{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,19]],"date-time":"2026-06-19T04:54:30Z","timestamp":1781844870941,"version":"3.54.5"},"reference-count":45,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,6,19]],"date-time":"2026-06-19T00:00:00Z","timestamp":1781827200000},"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. Med."],"abstract":"<jats:sec>\n                    <jats:title>Background<\/jats:title>\n                    <jats:p>Thyroid eye disease (TED) severity assessment using the EUGOGO severity classification is partly subjective and prone to interobserver variability. While MRI-derived anatomical measurements offer objective features, such as ocular protrusion and extraocular muscle thickness, these are underutilized in machine learning (ML) models that often rely on non-interpretable radiomic features. Moreover, the inclusion of longitudinal scans from the same patient may artificially inflate model performance due to unaccounted intra-patient correlations and temporal redundancy.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Purpose<\/jats:title>\n                    <jats:p>To develop an interpretable machine learning framework for objective TED severity stratification (mild, moderate-to-severe, sight-threatening) by quantitatively integrating established orbital anatomical parameters aligned with clinical assessment criteria, and to evaluate how data handling strategies influence model generalizability.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>We retrospectively analyzed 443 patients with TED, yielding 1,054 orbital MRI units from all available examinations. Two datasets were constructed: Dataset A included all available scans, while Dataset B retained only the first-visit orbital MRI units per patient (886 orbital units) to reduce temporal and repeat-visit bias. As bilateral orbits from the same patient were included, Dataset B reduces visit-related confounding but retains inherent inter-eye correlations. Six ML models\u2014Logistic Regression (LR), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), XGBoost, and LightGBM\u2014were trained and evaluated using cross-validation. Model performance was compared in terms of area under the receiver operating characteristic curve (AUC), F1-score, and recall. Feature importance was assessed using RF, XGBoost, and LightGBM.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>Under class-imbalance strategies, Random Forest with class weighting achieved the highest AUC (0.811). Random Forest with SMOTE achieved the highest recall (0.669), F1-score (0.648), and specificity (0.815). Performance on Dataset A demonstrated how unaccounted longitudinal scan correlations can inflate metrics, reinforcing the necessity of temporal deduplication and cautious handling of orbital-level correlations. Feature importance analysis consistently ranked ocular protrusion as the top predictor, followed by rectus muscle thicknesses and orbital geometric parameters.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusion<\/jats:title>\n                    <jats:p>Controlling for longitudinal redundancy and intra-patient correlations significantly impacts model evaluation and generalizability. Random Forest with class weighting demonstrated the best discriminative performance in our internal validation on temporally deduplicated first-visit scans. Rather than relying on isolated diagnostic thresholds, the framework integrates measurable anatomical parameters to generate predictions that complement categorical clinical grading. This approach emphasizes standardized quantification and workflow reproducibility, highlighting the need for rigorous data structuring and transparent feature alignment in medical AI.