date_received","label":"Date Received","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2026-08-10","name":"date_accepted","label":"Date Accepted","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2026-09-01","name":"date_epub","label":"Online publication date","group":{"name":"publication_dates","label":"Publication dates"}}]}}