{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,11,14]],"date-time":"2026-11-14T03:31:40Z","timestamp":1794627100462,"version":"4.9.1"},"reference-count":47,"publisher":"Springer Science and Business Media LLC","issue":"12","license":[{"start":{"date-parts":[[2022,9,15]],"date-time":"2022-09-15T00:00:00Z","timestamp":1663200000000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2022,9,15]],"date-time":"2022-09-15T00:00:00Z","timestamp":1663200000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["Nat. Biomed. Eng"],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>In tasks involving the interpretation of medical images, suitably trained machine-learning models often exceed the performance of medical experts. Yet such a high-level of performance typically requires that the models be trained with relevant datasets that have been painstakingly annotated by experts. Here we show that a self-supervised model trained on chest X-ray images that lack explicit annotations performs pathology-classification tasks with accuracies comparable to those of radiologists. On an external validation dataset of chest X-rays, the self-supervised model outperformed a fully supervised model in the detection of three pathologies (out of eight), and the performance generalized to pathologies that were not explicitly annotated for model training, to multiple image-interpretation tasks and to datasets from multiple institutions.<\/jats:p>","DOI":"10.1038\/s41551-022-00936-9","type":"journal-article","created":{"date-parts":[[2022,9,15]],"date-time":"2022-09-15T12:14:26Z","timestamp":1663244066000},"page":"1399-1406","update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":476,"title":["Expert-level detection of pathologies from unannotated chest X-ray images via self-supervised learning"],"prefix":"10.1038","volume":"6","author":[{"given":"Ekin","family":"Tiu","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ellie","family":"Talius","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Pujan","family":"Patel","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Curtis P.","family":"Langlotz","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Andrew Y.","family":"Ng","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8030-3727","authenticated-orcid":false,"given":"Pranav","family":"Rajpurkar","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"297","published-online":{"date-parts":[[2022,9,15]]},"reference":[{"key":"936_CR1","doi-asserted-by":"publisher","unstructured":"Rajpurkar, P., et al. 2017. CheXNet: radiologist-level pneumonia detection on chest X-Rays with deep learning. arXiv https:\/\/doi.org\/10.48550\/arXiv.1711.05225 (2017).","DOI":"10.48550\/arXiv.1711.05225"},{"key":"936_CR2","doi-asserted-by":"publisher","first-page":"60","DOI":"10.1016\/j.media.2017.07.005","volume":"42","author":"G Litjens","year":"2017","unstructured":"Litjens, G. et al. A survey on deep learning in medical image analysis. Med. Image Anal. 42, 60\u201388 (2017).","journal-title":"Med. Image Anal."},{"key":"936_CR3","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1186\/s12938-018-0544-y","volume":"17","author":"C Qin","year":"2018","unstructured":"Qin, C., Yao, D., Shi, Y. & Song, Z. Computer-aided detection in chest radiography based on artificial intelligence: a survey. Biomedical engineering online 17, 1\u201323 (2018).","journal-title":"Biomedical engineering online"},{"key":"936_CR4","doi-asserted-by":"publisher","unstructured":"Esteva, A. et al. Deep learning-enabled medical computer vision. NPJ Digit. Med. https:\/\/doi.org\/10.1038\/s41746-020-00376-2 (2021).","DOI":"10.1038\/s41746-020-00376-2"},{"key":"936_CR5","doi-asserted-by":"publisher","first-page":"221","DOI":"10.1146\/annurev-bioeng-071516-044442","volume":"19","author":"D Shen","year":"2017","unstructured":"Shen, D., Wu, G. & Suk, H.