{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,3]],"date-time":"2026-09-03T07:16:08Z","timestamp":1788419768499,"version":"build-2803163510"},"reference-count":24,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,9,3]],"date-time":"2026-09-03T00:00:00Z","timestamp":1788393600000},"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. Surg."],"abstract":"<jats:sec>\n                    <jats:title>Purpose<\/jats:title>\n                    <jats:p>This study introduces a novel multi-modal transformer-based framework that integrates clinical data, radiomic features, and deep learning representations for automated classification of laryngeal lesions from contact endoscopy images.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>\n                      We retrospectively enrolled 300 patients with laryngeal lesions from three independent medical centers (Centers A, B, and C), acquiring 7,847 contact endoscopy images with narrow band imaging. Clinical variables (18 features), radiomic features extracted using PyRadiomics (156 features), and deep learning representations from three state-of-the-art architectures, ConvNeXt V2, Swin Transformer V2, and EfficientNet V2 (3,328 combined features), were systematically integrated through a six-layer transformer encoder with multi-head cross-attention mechanisms. We evaluated three feature selection strategies (LASSO, mutual information, ReliefF) and three advanced classification architectures (Vision Transformer, Attention-MLP, Graph Neural Network). The framework was developed using data from Centers A and B (\n                      <jats:italic>n<\/jats:italic>\n                      \u2009=\u2009230 patients) with rigorous five-fold cross-validation, and validated on an independent external test set from Center C (\n                      <jats:italic>n<\/jats:italic>\n                      \u2009=\u200970 patients).\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>The optimal configuration (LASSO feature selection with Vision Transformer classifier) achieved accuracy of 96.1\u2009\u00b1\u20091.0% (AUC-ROC: 0.992\u2009\u00b1\u20090.007) on training, 93.8\u2009\u00b1\u20091.4% (AUC-ROC: 0.981\u2009\u00b1\u20090.012) on internal validation, and 91.4\u2009\u00b1\u20092.1% (AUC-ROC: 0.967\u2009\u00b1\u20090.019) on external testing, with sensitivity of 90.0\u2009\u00b1\u20092.8% and specificity of 92.5\u2009\u00b1\u20092.5% on the external cohort. The multi-modal fusion approach significantly outperformed all single-modality<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>\n                      clinical features alone (71.3% validation accuracy), radiomic features alone (84.2%), and the best individual deep learning model (87.6%), with improvements of 22.5, 9.6, and 6.2 percentage points respectively (all\n                      <jats:italic>p<\/jats:italic>\n                      \u2009&amp;lt;\u20090.001. Attention mechanism analysis revealed that the model dynamically weighted modality contributions, allocating 47.0% attention to deep features, 34.7% to radiomic features, and 18.3% attention to clinical features for correctly classified malignant cases.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusion<\/jats:title>\n                    <jats:p>This study presents the first comprehensive framework for laryngeal lesion classification that systematically integrates clinical data, radiomics, and deep learning through transformer-based fusion.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/fsurg.2026.1840300","type":"journal-article","created":{"date-parts":[[2026,9,3]],"date-time":"2026-09-03T06:41:35Z","timestamp":1788417695000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["A transformer-based multi-modal fusion framework for laryngeal lesion classification using contact endoscopy"],"prefix":"10.3389","volume":"13","author":[{"given":"Yunjing","family":"Duan","sequence":"first","affiliation":[{"name":"Department of Otolaryngology \u2013 Head and Neck Surgery, Shijiazhuang People's Hospital","place":["Shijiazhuang, Hebei, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Wei","family":"Li","sequence":"additional","affiliation":[{"name":"Department of Otolaryngology \u2013 Head and Neck Surgery, Shijiazhuang People's Hospital","place":["Shijiazhuang, Hebei, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,9,3]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"e472464","DOI":"10.1200\/EDBK-25-472464","article-title":"Artificial intelligence for head and neck squamous cell carcinoma: from diagnosis to treatment","volume":"45","author":"Song","year":"2025","journal-title":"Am Soc Clin Oncol Educ Book"},{"key":"B2","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1016\/j.isci.2025.114231","article-title":"A deep learning fusion network trained with medical records and laryngoscopic images in the early diagnosis of glottic carcinoma","volume":"28","author":"Shuai","year":"2025","journal-title":"iScience"},{"key":"B3","doi-asserted-by":"publisher","first-page":"16239","DOI":"10.1038\/s41598-025-01270-1","article-title":"Multiparameter MRI-based model integrating radiomics and deep learning for preoperative staging of laryngeal squamous cell carcinoma","volume":"15","author":"Xie","year":"2025","journal-title":"Sci Rep"},{"key":"B4","doi-asserted-by":"publisher","first-page":"1618349","DOI":"10.3389\/fonc.2025.1618349","article-title":"Enhancing clinical diagnosis of laryngeal cancer through fusion-based transfer learning with osprey optimisation algorithm using histology images","volume":"15","author":"Al-Kahtani","year":"2025","journal-title":"Front Oncol"},{"key":"B5","doi-asserted-by":"publisher","first-page":"92","DOI":"10.3892\/br.2024.1781","article-title":"Applications and challenges of neural networks in otolaryngology (review)","volume":"20","author":"Taciuc","year":"2024","journal-title":"Biomed Rep"},{"key":"B6","doi-asserted-by":"publisher","first-page":"2148","DOI":"10.1007\/s12672-025-03992-0","article-title":"Artificial