{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,13]],"date-time":"2026-08-13T04:30:40Z","timestamp":1786595440959,"version":"build-2736575974"},"reference-count":27,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,8,13]],"date-time":"2026-08-13T00:00:00Z","timestamp":1786579200000},"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>Introduction<\/jats:title>\n                    <jats:p>The automated classification of dermoscopic skin lesions is inherently challenging due to pronounced class imbalance, minimal inter-class variance, visual similarity across lesion types, and the requirement for clinically interpretable predictive outcomes.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>The present study designed a heterogeneous Vision Transformer ensemble framework for seven-class skin lesion classification using the HAM10000 dataset. The framework integrates three architecturally distinct backbones Swin-Tiny, ViT-Base, and DeiT-Small enhanced with a novel Regional Attention Wrapper (RAW) for spatially selective feature aggregation. The generated outputs are combined via a stacking protocol wherein a trained MLP meta-learner resolves class-aware disagreements among the models. Class imbalance is addressed using class-adaptive augmentation with class-weighted focal loss and MixUp regularisation.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>\n                      The proposed framework achieved 98.37% accuracy, weighted F1-score of 98.39%, and mean AUC of 0.999, surpassing all three individual backbones across all metrics. MEL misclassifications were reduced by 78% compared to the weakest baseline, confirmed by McNemar's test (\n                      <jats:italic>p<\/jats:italic>\n                      &amp;lt; 0.0001). Comprehensive ablation studies validate the contributions of the RAW module, each ensemble component, and each augmentation strategy. Three-fold cross-validation yields a mean accuracy of 96.14 \u00b1 0.36% and bootstrap confidence interval analysis confirms the reliability and reproducibility of the reported results.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Discussion<\/jats:title>\n                    <jats:p>An exhaustive explainability framework comprising Regional Attention Maps, GradCAM++, SHAP, and t-SNE provides complementary spatial, gradient-based, pixel-level, and embedding-level interpretability, ensuring clinical trust, transparency, and trustworthiness expected from an automated dermoscopy system.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/fmed.2026.1864593","type":"journal-article","created":{"date-parts":[[2026,8,13]],"date-time":"2026-08-13T04:23:59Z","timestamp":1786595039000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Multi-paradigm Vision Transformer ensemble with regional attention and MLP meta-fusion for explainable dermoscopic skin lesion classification"],"prefix":"10.3389","volume":"13","author":[{"given":"Nemili","family":"Sravani","sequence":"first","affiliation":[{"name":"School of Computer Science Engineering and Information Systems, Vellore Institute of Technology","place":["Vellore, India"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Srinivas","family":"Koppu","sequence":"additional","affiliation":[{"name":"School of Computer Science Engineering and Information Systems, Vellore Institute of Technology","place":["Vellore, India"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,8,13]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"110187","DOI":"10.1016\/j.bspc.2026.110187","article-title":"Developing an advanced deep learning framework for skin cancer detection and classification: empowering healthcare professionals with enhanced diagnostic tools and decision support","volume":"120","author":"Baladhandapani","year":"2026","journal-title":"Biomed Signal Process Control"},{"key":"B2","doi-asserted-by":"publisher","first-page":"1030","DOI":"10.3390\/diagnostics16071030","article-title":"Unveiling the importance of early detection of oral mucosal melanoma with non-invasive imaging techniques","volume":"16","author":"B\u0103l\u0103ceanu-Gur\u0103u","year":"2026","journal-title":"Diagnostics"},{"key":"B3","doi-asserted-by":"publisher","first-page":"64","DOI":"10.1007\/s40747-025-02179-y","article-title":"Semantic segmentation assisted deep ensemble feature learning model for skin-cancer detection and classification: SDENet","volume":"12","author":"Srilakshmi","year":"2026","journal-title":"Complex Intell Syst"},{"key":"B4","doi-asserted-by":"publisher","first-page":"181009","DOI":"10.1109\/ACCESS.2024.3510224","article-title":"Multi-model attentional fusion ensemble for accurate skin cancer classification","volume":"12","author":"Ahmed","year":"2024","journal-title":"IEEE Access"},{"key":"B5","doi-asserted-by":"publisher","first-page":"41405","DOI":"10.1007\/s11042-025-20797-z","article-title":"LesionAid: vision transformers-based skin lesion generation and classification-A practical review","volume":"84","author":"Rao","year":"2025","journal-title":"Multimed Tools Appl"},{"key":"B6","doi-asserted-by":"publisher","first-page":"460","DOI":"10.1007\/s44163-026-01162-x","article-title":"An advanced healthcare system with an automated ViT model for dermoscopic skin cancer identification","volume":"6","author":"Abugabah","year":"2026","journal-title":"Discov Artif Intell"},{"key":"B7","doi-asserted-by":"publisher","first-page":"415","DOI":"10.3390\/info14070415","article-title":"Multi-class skin cancer classification using vision transformer networks and convolutional neural network-based pre-trained models","volume":"14","author":"Arshed","year":"2023","journal-title":"Information"},{"key":"B8","doi-asserted-by":"publisher","first-page":"108787","DOI":"10.1016\/j.compbiolchem.2025.108787","article-title":"Vit-Ensemble: probabilistic voting based ensemble of Vision Transformers for tuberculosis detection using radiographs","volume":"120","author":"Pradhan","year":"2025","journal-title":"Comput Biol