{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,1]],"date-time":"2026-10-01T14:07:24Z","timestamp":1790863644735,"version":"4.1.0"},"reference-count":22,"publisher":"BMJ","issue":"10","license":[{"start":{"date-parts":[[2026,10,1]],"date-time":"2026-10-01T00:00:00Z","timestamp":1790812800000},"content-version":"unspecified","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by-nc\/4.0\/"}],"content-domain":{"domain":["bmj.com"],"crossmark-restriction":true},"short-container-title":["BMJ Open"],"accepted":{"date-parts":[[2026,9,24]]},"published-print":{"date-parts":[[2026,10]]},"abstract":"<jats:sec>\n                    <jats:title>Objectives<\/jats:title>\n                    <jats:p>To evaluate the diagnostic accuracy of a publicly available DenseNet-121 convolutional neural network (TorchXRayVision) for triaging chest radiographs of health assessment applicants at a tertiary hospital in Nepal.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Design<\/jats:title>\n                    <jats:p>Prospective, single-centre, shadow-mode diagnostic accuracy validation study. Reported in accordance with the Standards for Reporting of Diagnostic Accuracy Studies (STARD) 2015 checklist and STARD-Artificial Intelligence (AI)\/Developmental and Exploratory Clinical Investigations of DEcision support systems driven by Artificial Intelligence (DECIDE-AI) guidelines.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Setting<\/jats:title>\n                    <jats:p>Department of Radiology and Imaging, Patan Academy of Health Sciences\/Patan Hospital, Lalitpur, Nepal.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Participants<\/jats:title>\n                    <jats:p>826 consecutive health assessment applicants (foreign employment predeparture medical examination and student migration) undergoing chest radiography from 5 June 2026 to 20 June 2026 inclusive (16 days). Two cases were excluded due to Digital Imaging and Communications in Medicine technical failure.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Index test<\/jats:title>\n                    <jats:p>DenseNet-121 algorithm (TorchXRayVision library, densenet121-res224-all pretrained weights). A maximum aggregated pathology probability score was derived per radiograph and compared against a post hoc derived threshold of 0.6258 (selected as the highest threshold achieving the prespecified \u226595% sensitivity criterion).<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Reference standard<\/jats:title>\n                    <jats:p>Single-reader-per-case review by one of three radiologists\u2014two board-certified radiodiagnosticians (LS: 276 cases; DM: 275 cases) and one radiology resident (UB: 275 cases)\u2014each blinded to AI output, using a standardised data collection worksheet capturing binary classification (abnormal\/normal) and free-text findings.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>Of 826 radiographs, 41 (4.97%) were classified as abnormal by the reference standard. At the post hoc derived threshold of 0.6258, the DenseNet-121 algorithm achieved: sensitivity 95.12% (95% CI 83.9% to 98.7%), specificity 77.2% (95% CI 74.1% to 80.0%), area under the receiver operating characteristic curve 0.9583 (95% bootstrap CI 0.9225 to 0.9843), negative predictive value (NPV) 99.67% (95% Wilson CI 98.8% to 99.9%), positive predictive value 17.89% (95% Wilson CI 13.4% to 23.5%) and Cohen\u2019s \u03ba 0.237 (95% bootstrap CI 0.174 to 0.304). Brier score was 0.3621 (null Brier 0.0472) and expected calibration error was 0.564, confirming calibration failure due to score compression (range 0.52\u20130.72) despite preserved discrimination. The sensitivity estimate should be interpreted with caution given the relatively small number of reference-standard positives (n=41); the Wilson CI width of 14.8 percentage points (83.9\u201398.7%) reflects substantial uncertainty around this point estimate.