intelligence in head and neck cancer: a bibliometric and visualization analysis (1995\u20132025)","volume":"16","author":"Cai","year":"2025","journal-title":"Discover Oncology"},{"key":"B7","doi-asserted-by":"publisher","first-page":"3285","DOI":"10.3390\/cancers16193285","article-title":"Deep learning for image analysis in the diagnosis and management of esophageal cancer","volume":"16","author":"Theocharopoulos","year":"2024","journal-title":"Cancers (Basel)"},{"key":"B8","doi-asserted-by":"publisher","first-page":"7639","DOI":"10.2147\/JIR.S484485","article-title":"Deep-learning, radiomics and clinic based fusion models for predicting response to infliximab in crohn\u2019s disease patients: a multicentre, retrospective study","volume":"17","author":"Cai","year":"2024","journal-title":"J Inflamm Res"},{"key":"B9","doi-asserted-by":"publisher","first-page":"e14559","DOI":"10.1002\/acm2.14559","article-title":"Machine learning in image-based outcome prediction after radiotherapy: a review","volume":"26","author":"Yuan","year":"2025","journal-title":"J Appl Clin Med Phys"},{"key":"B10","author":"Kuo","year":""},{"key":"B11","doi-asserted-by":"publisher","first-page":"102060","DOI":"10.1016\/j.labinv.2024.102060","article-title":"Advancing precision medicine: algebraic topology and differential geometry in radiology and computational pathology","volume":"104","author":"Levenson","year":"2024","journal-title":"Lab Invest"},{"key":"B12","doi-asserted-by":"publisher","first-page":"813","DOI":"10.3390\/bioengineering12080813","article-title":"CT-Based Habitat radiomics combining multi-instance learning for early prediction of post-neoadjuvant lymph node metastasis in esophageal squamous cell carcinoma","volume":"12","author":"Peng","year":"2025","journal-title":"Bioengineering"},{"key":"B13","doi-asserted-by":"publisher","first-page":"1845","DOI":"10.3390\/jcm14061845","article-title":"Current role of artificial intelligence in the management of esophageal cancer","volume":"14","author":"Mela","year":"2025","journal-title":"J Clin Med"},{"key":"B14","doi-asserted-by":"publisher","first-page":"277","DOI":"10.1186\/s13244-024-01851-0","article-title":"Multimodality deep learning radiomics predicts pathological response after neoadjuvant chemoradiotherapy for esophageal squamous cell carcinoma","volume":"15","author":"Liu","year":"2024","journal-title":"Insights Imaging"},{"key":"B15","doi-asserted-by":"publisher","first-page":"803","DOI":"10.1016\/j.otc.2024.05.001","article-title":"Artificial intelligence in head and neck surgery","volume":"57","author":"Oliver","year":"2024","journal-title":"Otolaryngol Clin North Am"},{"key":"B16","doi-asserted-by":"publisher","first-page":"194","DOI":"10.1007\/s12672-025-04359-1","article-title":"Graph neural network-based multi-label classification of breast cancer using multiparametric MRI and clinical features","volume":"17","author":"Salehzehi","year":"2025","journal-title":"Discover Oncology"},{"key":"B17","doi-asserted-by":"publisher","first-page":"91","DOI":"10.1186\/s12880-026-02173-x","article-title":"Deep learning-based multimodal fusion of MRI and whole slide image for predicting neoadjuvant therapy response in locally advanced head and neck squamous cell carcinoma","volume":"26","author":"Kang","year":"2026","journal-title":"BMC Med Imaging"},{"key":"B18","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1016\/j.ebiom.2025.105663","article-title":"Deep learning informed multimodal fusion of radiology and pathology to predict outcomes in HPV-associated oropharyngeal squamous cell carcinoma","volume":"114","author":"Song","year":"2025","journal-title":"EBioMedicine"},{"key":"B19","doi-asserted-by":"publisher","first-page":"e74902","DOI":"10.2196\/74902","article-title":"A vision-language\u2013guided multimodal fusion network for glottic carcinoma early diagnosis: model development and validation study","volume":"13","author":"Jin","year":"2025","journal-title":"JMIR Med Inform"},{"key":"B20","author":"Farhadizadeh","year":""},{"key":"B21","doi-asserted-by":"publisher","first-page":"698","DOI":"10.1186\/s12967-023-04572-y","article-title":"Real-time detection of laryngopharyngeal cancer using an artificial intelligence-assisted system with multimodal data","volume":"21","author":"Li","year":"2023","journal-title":"J Transl Med"},{"key":"B22","doi-asserted-by":"publisher","first-page":"8157","DOI":"10.3390\/s21238157","article-title":"Deep convolution neural network for laryngeal cancer classification on contact endoscopy-narrow band imaging","volume":"21","author":"Esmaeili","year":"2021","journal-title":"Sensors"},{"key":"B23","doi-asserted-by":"publisher","first-page":"1798","DOI":"10.1002\/lary.29960","article-title":"Deep learning applied to white light and narrow band imaging videolaryngoscopy: toward real-time laryngeal cancer detection","volume":"132","author":"Azam","year":"2022","journal-title":"Laryngoscope"},{"key":"B24","doi-asserted-by":"publisher","first-page":"1582011","DOI":"10.3389\/fonc.2025.1582011","article-title":"Comparison of clinical nasal endoscopy, optical biopsy, and artificial intelligence in early diagnosis and treatment planning in laryngeal cancer: a prospective observational study","volume":"15","author":"Hu","year":"2025","journal-title":"Front Oncol"}],"container-title":["Frontiers in Surgery"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fsurg.2026.1840300\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,3]],"date-time":"2026-09-03T06:41:40Z","timestamp":1788417700000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fsurg.2026.1840300\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,3]]},"references-count":24,"alternative-id":["10.3389\/fsurg.2026.1840300"],"URL":"https:\/\/doi.org\/10.3389\/fsurg.2026.1840300","relation":{},"ISSN":["2296-875X"],"issn-type":[{"value":"2296-875X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,3]]},"article-number":"1840300"}}