{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,24]],"date-time":"2026-08-24T11:25:44Z","timestamp":1787570744120,"version":"build-2736575974"},"reference-count":30,"publisher":"Oxford University Press (OUP)","issue":"Supplement_2","license":[{"start":{"date-parts":[[2026,8,24]],"date-time":"2026-08-24T00:00:00Z","timestamp":1787529600000},"content-version":"vor","delay-in-days":23,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100001592","name":"Higher Education Authority","doi-asserted-by":"publisher","award":["10121436"],"award-info":[{"award-number":["10121436"]}],"id":[{"id":"10.13039\/501100001592","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001592","name":"Higher Education Authority","doi-asserted-by":"publisher","award":["101233450"],"award-info":[{"award-number":["101233450"]}],"id":[{"id":"10.13039\/501100001592","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,8,1]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:sec>\n                    <jats:title>Motivation<\/jats:title>\n                    <jats:p>Perineural invasion (PNI) is an important pathological phenotype associated with poor prognosis in multiple malignancies. The primary detection method is visual inspection of whole slide images (WSIs), which is labor-intensive, time-consuming, subjective, and prone to high inter-observer variability. Developing reliable, accurate deep learning models for PNI detection is constrained by the lack of pixel-level annotated WSIs.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>We evaluated four backbone architectures and two different approaches to improve PNI detection in a multi-organ dataset of colon, prostate, and pancreatic adenocarcinomas. We report three key findings. First, two pathology-pretrained foundation models, Virchow-2 and UNI, substantially outperformed ImageNet-pretrained CNNs (EfficientNet-B3, ConvNeXt-2), with distinct baseline error profiles reflecting differences in pretraining-data composition. Second, a data curation strategy driven by confidence-based pseudo-labelling (threshold P\u2009&amp;gt;\u2009.9) with human-in-the-loop review expanded the dataset from 262 to 352 WSIs, yielding a 12.4% relative F1 improvement (0.740 to 0.832) and a 55.5% reduction in false positives per slide; an ablation attributed 70% of the F1 gain to data volume and 41% of the FP reduction to benign-class curation. Third, per-organ analysis revealed that the primary driver of this improvement was not data volume alone but the targeted annotation enrichment of underrepresented morphologies in adjacent-normal and benign tissue, including desmoplastic stroma, crypts, and small blood vessels, that had been a systematic source of false positive predictions across tissue types.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Availability<\/jats:title>\n                    <jats:p>Our implementation is available at https:\/\/github.com\/AhmadAlkhan\/PNI_SSL.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1093\/bioinformatics\/btag461","type":"journal-article","created":{"date-parts":[[2026,6,24]],"date-time":"2026-06-24T11:56:15Z","timestamp":1782302175000},"source":"Crossref","is-referenced-by-count":0,"title":["Semi-supervised learning for automated perineural invasion detection in multi-organ H&amp;E whole slide images"],"prefix":"10.1093","volume":"42","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-5130-9767","authenticated-orcid":false,"given":"A","family":"Alkhan","sequence":"first","affiliation":[{"name":"School of Medicine, Limerick Digital Cancer Research, and Health Research Institute, University of Limerick , Co Limerick, V94 T9PX,","place":["Ireland"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9535-6461","authenticated-orcid":false,"given":"M","family":"Lynch","sequence":"additional","affiliation":[{"name":"School of Medicine, Limerick Digital Cancer Research, and Health Research Institute, University of Limerick , Co Limerick, V94 T9PX,","place":["Ireland"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5137-1253","authenticated-orcid":false,"given":"E J","family":"Ryan","sequence":"additional","affiliation":[{"name":"Department of Biological Science, Limerick Digital Cancer Research Centre, Health Research Institute, University of Limerick, Co Limerick, V94 T9PX, Ireland"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0009-8404-6906","authenticated-orcid":false,"given":"M","family":"Lavelle","sequence":"additional","affiliation":[{"name":"Pathology