-I. Deep learning in medical image analysis. Annu. Rev. Biomed. Eng. 19, 221\u2013248 (2017).","journal-title":"Annu. Rev. Biomed. Eng."},{"key":"936_CR6","doi-asserted-by":"publisher","first-page":"103678","DOI":"10.1016\/j.engappai.2020.103678","volume":"92","author":"O Fink","year":"2020","unstructured":"Fink, O. et al. Potential, challenges and future directions for deep learning in prognostics and health management applications. Eng. Appl. Artif. Intell. 92, 103678 (2020).","journal-title":"Eng. Appl. Artif. Intell."},{"key":"936_CR7","doi-asserted-by":"publisher","unstructured":"Smit, A., et al. 2020. CheXbert: combining automatic labelers and expert annotations for accurate radiology report labeling using BERT. arXiv https:\/\/doi.org\/10.48550\/arXiv.2004.09167 (2020).","DOI":"10.48550\/arXiv.2004.09167"},{"key":"936_CR8","doi-asserted-by":"crossref","unstructured":"Irvin, J., et al. Chexpert: a large chest radiograph dataset with uncertainty labels and expert comparison. In Proc. AAAI Conference on Artificial Intelligence, 33:590\u2013597 (AAAI Press, 2019).","DOI":"10.1609\/aaai.v33i01.3301590"},{"key":"936_CR9","unstructured":"Erhan, D., A. Courville, Y. Bengio, and P. Vincent. Why does unsupervised pre-training help deep learning? In Proc. Thirteenth International Conference on Artificial Intelligence and Statistics (eds Teh, Y. W. & Titterington, T.) 9:201\u2013208 (PMLR, 2010)."},{"key":"936_CR10","doi-asserted-by":"crossref","unstructured":"Tan, C., Sun, F., Kong, T., Zhang, W., Yang, C., & Liu, C. A survey in deep transfer learning. In Artificial Neural Networks and Machine Learning \u2013 ICANN 2018 270\u2013279 (Springer Int. Publishing, Cham, 2018).","DOI":"10.1007\/978-3-030-01424-7_27"},{"key":"936_CR11","unstructured":"Chen, T., S. Kornblith, M. Norouzi, and G. Hinton. A simple framework for contrastive learning of visual representations. In International Conference on Machine Learning 1597\u20131607 (PMLR, 2020)."},{"key":"936_CR12","doi-asserted-by":"crossref","unstructured":"He, K., H. Fan, Y. Wu, S. Xie, and R. Girshick. Momentum contrast for unsupervised visual representation learning. In Proc. IEEE\/CVF Conference on Computer Vision and Pattern Recognition 9729\u20139738 (CVPR, 2020).","DOI":"10.1109\/CVPR42600.2020.00975"},{"key":"936_CR13","doi-asserted-by":"publisher","unstructured":"Vu, Y. N. T., et al. MedAug: contrastive learning leveraging patient metadata improves representations for chest X-ray interpretation. arXiv https:\/\/doi.org\/10.48550\/arXiv.2102.10663 (2021).","DOI":"10.48550\/arXiv.2102.10663"},{"key":"936_CR14","doi-asserted-by":"publisher","unstructured":"Zhang, Y., H. Jiang, Y. Miura, C. D. Manning, and C. P. Langlotz. Contrastive learning of medical visual representations from paired images and text. arXiv https:\/\/doi.org\/10.48550\/arXiv.2010.00747 (2020).","DOI":"10.48550\/arXiv.2010.00747"},{"key":"936_CR15","unstructured":"Radford, A., Kim, J. W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., ... & Sutskever, I. Learning transferable visual models from natural language supervision. In Proc. 38th International Conference on Machine Learning 39:8748\u20138763 (PMLR, 2021)."},{"key":"936_CR16","doi-asserted-by":"publisher","first-page":"2251","DOI":"10.1109\/TPAMI.2018.2857768","volume":"41","author":"Y Xian","year":"2019","unstructured":"Xian, Y., Lampert, C. H., Schiele, B. & Akata, Z. Zero-shot learning\u2014a comprehensive evaluation of the good, the bad and the ugly. IEEE Trans. Pattern Anal. Mach. Intell. 41, 2251\u20132265 (2019).","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell."},{"key":"936_CR17","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1038\/s41597-019-0322-0","volume":"6","author":"AEW Johnson","year":"2019","unstructured":"Johnson, A. E. W. et al. MIMIC-CXR, a de-identified publicly available database of chest radiographs with free-text reports. Sci. Data 6, 1\u20138 (2019).","journal-title":"Sci. Data"},{"key":"936_CR18","doi-asserted-by":"publisher","unstructured":"Sowrirajan, H., J. Yang, A. Y. Ng, and P. Rajpurkar. MoCo-CXR: pretraining improves representation and transferability of chest X-ray models. arXiv https:\/\/doi.org\/10.48550\/arXiv.2010.05352 (2021).","DOI":"10.48550\/arXiv.2010.05352"},{"key":"936_CR19","doi-asserted-by":"publisher","unstructured":"Pooch, E. H. P., P. L. Ballester, and R. C. Barros. Can we trust deep learning models diagnosis? The impact of domain shift in chest radiograph classification. arXiv https:\/\/doi.org\/10.48550\/arXiv.1909.01940 (2019).","DOI":"10.48550\/arXiv.1909.01940"},{"key":"936_CR20","doi-asserted-by":"publisher","first-page":"e1002686","DOI":"10.1371\/journal.pmed.1002686","volume":"15","author":"P Rajpurkar","year":"2018","unstructured":"Rajpurkar, P. et al. Deep learning for chest radiograph diagnosis: a retrospective comparison of the CheXNeXt algorithm to practicing radiologists. PLoS Med. 15, e1002686 (2018).","journal-title":"PLoS Med."},{"key":"936_CR21","doi-asserted-by":"crossref","unstructured":"Huang, S.-C., L. Shen, M. P. Lungren, and S. Yeung. GLoRIA: a multimodal global-local representation learning framework for label-efficient medical image recognition. In Proc. IEEE\/CVF International Conference on Computer Vision 3942\u20133951 (ICCV, 2021).","DOI":"10.1109\/ICCV48922.2021.00391"},{"key":"936_CR22","doi-asserted-by":"publisher","unstructured":"Hayat, N., H. Lashen, and F. E. Shamout. Multi-label generalized zero shot learning for the classification of disease in chest radiographs. arXiv https:\/\/doi.org\/10.48550\/arXiv.2107.06563 (2021).","DOI":"10.48550\/arXiv.2107.06563"},{"key":"936_CR23","doi-asserted-by":"publisher","unstructured":"Wang, X., Z. Xu, L. Tam, D. Yang, and D. Xu. Self-supervised image-text pre-training with mixed data in chest X-rays. arXiv https:\/\/doi.org\/10.48550\/arXiv.2103.16022 (2021).","DOI":"10.48550\/arXiv.2103.16022"},{"key":"936_CR24","first-page":"3285","volume":"28","author":"A Avdic","year":"2020","unstructured":"Avdic, A., Marovac, U. & Jankovic, D. Automated labeling of terms in medical reports in Serbian. Turk. J. Electr. Eng. Comput. Sci. 28, 3285\u20133303 (2020).","journal-title":"Turk. J. Electr. Eng. Comput. Sci."},{"key":"936_CR25","unstructured":"Haug, P. J., et al. 2014. Developing a section labeler for clinical documents. AMIA Annu. Symp. Proc. 636\u2013644 (2014)."},{"key":"936_CR26","doi-asserted-by":"publisher","first-page":"244","DOI":"10.1109\/JBHI.2017.2700722","volume":"22","author":"JX Qiu","year":"2018","unstructured":"Qiu, J. X., Yoon, H.-J., Fearn, P. A. & Tourassi, G. D. Deep learning for automated extraction of primary sites from cancer pathology reports. IEEE J. Biomed. Health Inform. 22, 244\u2013251 (2018).","journal-title":"IEEE J. Biomed. Health Inform."},{"key":"936_CR27","doi-asserted-by":"publisher","first-page":"107","DOI":"10.1145\/3446776","volume":"64","author":"C Zhang","year":"2021","unstructured":"Zhang, C., Bengio, S., Hardt, M., Recht, B. & Vinyals, O. Understanding deep learning (still) requires rethinking generalization. Commun. ACM 64, 107\u2013115 (2021).","journal-title":"Commun. ACM"},{"key":"936_CR28","unstructured":"Arjovsky, M.. Out of Distribution Generalization in Machine Learning (ed. Bottou, L.) PhD thesis, New York Univ. https:\/\/www.proquest.com\/dissertations-theses\/out-distribution-generalization-machine-learning\/docview\/2436913706\/se-2 (2020)."