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/fmed.2026.1840814","type":"journal-article","created":{"date-parts":[[2026,6,19]],"date-time":"2026-06-19T04:29:58Z","timestamp":1781843398000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Interpretable machine learning for severity classification of thyroid eye disease using orbital anatomical features"],"prefix":"10.3389","volume":"13","author":[{"given":"Ruixin","family":"Shi","sequence":"first","affiliation":[{"name":"School of Nursing, Nanjing University of Chinese Medicine","place":["Nanjing, Jiangsu, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Leiming","family":"Gao","sequence":"additional","affiliation":[{"name":"School of Nursing, Nanjing University of Chinese Medicine","place":["Nanjing, Jiangsu, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shengzhi","family":"Jiao","sequence":"additional","affiliation":[{"name":"School of Nursing, Nanjing University of Chinese Medicine","place":["Nanjing, Jiangsu, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Liuzi","family":"Wang","sequence":"additional","affiliation":[{"name":"School of Nursing, Nanjing University of Chinese Medicine","place":["Nanjing, Jiangsu, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jianing","family":"Li","sequence":"additional","affiliation":[{"name":"Department of Nursing, Affiliated Hospital of Traditional Chinese and Western Medicine, Nanjing University of Chinese Medicine","place":["Nanjing, Jiangsu, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Bei","family":"Wang","sequence":"additional","affiliation":[{"name":"Department of Nursing, Affiliated Hospital of Traditional Chinese and Western Medicine, Nanjing University of Chinese Medicine","place":["Nanjing, Jiangsu, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,6,19]]},"reference":[{"key":"ref1","doi-asserted-by":"publisher","first-page":"1124775","DOI":"10.3389\/fcell.2023.1124775","article-title":"Research progress and application of artificial intelligence in thyroid associated ophthalmopathy","volume":"11","author":"Diao","year":"2023","journal-title":"Front Cell Dev Biol"},{"key":"ref2","doi-asserted-by":"publisher","first-page":"73","DOI":"10.3760\/cma.j.cn112142-20221025-00542","article-title":"Interpretation of the key points of the Chinese guidelines for the diagnosis and treatment of thyroid-related eye diseases (2022)","volume":"59","author":"Huifang","year":"2023","journal-title":"Chinese J Ophthalmology"},{"key":"ref3","doi-asserted-by":"publisher","first-page":"281","DOI":"10.1097\/IOP.0000000000002318","article-title":"Evaluation of depression and anxiety in a diverse population with thyroid eye disease using the Nationwide NIH all of us database","volume":"39","author":"Lee","year":"2023","journal-title":"Ophthalmic Plast Reconstr Surg"},{"key":"ref4","doi-asserted-by":"publisher","first-page":"e230247","DOI":"10.1530\/ETJ-23-0247","article-title":"Psychiatric complications in graves' disease","volume":"13","author":"Holmberg","year":"2024","journal-title":"Eur Thyroid J"},{"key":"ref5","doi-asserted-by":"publisher","first-page":"166","DOI":"10.1001\/jamaophthalmol.2022.5665","article-title":"Insight into quality of life in patients with moderate to severe thyroid eye disease","volume":"141","author":"Carter","year":"2023","journal-title":"JAMA Ophthalmol"},{"key":"ref6","doi-asserted-by":"publisher","first-page":"S27","DOI":"10.1210\/clinem\/dgac230","article-title":"Measuring health-related quality of life in thyroid eye disease","volume":"107","author":"Sharma","year":"2022","journal-title":"J Clin Endocrinol Metab"},{"key":"ref7","doi-asserted-by":"publisher","first-page":"68","DOI":"10.1007\/s10792-024-02952-x","article-title":"Evaluation of disease-specific quality of life and its influencing factors in Bulgarian patients with graves' orbitopathy","volume":"44","author":"Stoynova","year":"2024","journal-title":"Int Ophthalmol"},{"key":"ref8","doi-asserted-by":"publisher","first-page":"G43","DOI":"10.1530\/EJE-21-0479","article-title":"The 2021 European group on graves' orbitopathy (EUGOGO) clinical practice guidelines for the medical management of graves' orbitopathy","volume":"185","author":"Bartalena","year":"2021","journal-title":"Eur