Department, University Hospital Limerick , Co Limerick, V94 F858,","place":["Ireland"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1395-9734","authenticated-orcid":false,"given":"A C","family":"Culhane","sequence":"additional","affiliation":[{"name":"School of Medicine, Limerick Digital Cancer Research, and Health Research Institute, University of Limerick , Co Limerick, V94 T9PX,","place":["Ireland"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"286","published-online":{"date-parts":[[2026,8,21]]},"reference":[{"key":"2026082407033770000_btag461-B1","doi-asserted-by":"crossref","first-page":"850","DOI":"10.1038\/s41591-024-02857-3","article-title":"Towards a general-purpose foundation model for computational pathology","volume":"30","author":"Chen","year":"2024","journal-title":"Nat Med"},{"key":"2026082407033770000_btag461-B2","first-page":"1","article-title":"Perineural invasion of cancer: a complex crosstalk between cells and molecules in the perineural niche","volume":"9","author":"Chen","year":"2019","journal-title":"Am J Cancer Res"},{"key":"2026082407033770000_btag461-B3","doi-asserted-by":"crossref","first-page":"284","DOI":"10.1111\/his.14107","article-title":"Perineural invasion by prostate adenocarcinoma in needle biopsies predicts bone metastasis: ten year data from the TROG 03.04 RADAR Trial","volume":"77","author":"Delahunt","year":"2020","journal-title":"Histopathology"},{"key":"2026082407033770000_btag461-B4","doi-asserted-by":"crossref","DOI":"10.1016\/j.pathol.2020.08.006","article-title":"Benign mimics of prostate cancer","volume":"53","author":"Egevad","year":"2021","journal-title":"Pathology"},{"key":"2026082407033770000_btag461-B5","doi-asserted-by":"crossref","first-page":"1044026","DOI":"10.3389\/fonc.2022.1044026","article-title":"A semi-supervised learning approach with consistency regularization for tumor histopathological images analysis","volume":"12","author":"Jiang","year":"2022","journal-title":"Front Oncol"},{"key":"2026082407033770000_btag461-B6","doi-asserted-by":"crossref","first-page":"73","DOI":"10.1007\/s00428-022-03326-3","article-title":"Detection of perineural invasion in prostate needle biopsies with deep neural networks","volume":"481","author":"Kartasalo","year":"2022","journal-title":"Virchows Arch"},{"key":"2026082407033770000_btag461-B7","doi-asserted-by":"crossref","first-page":"382","DOI":"10.5858\/133.3.382","article-title":"Chronic pancreatitis and the differential diagnosis versus pancreatic cancer","volume":"133","author":"Kl\u00f6ppel","year":"2009","journal-title":"Arch Pathol Lab Med"},{"key":"2026082407033770000_btag461-B8","author":"Lee","year":"2013"},{"key":"2026082407033770000_btag461-B9","doi-asserted-by":"crossref","first-page":"310","DOI":"10.1038\/s41698-025-01104-3","article-title":"Computational pathology annotation enhances the resolution and interpretation of breast cancer spatial transcriptomics data","volume":"9","author":"Li","year":"2025","journal-title":"npj Precis Oncol"},{"key":"2026082407033770000_btag461-B10","doi-asserted-by":"crossref","first-page":"5131","DOI":"10.1200\/JCO.2009.22.4949","article-title":"Perineural invasion is an independent predictor of outcome in colorectal cancer","volume":"27","author":"Liebig","year":"2009","journal-title":"J Clin Oncol"},{"key":"2026082407033770000_btag461-B11","doi-asserted-by":"crossref","first-page":"3379","DOI":"10.1002\/cncr.24396","article-title":"Perineural invasion in cancer","volume":"115","author":"Liebig","year":"2009","journal-title":"Cancer"},{"key":"2026082407033770000_btag461-B12","doi-asserted-by":"crossref","first-page":"555","DOI":"10.1038\/s41551-020-00682-w","article-title":"Data-efficient and weakly supervised computational pathology on whole-slide images","volume":"5","author":"Lu","year":"2021","journal-title":"Nat Biomed Eng"},{"key":"2026082407033770000_btag461-B14","doi-asserted-by":"crossref","first-page":"62","DOI":"10.1109\/TSMC.1979.4310076","article-title":"A threshold selection method from gray-level histograms","volume":"9","author":"Otsu","year":"1979","journal-title":"IEEE Trans Syst Man Cybern"},{"key":"2026082407033770000_btag461-B15","author":"PAIP","year":"2021"},{"key":"2026082407033770000_btag461-B16","first-page":"1649","article-title":"Efficient perineural invasion detection of histopathological images using U-Net","volume":"11","author":"Park","year":"2022","journal-title":"Electronics (Basel)"},{"key":"2026082407033770000_btag461-B17","doi-asserted-by":"crossref","first-page":"105","DOI":"10.1016\/j.suronc.2017.01.007","article-title":"The influence of neural invasion on survival and tumor recurrence in pancreatic ductal adenocarcinoma \u2013 a systematic review and meta-analysis","volume":"26","author":"Schorn","year":"2017","journal-title":"Surg Oncol"},{"key":"2026082407033770000_btag461-B18","doi-asserted-by":"crossref","first-page":"5793","DOI":"10.3390\/cancers14235793","article-title":"Perineural invasion in pancreatic ductal adenocarcinoma: from molecules towards drugs of clinical relevance","volume":"14","author":"Selvaggi","year":"2022","journal-title":"Cancers (Basel)"},{"key":"2026082407033770000_btag461-B19","author":"SNUH","year":"2021"},{"key":"2026082407033770000_btag461-B20","author":"Sohn","year":"2020"},{"key":"2026082407033770000_btag461-B21","doi-asserted-by":"crossref","DOI":"10.3390\/ijms21155486","article-title":"Crosstalk between tumor and stromal cells in pancreatic ductal adenocarcinoma","volume":"21","author":"Sperb","year":"2020","journal-title":"Int J Mol Sci"},{"key":"2026082407033770000_btag461-B22","author":"Tan","year":"2019"},{"key":"2026082407033770000_btag461-B23","article-title":"Mean teachers are better role models: weight-averaged consistency targets improve semi-supervised deep learning results","volume":"30","author":"Tarvainen","year":"2017","journal-title":"Adv Neural Inf Process Syst"},{"key":"2026082407033770000_btag461-B24","doi-asserted-by":"crossref","first-page":"22","DOI":"10.1038\/modpathol.2017.136","article-title":"Benign mimics of prostatic adenocarcinoma","volume":"31","author":"Trpkov","year":"2018","journal-title":"Modern Pathol"},{"key":"2026082407033770000_btag461-B25","doi-asserted-by":"crossref","first-page":"373","DOI":"10.1007\/s10994-019-05855-6","article-title":"A survey on semi-supervised learning","volume":"109","author":"van Engelen","year":"2020","journal-title":"Mach Learn"},{"key":"2026082407033770000_btag461-B26","doi-asserted-by":"crossref","first-page":"11","DOI":"10.1016\/j.critrevonc.2017.02.005","article-title":"The role of perineural invasion in predicting survival in patients with primary operable colorectal cancer: a systematic review","volume":"112","author":"van Wyk","year":"2017","journal-title":"Crit Rev Oncol Hematol"},{"key":"2026082407033770000_btag461-B27","doi-asserted-by":"crossref","first-page":"2924","DOI":"10.1038\/s41591-024-03141-0","article-title":"A foundation model for clinical-grade computational pathology and rare cancers detection","volume":"30","author":"Vorontsov","year":"2024","journal-title":"Nat Med"},{"key":"2026082407033770000_btag461-B28","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1007\/s13402-023-00857-y","article-title":"Perineural invasion in colorectal cancer: mechanisms of action and clinical relevance","volume":"47","author":"Wang","year":"2024","journal-title":"Cell Oncol (Dordr)"},{"key":"2026082407033770000_btag461-B29","doi-asserted-by":"crossref","first-page":"4183","DOI":"10.18632\/oncotarget.6770","article-title":"The prognostic role of desmoplastic stroma in pancreatic ductal adenocarcinoma","volume":"7","author":"Wang","year":"2016","journal-title":"Oncotarget"},{"key":"2026082407033770000_btag461-B30","doi-asserted-by":"publisher","author":"Woo","year":"2023","DOI":"10.1109\/CVPR52729.2023.01548"},{"key":"2026082407033770000_btag461-B31","author":"Zimmermann","year":"2024"}],"container-title":["Bioinformatics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article-pdf\/42\/Supplement_2\/btag461\/70726329\/btag461.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article-pdf\/42\/Supplement_2\/btag461\/70726329\/btag461.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,24]],"date-time":"2026-08-24T11:03:44Z","timestamp":1787569424000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article\/doi\/10.1093\/bioinformatics\/btag461\/8767309"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,8,1]]},"references-count":30,"journal-issue":{"issue":"Supplement_2","published-print":{"date-parts":[[2026,8,1]]}},"URL":"https:\/\/doi.org\/10.1093\/bioinformatics\/btag461","relation":{},"ISSN":["1367-4803","1367-4811"],"issn-type":[{"value":"1367-4803","type":"print"},{"value":"1367-4811","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2026,8]]},"published":{"date-parts":[[2026,8,1]]},"article-number":"btag461"}}