{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,15]],"date-time":"2026-07-15T18:13:04Z","timestamp":1784139184733,"version":"3.55.0"},"reference-count":19,"publisher":"Public Library of Science (PLoS)","issue":"7","license":[{"start":{"date-parts":[[2026,7,15]],"date-time":"2026-07-15T00:00:00Z","timestamp":1784073600000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"SERB, DST, Government of India","award":["EEQ\/2023\/000053"],"award-info":[{"award-number":["EEQ\/2023\/000053"]}]},{"DOI":"10.13039\/501100004242","name":"Princess Nourah Bint Abdulrahman University","doi-asserted-by":"publisher","award":["PNURSP2026R300"],"award-info":[{"award-number":["PNURSP2026R300"]}],"id":[{"id":"10.13039\/501100004242","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["www.plosone.org"],"crossmark-restriction":false},"short-container-title":["PLoS One"],"abstract":"<jats:sec id=\"sec001\">\n                    <jats:title>Background<\/jats:title>\n                    <jats:p>Diagnosing breast cancer with mammography continues to be an incidentally challenging process because of the numerous different ways of acquiring the images, the vast difference in appearance between lesions, and the minimal use of historical information regarding patients being considered when making a diagnosis. In addition, many current deep learning algorithms use static weight values based on the most recent image taken only, often failing to relate a current case to similar historical examples and to incorporate retrieval-based contextual evidence during inference.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec id=\"sec002\">\n                    <jats:title>Introduction<\/jats:title>\n                    <jats:p>ViT-MultiRAGNet is a retrieval-augmented Vision Transformer framework designed to integrate multi-view 2D mammographic features with memory-bank-based contextual evidence from historically similar cases. The proposed approach enhances diagnostic accuracy by combining global 2D mammographic features, semantic lesion representations, and historical case evidence retrieved from a fold-wise memory bank during inference.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec id=\"sec003\">\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>Multi-view 2D mammograms were processed using a Vision Transformer encoder to extract global contextual features, while a retrieval-augmented memory bank was used to identify diagnostically similar historical cases for evidence-guided feature enrichment. A Retrieval-Augmented Generation (RAG) module retrieves the top-k most similar feature embeddings from a fold-wise training memory bank and combines them with the current query representation using multi-head cross-attention. The evaluation of the models included utilizing accuracy, area under the receiver operating curve (AUC-ROC), Dice Coefficient, and F1 score metrics at a 95% bootstrapped confidence interval using five-fold stratified cross-validation, with findings assessed using Wilcoxon signed-rank statistical tests.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec id=\"sec004\">\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>The accuracy of ViT-MultiRAGNet is calculated as a value of 0.978\u2009\u00b1\u20090.009 and AUC-ROC 0.998\u2009\u00b1\u20090.009 (Wilcoxon p\u2009&lt;\u20090.01), which is significantly higher than unimodal benchmarks and fusion techniques on the RTM dataset. The calculations for the CBIS-DDSM evaluation were 0.961\u2009\u00b1\u20090.011 for accuracy and 0.989\u2009\u00b1\u20090.010 for AUC-ROC. The calculations for segmentation evaluation were 0.882 (RTM) and 0.795 (CBIS-DDSM), however both have a higher degree of alignment at the edge. The RAG mechanism improved retrieval-guided fusion compared with simple feature concatenation while maintaining efficient inference performance of 0.31 seconds per image.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec id=\"sec005\">\n                    <jats:title>Conclusion<\/jats:title>\n                    <jats:p>The combination of ViT-based global 2D mammographic representation, retrieval-augmented fusion, and memory-guided contextual evidence improves diagnostic performance, robustness, and clinical interpretability. This Framework appears to be a good candidate for an Evidence-based Decision Support System which will provide a transparent method of helping clinicians make decisions regarding screening and diagnosis of breast cancer.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1371\/journal.pone.0349865","type":"journal-article","created":{"date-parts":[[2026,7,15]],"date-time":"2026-07-15T17:49:39Z","timestamp":1784137779000},"page":"e0349865","update-policy":"https:\/\/doi.org\/10.1371\/journal.pone.corrections_policy","source":"Crossref","is-referenced-by-count":0,"title":["ViT-MultiRAGNet: A scalable and reliable retrieval-augmented Vision Transformer framework for memory-guided feature fusion multi-modal mammogram classification"],"prefix":"10.1371","volume":"21","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-8957-8532","authenticated-orcid":true,"given":"N. Thirupathi","family":"Rao","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-3846-8269","authenticated-orcid":true,"given":"CH V. V.","family":"Ramana","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Faten Khalid","family":"Karim","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-3909-2227","authenticated-orcid":true,"given":"Ayman","family":"Aljarbouh","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9234-5898","authenticated-orcid":true,"given":"Samih