},{"key":"936_CR29","doi-asserted-by":"publisher","unstructured":"Radford, A., et al. Learning transferable visual models from natural language supervision. arXiv https:\/\/doi.org\/10.48550\/arXiv.2103.00020 (2021).","DOI":"10.48550\/arXiv.2103.00020"},{"key":"936_CR30","doi-asserted-by":"publisher","unstructured":"Liu, P., et al. Pre-train, prompt, and predict: a systematic survey of prompting methods in natural language processing. arXiv https:\/\/doi.org\/10.48550\/arXiv.2107.13586 (2021).","DOI":"10.48550\/arXiv.2107.13586"},{"key":"936_CR31","doi-asserted-by":"publisher","first-page":"5","DOI":"10.1148\/radiol.2351020759","volume":"235","author":"HS Patterson","year":"2005","unstructured":"Patterson, H. S. & Sponaugle, D. N. Is infiltrate a useful term in the interpretation of chest radiographs? Physician survey results. Radiology 235, 5\u20138 (2005).","journal-title":"Radiology"},{"key":"936_CR32","doi-asserted-by":"crossref","unstructured":"Liang, Z.-P., and P. C. Lauterbur. Principles of Magnetic Resonance Imaging (SPIE Optical Engineering Press Belllingham, 2000).","DOI":"10.1109\/9780470545652"},{"key":"936_CR33","doi-asserted-by":"publisher","first-page":"102","DOI":"10.1016\/j.zemedi.2018.11.002","volume":"29","author":"AS Lundervold","year":"2019","unstructured":"Lundervold, A. S. & Lundervold, A. An overview of deep learning in medical imaging focusing on MRI. Z. Med. Phys. 29, 102\u2013127 (2019).","journal-title":"Z. Med. Phys."},{"key":"936_CR34","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-020-77258-w","volume":"10","author":"Y Kim","year":"2020","unstructured":"Kim, Y. et al. Validation of deep learning natural language processing algorithm for keyword extraction from pathology reports in electronic health records. Sci. Rep. 10, 20265 (2020).","journal-title":"Sci. Rep."},{"key":"936_CR35","doi-asserted-by":"publisher","first-page":"775","DOI":"10.1038\/s41591-021-01343-4","volume":"27","author":"J van der Laak","year":"2021","unstructured":"van der Laak, J., Litjens, G. & Ciompi, F. Deep learning in histopathology: the path to the clinic. Nat. Med. 27, 775\u201384. (2021).","journal-title":"Nat. Med."},{"key":"936_CR36","doi-asserted-by":"publisher","unstructured":"Han, Y., C. Chen, A. H. Tewfik, Y. Ding, and Y. Peng. Pneumonia detection on chest X-ray using radiomic features and contrastive learning. arXiv https:\/\/doi.org\/10.48550\/arXiv.2101.04269 (2021).","DOI":"10.48550\/arXiv.2101.04269"},{"key":"936_CR37","doi-asserted-by":"publisher","first-page":"337","DOI":"10.1016\/j.jacr.2016.09.039","volume":"14","author":"SI Kamel","year":"2017","unstructured":"Kamel, S. I., Levin, D. C., Parker, L. & Rao, V. M. Utilization trends in noncardiac thoracic imaging, 2002\u20132014. J. Am. Coll. Radiology 14, 337\u2013342 (2017).","journal-title":"J. Am. Coll. Radiology"},{"key":"936_CR38","doi-asserted-by":"crossref","unstructured":"Cardoso, J., Van Nguyen, H., Heller, N., Abreu, P. H., Isgum, I., Silva, W., ... & Abbasi, S. in Interpretable and Annotation-Efficient Learning for Medical Image Computing 103\u2013111 (Springer Nature, 2020).","DOI":"10.1007\/978-3-030-61166-8_30"},{"key":"936_CR39","doi-asserted-by":"publisher","first-page":"2642","DOI":"10.1109\/TMI.2021.3054817","volume":"40","author":"A Paul","year":"2021","unstructured":"Paul, A. et al. Generalized zero-shot chest X-ray diagnosis through trait-guided multi-view semantic embedding with self-training. IEEE Trans. Med. Imaging 40, 2642\u20132655 (2021).","journal-title":"IEEE Trans. Med. Imaging"},{"key":"936_CR40","doi-asserted-by":"publisher","unstructured":"Raghu, M., C. Zhang, J. M. Kleinberg, and S. Bengio. Transfusion: understanding transfer learning with applications to medical imaging. arXiv https:\/\/doi.org\/10.48550\/arXiv.1902.07208 (2019).","DOI":"10.48550\/arXiv.1902.07208"},{"key":"936_CR41","doi-asserted-by":"publisher","first-page":"100005","DOI":"10.1016\/j.ibmed.2020.100005","volume":"3","author":"M Rezaei","year":"2020","unstructured":"Rezaei, M. & Shahidi, M. Zero-shot learning and its applications from autonomous vehicles to COVID-19 diagnosis: a review. Intell. Based Med. 3, 100005 (2020).","journal-title":"Intell. Based Med."