J Endocrinol"},{"key":"ref9","doi-asserted-by":"publisher","first-page":"249125","DOI":"10.1155\/2015\/249125","article-title":"Graves' Ophthalmopathy: VISA versus EUGOGO classification, assessment, and management","volume":"2015","author":"Barrio-Barrio","year":"2015","journal-title":"J Ophthalmol"},{"key":"ref10","doi-asserted-by":"publisher","first-page":"94","DOI":"10.1016\/j.ajo.2023.03.027","article-title":"Inter-observer variability of clinical activity score: assessments in patients with thyroid eye disease","volume":"252","author":"Perros","year":"2023","journal-title":"Am J Ophthalmol"},{"key":"ref11","doi-asserted-by":"publisher","first-page":"609","DOI":"10.5603\/EP.a2021.0078","article-title":"Orbital decompression for thyroid eye disease: surgical treatment outcomes in endocrinological assessment","volume":"72","author":"Jurek-Matusiak","year":"2021","journal-title":"Endokrynol Pol"},{"key":"ref12","doi-asserted-by":"publisher","first-page":"388","DOI":"10.3390\/diagnostics15030388","article-title":"The value of MRI and Radiomics for the diagnostic evaluation of thyroid-associated Ophthalmopathy","volume":"15","author":"Zhou","year":"2025","journal-title":"Diagnostics"},{"key":"ref13","doi-asserted-by":"publisher","first-page":"206","DOI":"10.1007\/s10792-024-03141-6","article-title":"Asymmetric proptosis in thyroid eye disease","volume":"44","author":"Rana","year":"2024","journal-title":"Int Ophthalmol"},{"key":"ref14","doi-asserted-by":"publisher","first-page":"251","DOI":"10.1080\/01676830.2022.2088807","article-title":"Evolution of asymmetric proptosis during the active phase of thyroid eye disease","volume":"42","author":"Tran","year":"2023","journal-title":"Orbit"},{"key":"ref15","doi-asserted-by":"publisher","first-page":"1537957","DOI":"10.3389\/fendo.2025.1537957","article-title":"Application progress of magnetic resonance imaging in thyroid associated ophthalmopathy","volume":"16","author":"Song","year":"2025","journal-title":"Front Endocrinol"},{"key":"ref16","doi-asserted-by":"publisher","first-page":"321","DOI":"10.13389\/j.cnki.rao.2025.0057","article-title":"Application of imaging in the evaluation of thyroid-related ophthalmopathy","volume":"45","author":"Xiaowei","year":"2025","journal-title":"New Progress Ophthalmology"},{"key":"ref17","doi-asserted-by":"publisher","first-page":"1722","DOI":"10.1136\/bjo-2022-321833","article-title":"Ensemble neural network model for detecting thyroid eye disease using external photographs","volume":"107","author":"Karlin","year":"2023","journal-title":"Br J Ophthalmol"},{"key":"ref18","doi-asserted-by":"publisher","first-page":"1365350","DOI":"10.3389\/fendo.2024.1365350","article-title":"An ensemble deep learning diagnostic system for determining clinical activity scores in thyroid-associated ophthalmopathy: integrating multi-view multimodal images from anterior segment slit-lamp photographs and facial images","volume":"15","author":"Yan","year":"2024","journal-title":"Front Endocrinol (Lausanne)"},{"key":"ref19","doi-asserted-by":"publisher","first-page":"39","DOI":"10.1186\/s12886-020-01783-5","article-title":"Detection of active and inactive phases of thyroid-associated ophthalmopathy using deep convolutional neural network","volume":"21","author":"Lin","year":"2021","journal-title":"BMC Ophthalmol"},{"key":"ref20","doi-asserted-by":"publisher","first-page":"102974","DOI":"10.1016\/j.displa.2025.102974","article-title":"Semi-supervised pre-training based multi-task network for thyroid-associated ophthalmopathy classification","volume":"87","author":"Yang","year":"2025","journal-title":"Displays"},{"key":"ref21","doi-asserted-by":"publisher","first-page":"2640","DOI":"10.3390\/jcm12072640","article-title":"Application of an interpretable machine learning for estimating severity of graves' Orbitopathy based on initial finding","volume":"12","author":"Lee","year":"2023","journal-title":"J Clin