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Cross-validated results<\/jats:title>\n                    <jats:p>10-fold cross-validation yielded bias-corrected sensitivity 95.12% (95% Wilson CI 83.9% to 98.7%; optimism 0.00 pp) and specificity 75.80% (95% Wilson CI 72.7% to 78.7%; optimism+1.40 pp), confirming by internal validation that primary metrics are not materially inflated by circular optimisation; independent external validation was not performed.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusions<\/jats:title>\n                    <jats:p>The DenseNet-121 algorithm demonstrated high point-estimate sensitivity and excellent discrimination for chest radiograph triage in a Nepali health-assessment population, supporting its potential as a radiographic abnormality rule-out triage tool (NPV 99.67%); this does not constitute microbiological exclusion of active pulmonary tuberculosis. Systematic score compression\u2014preserved discrimination despite calibration shift\u2014is a quantifiable marker of low- and middle-income country distributional shift. Prospective local calibration studies and independent external validation are warranted before operational deployment.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1136\/bmjopen-2026-124868","type":"journal-article","created":{"date-parts":[[2026,10,1]],"date-time":"2026-10-01T13:42:11Z","timestamp":1790862131000},"page":"e124868","update-policy":"https:\/\/doi.org\/10.1136\/crossmarkpolicy","source":"Crossref","is-referenced-by-count":0,"title":["Diagnostic accuracy of a DenseNet-121 deep learning algorithm for chest radiograph triage in health assessment applicants: a prospective shadow-mode validation study in Nepal"],"prefix":"10.1136","volume":"16","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-0814-3332","authenticated-orcid":false,"given":"Lochan","family":"Shrestha","sequence":"first","affiliation":[{"name":"Department of Radiology and Imaging, Patan Academy of Health Sciences, Lalitpur, Bagmati, Nepal"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-6177-7233","authenticated-orcid":false,"given":"Dinesh","family":"Maharjan","sequence":"additional","affiliation":[{"name":"Department of Radiology and Imaging, Patan Academy of Health Sciences, Lalitpur, Bagmati, Nepal"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0001-4898-5748","authenticated-orcid":false,"given":"Uttam","family":"Bista","sequence":"additional","affiliation":[{"name":"Department of Radiology and Imaging, Patan Academy of Health Sciences, Lalitpur, Bagmati, Nepal"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"239","published-online":{"date-parts":[[2026,10,1]]},"reference":[{"key":"2026100106201495000_16.10.e124868.1","unstructured":"World Health Organization . Health of migrants. WHO; 2022."},{"key":"2026100106201495000_16.10.e124868.2","unstructured":"World Health Organization . Global tuberculosis report 2023. Geneva WHO; 2023."},{"key":"2026100106201495000_16.10.e124868.3","unstructured":"Nepal National Tuberculosis Programme . Report of the 2023 Nepal TB programme review. Kathmandu: Nepal NTP, 2024."},{"key":"2026100106201495000_16.10.e124868.4","unstructured":"World Health Organization . Nepal successfully concludes joint review mission for tuberculosis. WHO; 2023."},{"key":"2026100106201495000_16.10.e124868.5","unstructured":"Wang X , Peng Y , Lu L , et al . ChestX-ray8: hospital-scale chest x-ray database and benchmarks. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition; 2017:2097\u2013106."},{"key":"2026100106201495000_16.10.e124868.6","doi-asserted-by":"crossref","first-page":"590","DOI":"10.1609\/aaai.v33i01.3301590","article-title":"CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and Expert Comparison","volume":"33","author":"Irvin","year":"2019","journal-title":"AAAI"},{"key":"2026100106201495000_16.10.e124868.7","doi-asserted-by":"crossref","unstructured":"Huang G , Liu Z , Van Der Maaten L , et al . Densely connected convolutional networks. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR); 4700. Honolulu, HI. doi:10.1109\/CVPR.2017.243 Available: https:\/\/ieeexplore.ieee.org\/xpl\/mostRecentIssue.jsp?punumber=8097368","DOI":"10.1109\/CVPR.2017.243"},{"key":"2026100106201495000_16.10.e124868.8","unstructured":"Rajpurkar P , Irvin J , Zhu K , et al . CheXNet: Radiologist-level pneumonia detection on chest X-rays with deep learning. arXiv Preprint arXiv:1711 2017."},{"key":"2026100106201495000_16.10.e124868.9","first-page":"231","article-title":"TorchXRayVision: A library of chest X-ray datasets and models","volume":"172","author":"Cohen","year":"2022","journal-title":"Proc Mach Learn Res Med Imaging"},{"key":"2026100106201495000_16.10.e124868.10","first-page":"874641","article-title":"Can we trust deep learning models diagnosis? The impact of domain shift in chest radiograph classification","volume":"9","author":"Pooch","year":"2022","journal-title":"Front Med (Lausanne)"},{"key":"2026100106201495000_16.10.e124868.11","doi-asserted-by":"crossref","DOI":"10.1371\/journal.pmed.1002683","article-title":"Variable generalization performance of a deep learning model to detect pneumonia in chest radiographs: A cross-sectional study","volume":"15","author":"Zech","year":"2018","journal-title":"PLoS Med"},{"key":"2026100106201495000_16.10.e124868.12","first-page":"689","article-title":"DECIDE-AI: new reporting guidance to bridge the development-to-implementation gap in clinical artificial intelligence","volume":"28","author":"Vasey","year":"2022","journal-title":"Nat Med"},{"key":"2026100106201495000_16.10.e124868.13","doi-asserted-by":"crossref","DOI":"10.1136\/bmj.h5527","article-title":"STARD 2015: an updated list of essential items for reporting diagnostic accuracy studies","volume":"351","author":"Bossuyt","year":"2015","journal-title":"BMJ"},{"key":"2026100106201495000_16.10.e124868.14","doi-asserted-by":"crossref","DOI":"10.1136\/bmjopen-2020-047709","article-title":"Developing a reporting guideline for artificial intelligence-centred diagnostic test accuracy studies: the STARD-AI protocol","volume":"11","author":"Sounderajah","year":"2021","journal-title":"BMJ Open"},{"key":"2026100106201495000_16.10.e124868.15","doi-asserted-by":"publisher","DOI":"10.1111\/j.1553-2712.1996.tb03538.x"},{"key":"2026100106201495000_16.10.e124868.16","doi-asserted-by":"publisher","DOI":"10.1016\/j.ejrad.2024.111783"},{"key":"2026100106201495000_16.10.e124868.17","unstructured":"Shrestha L , Maharjan D , Bista U . DenseNet-121 chest radiograph triage validation \u2014 nepal (PAHS). GitHub repository; 2026. Available: https:\/\/github.com\/lochanshrestha-dev\/densenet121-cxr-triage-nepal"},{"key":"2026100106201495000_16.10.e124868.18","doi-asserted-by":"publisher","DOI":"10.1175\/1520-0493(1950)078<0001:VOFEIT>2.0.CO;2"},{"key":"2026100106201495000_16.10.e124868.19","unstructured":"Guo C , Pleiss G , Sun Y , et al . On calibration of modern neural networks. Proc 34th Int Conf Mach Learn; 2017:1321\u201330."},{"key":"2026100106201495000_16.10.e124868.20","doi-asserted-by":"publisher","DOI":"10.2307\/2276774"},{"key":"2026100106201495000_16.10.e124868.21","doi-asserted-by":"publisher","DOI":"10.1016\/0895-4356(90)90158-L"},{"key":"2026100106201495000_16.10.e124868.22","doi-asserted-by":"publisher","DOI":"10.2307\/2531595"}],"container-title":["BMJ Open"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.1136\/bmjopen-2026-124868","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,10,1]],"date-time":"2026-10-01T13:42:17Z","timestamp":1790862137000},"score":1,"resource":{"primary":{"URL":"https:\/\/bmjopen.bmj.com\/lookup\/doi\/10.1136\/bmjopen-2026-124868"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,10]]},"references-count":22,"journal-issue":{"issue":"10","published-online":{"date-parts":[[2026,10,1]]},"published-print":{"date-parts":[[2026,10]]}},"alternative-id":["10.1136\/bmjopen-2026-124868"],"URL":"https:\/\/doi.org\/10.1136\/bmjopen-2026-124868","relation":{},"ISSN":["2044-6055","2044-6055"],"issn-type":[{"value":"2044-6055","type":"print"},{"value":"2044-6055","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,10]]}}}