},{"key":"936_CR42","doi-asserted-by":"publisher","unstructured":"Sennrich, R., B. Haddow, and A. Birch. Neural machine translation of rare words with subword units. arXiv https:\/\/doi.org\/10.48550\/arXiv.1508.07909 (2015).","DOI":"10.48550\/arXiv.1508.07909"},{"key":"936_CR43","doi-asserted-by":"publisher","first-page":"2251","DOI":"10.1109\/TPAMI.2018.2857768","volume":"41","author":"Y Xian","year":"2018","unstructured":"Xian, Y., Lampert, C. H., Schiele, B. & Akata, Z. Zero-shot learning\u2014a comprehensive evaluation of the good, the bad and the ugly. IEEE Trans. Pattern Anal. Mach. Intell. 41, 2251\u20132265 (2018).","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell."},{"key":"936_CR44","doi-asserted-by":"publisher","unstructured":"Yuan, Z., Y. Yan, M. Sonka, and T. Yang. Robust deep AUC maximization: a new surrogate loss and empirical studies on medical image classification. arXiv https:\/\/doi.org\/10.48550\/arXiv.2012.03173 (2020).","DOI":"10.48550\/arXiv.2012.03173"},{"key":"936_CR45","doi-asserted-by":"crossref","unstructured":"Pooch, E. H., Ballester, P., & Barros, R. C. Can we trust deep learning based diagnosis? The impact of domain shift in chest radiograph classification. In International Workshop on Thoracic Image Analysis pp. 74\u201383 (Springer, Cham, 2020).","DOI":"10.1007\/978-3-030-62469-9_7"},{"key":"936_CR46","doi-asserted-by":"publisher","first-page":"101797","DOI":"10.1016\/j.media.2020.101797","volume":"66","author":"A Bustos","year":"2020","unstructured":"Bustos, A., Pertusa, A., Salinas, J.-M. & de la Iglesia-Vay\u00e1, M. PadChest: a large chest X-ray image dataset with multi-label annotated reports. Med. Image Anal. 66, 101797 (2020).","journal-title":"Med. Image Anal."},{"key":"936_CR47","doi-asserted-by":"publisher","unstructured":"Gaillard, F. Tension pneumothorax. Case study. Radiopaedia.org https:\/\/doi.org\/10.53347\/rID-10558 (2010).","DOI":"10.53347\/rID-10558"}],"container-title":["Nature Biomedical Engineering"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.nature.com\/articles\/s41551-022-00936-9.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s41551-022-00936-9","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s41551-022-00936-9.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,9,4]],"date-time":"2025-09-04T10:57:31Z","timestamp":1756983451000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.nature.com\/articles\/s41551-022-00936-9"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2022,9,15]]},"references-count":47,"journal-issue":{"issue":"12","published-online":{"date-parts":[[2022,12]]}},"alternative-id":["936"],"URL":"https:\/\/doi.org\/10.1038\/s41551-022-00936-9","relation":{},"ISSN":["2157-846X"],"issn-type":[{"value":"2157-846X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2022,9,15]]},"assertion":[{"value":"30 November 2021","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"7 August 2022","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"15 September 2022","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"The authors declare no competing interests.","order":1,"name":"Ethics","group":{"name":"EthicsHeading","label":"Competing interests"}}]}}