M.","family":"Mostafa","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"340","published-online":{"date-parts":[[2026,7,15]]},"reference":[{"key":"pone.0349865.ref001","article-title":"An image is worth 16x16 words: transformers for image recognition at scale","author":"A Dosovitskiy","year":"2020","journal-title":"International Conference on Learning Representations (ICLR)"},{"key":"pone.0349865.ref002","article-title":"Retrieval-augmented generation for knowledge-intensive NLP tasks","author":"P Lewis","year":"2020","journal-title":"Neural Information Processing Systems (NeurIPS)"},{"key":"pone.0349865.ref003","article-title":"Swin transformer: hierarchical Vision Transformer using shifted windows","author":"Z Liu","year":"2021","journal-title":"IEEE\/CVF International Conference on Computer Vision (ICCV)"},{"key":"pone.0349865.ref004","article-title":"Transformers improve breast cancer diagnosis from multi-view mammograms","author":"X Chen","year":"2022","journal-title":"Med Image Anal"},{"key":"pone.0349865.ref005","article-title":"Transformer-based multi-modal learning for breast cancer diagnosis","author":"X Wang","year":"2025","journal-title":"Artif Intell Med"},{"key":"pone.0349865.ref006","article-title":"Cross-attention guided loss-based deep dual-branch transformer network for breast cancer segmentation and classification","author":"BE Enes","year":"2024","journal-title":"IEEE Trans Med Imaging"},{"key":"pone.0349865.ref007","article-title":"Detection of breast cancer in digital breast tomosynthesis using Vision Transformers","author":"I Kassis","year":"2024","journal-title":"Med Phys"},{"key":"pone.0349865.ref008","article-title":"Evaluating deep learning models on large-scale histopathological datasets","author":"S Eskandari","year":"2024","journal-title":"Comput Methods Pathol"},{"issue":"2","key":"pone.0349865.ref009","doi-asserted-by":"crossref","first-page":"178","DOI":"10.3390\/diagnostics13020178","article-title":"Vision-transformer-based transfer learning for mammogram classification","volume":"13","author":"G Ayana","year":"2023","journal-title":"Diagnostics (Basel)"},{"issue":"4","key":"pone.0349865.ref010","doi-asserted-by":"crossref","first-page":"461","DOI":"10.3390\/analytics3040026","article-title":"Breast cancer classification using fine-tuned SWIN transformer model on mammographic images","volume":"3","author":"O Tanimola","year":"2024","journal-title":"Analytics"},{"key":"pone.0349865.ref011","article-title":"MV-Swin-T: mammogram classification with multi-view swin transformer","author":"S Sarker","year":"2024","journal-title":"Proc IEEE Int Symp Biomed Imaging"},{"issue":"2","key":"pone.0349865.ref012","doi-asserted-by":"crossref","first-page":"121","DOI":"10.11648\/j.ajcst.20250802.16","article-title":"Transformer-based explainable deep learning for breast cancer detection in mammography: the MammoFormer framework","volume":"8","author":"O Peter","year":"2025","journal-title":"AJCST"},{"key":"pone.0349865.ref013","unstructured":"Osman S, Turzo M, Rusho M, Haider M, Sajin S, Behesti H, et al. Fine-tuned transformer models for breast cancer detection and classification. 2025. https:\/\/doi.org\/10.13140\/RG.2.2.10874.81601\/1"},{"key":"pone.0349865.ref014","doi-asserted-by":"crossref","first-page":"6562","DOI":"10.1038\/s41598-025-89451-w","article-title":"An enhanced denoising system for mammogram images using deep transformer model with fusion of local and global features","volume":"15","author":"AR Singh","year":"2025","journal-title":"Sci Rep"},{"issue":"4","key":"pone.0349865.ref015","doi-asserted-by":"crossref","first-page":"461","DOI":"10.3390\/analytics3040026","article-title":"Breast cancer classification using fine-tuned SWIN transformer model on mammographic images","volume":"3","author":"O Tanimola","year":"2024","journal-title":"Analytics"},{"issue":"04","key":"pone.0349865.ref016","doi-asserted-by":"crossref","first-page":"1298","DOI":"10.47857\/irjms.2025.v06i04.07120","article-title":"An enhanced breast cancer detection in mammograms using Vision Transformers and data augmentation","volume":"06","author":"D Sherine H","year":"2025","journal-title":"IRJMS"},{"issue":"4","key":"pone.0349865.ref017","first-page":"319","article-title":"Advanced mammography based breast cancer classification using Vision Transformer and swin models: a study on ddsm dataset","volume":"18","author":"H Meylem","year":"2025","journal-title":"J Info Technol"},{"key":"pone.0349865.ref018","article-title":"TT-stack: a transformer-based tiered-stacking ensemble framework with meta-learning for automated breast cancer detection in mammography.","author":"S Osman","year":"2025"},{"issue":"1","key":"pone.0349865.ref019","first-page":"1","article-title":"A curated mammography data set for use in computer-aided detection and diagnosis research","volume":"4","author":"RS Lee","year":"2017","journal-title":"Sci Data"}],"container-title":["PLOS One"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pone.0349865","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,15]],"date-time":"2026-07-15T17:49:48Z","timestamp":1784137788000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pone.0349865"}},"subtitle":[],"editor":[{"given":"Taikyeong Ted","family":"Jeong","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"editor"}]}],"short-title":[],"issued":{"date-parts":[[2026,7,15]]},"references-count":19,"journal-issue":{"issue":"7","published-online":{"date-parts":[[2026,7,15]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pone.0349865","relation":{},"ISSN":["1932-6203"],"issn-type":[{"value":"1932-6203","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7,15]]}}}