Med"},{"key":"ref22","doi-asserted-by":"publisher","first-page":"14","DOI":"10.1007\/s10792-024-03385-2","article-title":"The value of signal intensity ratios of orbital tissue to white matter of orbital MRI in evaluating graves' orbitopathy","volume":"45","author":"Sun","year":"2024","journal-title":"Int Ophthalmol"},{"key":"ref23","doi-asserted-by":"publisher","first-page":"213","DOI":"10.1007\/s10792-024-03138-1","article-title":"Whole-orbit-based multiparametric assessment of disease activity of thyroid eye disease on Dixon MRI","volume":"44","author":"Xia","year":"2024","journal-title":"Int Ophthalmol"},{"key":"ref24","doi-asserted-by":"publisher","first-page":"481","DOI":"10.3969\/j.issn.1673-5765.2024.05.001","article-title":"Dilemmas and opportunities of clinical predictive modelling","volume":"19","author":"Gu","year":"2024","journal-title":"Chinese J Stroke"},{"key":"ref25","doi-asserted-by":"publisher","first-page":"1454","DOI":"10.3969\/j.issn.1674-4055.2018.12.04","article-title":"Clinical prediction models: basic concepts, application scenarios and research ideas","volume":"10","author":"H","year":"2018","journal-title":"Chinese J Evidence-Based Cardiovascular Medicine"},{"key":"ref26","doi-asserted-by":"publisher","first-page":"3374","DOI":"10.1109\/TNNLS.2019.2944455","article-title":"Guide subspace learning for unsupervised domain adaptation","volume":"31","author":"Zhang","year":"2019","journal-title":"IEEE Trans Neural Netw Learn Syst"},{"key":"ref27","doi-asserted-by":"publisher","first-page":"890","DOI":"10.19538\/j.nk2023110103","article-title":"Interpretation of diagnosis and internal medicine treatment in the Chinese guidelines for the diagnosis and treatment of thyroid-related eye diseases (2022)","volume":"43","author":"Y","year":"2023","journal-title":"Chinese J Practical Internal Medicine"},{"key":"ref28","doi-asserted-by":"publisher","first-page":"1439","DOI":"10.1089\/thy.2022.0251","article-title":"Management of Thyroid eye Disease: a consensus statement by the American Thyroid Association and the European thyroid association","volume":"32","author":"Burch","year":"2022","journal-title":"Thyroid"},{"key":"ref29","doi-asserted-by":"publisher","first-page":"1111","DOI":"10.3760\/cma.j.cn115989-20230512-00175","article-title":"Interpretation of the ATA\/ETA consensus on the management of thyroid eye disease","volume":"41","author":"Wei","year":"2023","journal-title":"Chinese J Experimental Ophthalmology"},{"key":"ref30","doi-asserted-by":"publisher","first-page":"12482","DOI":"10.1038\/s41598-025-97565-4","article-title":"Privacy-preserving federated learning for collaborative medical data mining in multi-institutional settings","volume":"15","author":"Haripriya","year":"2025","journal-title":"Sci Rep"},{"key":"ref31","doi-asserted-by":"publisher","first-page":"137","DOI":"10.3233\/SHTI220418","article-title":"Robust random Forest-based all-relevant feature ranks for trustworthy AI","volume":"294","author":"Pfeifer","year":"2022","journal-title":"Stud Health Technol Inform"},{"key":"ref32","doi-asserted-by":"publisher","first-page":"bbad002","DOI":"10.1093\/bib\/bbad002","article-title":"A review on longitudinal data analysis with random forest","volume":"24","author":"Hu","year":"2023","journal-title":"Brief Bioinform"},{"key":"ref33","doi-asserted-by":"publisher","first-page":"103665","DOI":"10.1016\/j.media.2025.103665","article-title":"A novel explainable AI framework for medical image classification integrating statistical, visual, and rule-based methods","volume":"105","author":"Ullah","year":"2025","journal-title":"Med Image Anal"},{"key":"ref34","doi-asserted-by":"publisher","first-page":"749","DOI":"10.1038\/nrclinonc.2017.141","article-title":"Radiomics: the bridge between medical imaging and personalized medicine","volume":"14","author":"Lambin","year":"2017","journal-title":"Nat Rev Clin Oncol"},{"key":"ref35","doi-asserted-by":"publisher","first-page":"9074","DOI":"10.1007\/s00330-023-09931-3","article-title":"Quantitative