Chem"},{"key":"B9","doi-asserted-by":"publisher","first-page":"68","DOI":"10.1109\/TTS.2023.3234203","article-title":"Challenges of deep learning in medical image analysis\u2013improving explainability and trust","volume":"4","author":"Dhar","year":"2023","journal-title":"IEEE Trans Technol Soc"},{"key":"B10","doi-asserted-by":"publisher","first-page":"67","DOI":"10.1016\/j.nbt.2022.05.002","article-title":"Explainability and causability for artificial intelligence-supported medical image analysis in the context of the European In Vitro Diagnostic Regulation","volume":"70","author":"M\u00fcller","year":"2022","journal-title":"New Biotechnol"},{"key":"B11","doi-asserted-by":"publisher","first-page":"2227","DOI":"10.1007\/s10278-023-00862-5","article-title":"Deep learning-based skin lesion multi-class classification with global average pooling improvement","volume":"36","author":"Raghavendra","year":"2023","journal-title":"J Digit Imaging"},{"key":"B12","doi-asserted-by":"publisher","first-page":"28850","DOI":"10.1038\/s41598-024-80087-w","article-title":"Enhanced MobileNet for skin cancer image classification with fused spatial channel attention mechanism","volume":"14","author":"Cheng","year":"2024","journal-title":"Sci Rep"},{"key":"B13","doi-asserted-by":"publisher","first-page":"108929","DOI":"10.1016\/j.bspc.2025.108929","article-title":"Cutting-edge CNN-based skin cancer detection with batch normalization and advanced imbalance learning for superior medical image classification","volume":"113","author":"Govindu","year":"2026","journal-title":"Biomed Signal Process Control"},{"key":"B14","doi-asserted-by":"publisher","first-page":"25709","DOI":"10.1007\/s11042-024-19864-8","article-title":"Melanoma skin cancer detection based on deep learning methods and binary Harris Hawk optimization","volume":"84","author":"Jaber","year":"2025","journal-title":"Multimed Tools Appl"},{"key":"B15","doi-asserted-by":"publisher","first-page":"4633","DOI":"10.1038\/s41598-024-82241-w","article-title":"A comprehensive analysis of deep learning and transfer learning techniques for skin cancer classification","volume":"15","author":"Shakya","year":"2025","journal-title":"Sci Rep"},{"key":"B16","doi-asserted-by":"publisher","first-page":"45660","DOI":"10.1038\/s41598-025-30534-z","article-title":"Multimodal deep learning ensemble framework for skin cancer detection","volume":"15","author":"Saeed","year":"2025","journal-title":"Sci Rep"},{"key":"B17","doi-asserted-by":"publisher","first-page":"4938","DOI":"10.1038\/s41598-025-89230-7","article-title":"A robust deep learning framework for multiclass skin cancer classification","volume":"15","author":"Ozdemir","year":"2025","journal-title":"Sci Rep"},{"key":"B18","doi-asserted-by":"publisher","first-page":"40204","DOI":"10.1038\/s41598-025-24081-w","article-title":"Transformer-aided skin cancer classification using VGG19-based feature encoding","volume":"15","author":"Najjar","year":"2025","journal-title":"Sci Rep"},{"key":"B19","doi-asserted-by":"publisher","first-page":"4005","DOI":"10.3390\/app14104005","article-title":"SkinSwinViT: a lightweight transformer-based method for multiclass skin lesion classification with enhanced generalization capabilities","volume":"14","author":"Tang","year":"2024","journal-title":"Appl Sci"},{"key":"B20","doi-asserted-by":"publisher","first-page":"107819","DOI":"10.1016\/j.bspc.2025.107819","article-title":"White patchy skin lesion classification using feature enhancement and interaction transformer module","volume":"107","author":"Li","year":"2025","journal-title":"Biomed Signal Process Control"},{"key":"B21","doi-asserted-by":"publisher","first-page":"180161","DOI":"10.1038\/sdata.2018.161","article-title":"The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions","volume":"5","author":"Tschandl","year":"2018","journal-title":"Sci Data"},{"key":"B22","first-page":"10012","article-title":"Swin transformer: hierarchical vision transformer using shifted windows","volume-title":"Proceedings of the IEEE\/CVF international conference on computer vision","author":"Liu","year":"2021"},{"key":"B23","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2010.11929","article-title":"An image is worth 16x16 words: Transformers for image recognition at scale","author":"Dosovitskiy","year":"2020","journal-title":"arXiv"},{"key":"B24","first-page":"10347","article-title":"Training data-efficient image transformers & distillation through attention","volume-title":"International conference on machine learning","author":"Touvron","year":"2021"},{"key":"B25","doi-asserted-by":"publisher","first-page":"8","DOI":"10.69882\/adba.ai.2026012","article-title":"Comparative evaluation of convolutional neural network architectures for automated skin cancer classification: a study on the ISIC 2018 dataset","volume":"2","author":"Kingni","year":"2026","journal-title":"Artif Intell Appl Sci"},{"key":"B26","doi-asserted-by":"publisher","first-page":"32","DOI":"10.69882\/adba.ai.2026015","article-title":"Bridging the gap between theoretical performance and clinical utility in multi-class skin lesion diagnosis","volume":"2","author":"S\u00f6nmez","year":"2026","journal-title":"Artif Intell Appl Sci"},{"key":"B27","unstructured":"Tschandl\n              P\n            \n          \n          Skin Cancer MNIST: HAM10000\n          \n          2018"}],"container-title":["Frontiers in Medicine"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fmed.2026.1864593\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,13]],"date-time":"2026-08-13T04:24:01Z","timestamp":1786595041000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fmed.2026.1864593\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,8,13]]},"references-count":27,"alternative-id":["10.3389\/fmed.2026.1864593"],"URL":"https:\/\/doi.org\/10.3389\/fmed.2026.1864593","relation":{},"ISSN":["2296-858X"],"issn-type":[{"value":"2296-858X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,8,13]]},"article-number":"1864593"}}