assessment of extraocular muscles in graves' ophthalmopathy using T1 mapping","volume":"33","author":"Zhu","year":"2023","journal-title":"Eur Radiol"},{"key":"ref36","doi-asserted-by":"publisher","first-page":"16939","DOI":"10.1038\/s41598-023-43942-w","article-title":"Effects of various extraocular muscle enlargement patterns on muscle diameter index in graves ophthalmopathy patients: a retrospective cohort study","volume":"13","author":"Yang","year":"2023","journal-title":"Sci Rep"},{"key":"ref37","doi-asserted-by":"publisher","first-page":"2651","DOI":"10.1007\/s00417-024-06424-x","article-title":"Investigation of lacrimal gland and extraocular muscle in thyroid eye disease patients with severe subjective dry eye disease: a multiparametric magnetic resonance imaging study","volume":"262","author":"Liao","year":"2024","journal-title":"Graefes Arch Clin Exp Ophthalmol"},{"key":"ref38","doi-asserted-by":"publisher","first-page":"10","DOI":"10.1080\/01676830.2025.2528979","article-title":"Correlation between quantified magnetic resonance imaging (MRI) findings and \"MRI score\" in graves' ophthalmopathy","volume":"45","author":"Ueki","year":"2025","journal-title":"Orbit"},{"key":"ref39","doi-asserted-by":"publisher","first-page":"936819","DOI":"10.3389\/fmed.2022.936819","article-title":"Risk prediction of dysthyroid optic neuropathy based on CT imaging features combined the bony orbit with the soft tissue structures","volume":"9","author":"Cheng","year":"2022","journal-title":"Front Med"},{"key":"ref40","doi-asserted-by":"publisher","first-page":"77","DOI":"10.1080\/01676830.2019.1600150","article-title":"Visual field and orbital computed tomography correlation in dysthyroid optic neuropathy due to thyroid eye disease","volume":"39","author":"Starks","year":"2020","journal-title":"Orbit"},{"key":"ref41","doi-asserted-by":"publisher","first-page":"4419","DOI":"10.4103\/ijo.IJO_719_22","article-title":"Dysthyroid optic neuropathy: demographics, risk factors, investigations, and management outcomes","volume":"70","author":"Poonam","year":"2022","journal-title":"Indian J Ophthalmol"},{"key":"ref42","doi-asserted-by":"publisher","first-page":"2353","DOI":"10.1158\/0008-5472","article-title":"Paving a path to clinical impact with Radiomics: enabling reproducibility and reach","volume":"85","author":"El-Jammal","year":"2025","journal-title":"Cancer Res"},{"key":"ref43","doi-asserted-by":"publisher","first-page":"9","DOI":"10.1016\/j.ejmp.2021.02.006","article-title":"AI applications to medical images: from machine learning to deep learning","volume":"83","author":"Castiglioni","year":"2021","journal-title":"Phys Med"},{"key":"ref44","doi-asserted-by":"publisher","first-page":"6","DOI":"10.1167\/iovs.65.5.6","article-title":"Deep-learning based automated segmentation and quantitative volumetric analysis of orbital muscle and fat for diagnosis of thyroid eye disease","volume":"65","author":"Alkhadrawi","year":"2024","journal-title":"Invest Ophthalmol Vis Sci"},{"key":"ref45","doi-asserted-by":"publisher","first-page":"130","DOI":"10.1186\/s41747-024-00529-y","article-title":"Image biomarkers and explainable AI: handcrafted features versus deep learned features","volume":"8","author":"Rundo","year":"2024","journal-title":"Eur Radiol Exp"}],"container-title":["Frontiers in Medicine"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fmed.2026.1840814\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,19]],"date-time":"2026-06-19T04:30:03Z","timestamp":1781843403000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fmed.2026.1840814\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6,19]]},"references-count":45,"alternative-id":["10.3389\/fmed.2026.1840814"],"URL":"https:\/\/doi.org\/10.3389\/fmed.2026.1840814","relation":{},"ISSN":["2296-858X"],"issn-type":[{"value":"2296-858X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,6,19]]},"article-number":"1840814"}}