{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,16]],"date-time":"2026-04-16T10:38:39Z","timestamp":1776335919344,"version":"3.51.2"},"reference-count":34,"publisher":"Proceedings of the National Academy of Sciences","issue":"48","license":[{"start":{"date-parts":[[2020,11,11]],"date-time":"2020-11-11T00:00:00Z","timestamp":1605052800000},"content-version":"vor","delay-in-days":184,"URL":"https:\/\/www.pnas.org\/site\/aboutpnas\/licenses.xhtml"}],"funder":[{"DOI":"10.13039\/501100000266","name":"RCUK | Engineering and Physical Sciences Research Council","doi-asserted-by":"publisher","award":["EP\/L003457\/1"],"award-info":[{"award-number":["EP\/L003457\/1"]}],"id":[{"id":"10.13039\/501100000266","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100000288","name":"Royal Society","doi-asserted-by":"publisher","award":["1"],"award-info":[{"award-number":["1"]}],"id":[{"id":"10.13039\/501100000288","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100000275","name":"Leverhulme Trust","doi-asserted-by":"publisher","award":["1"],"award-info":[{"award-number":["1"]}],"id":[{"id":"10.13039\/501100000275","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100007065","name":"Nvidia","doi-asserted-by":"publisher","award":["1"],"award-info":[{"award-number":["1"]}],"id":[{"id":"10.13039\/100007065","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100002790","name":"Canadian Network for Research and Innovation in Machining Technology, Natural Sciences and Engineering Research Council of Canada","doi-asserted-by":"publisher","award":["611675"],"award-info":[{"award-number":["611675"]}],"id":[{"id":"10.13039\/501100002790","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100000780","name":"European Commission","doi-asserted-by":"publisher","award":["655282"],"award-info":[{"award-number":["655282"]}],"id":[{"id":"10.13039\/501100000780","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100000780","name":"European Commission","doi-asserted-by":"publisher","award":["SFRH\/BPD\/118714\/2016"],"award-info":[{"award-number":["SFRH\/BPD\/118714\/2016"]}],"id":[{"id":"10.13039\/501100000780","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["www.pnas.org"],"crossmark-restriction":true},"short-container-title":["Proc. Natl. Acad. Sci. U.S.A."],"published-print":{"date-parts":[[2020,12]]},"abstract":"<jats:p>Deep learning, due to its unprecedented success in tasks such as image classification, has emerged as a new tool in image reconstruction with potential to change the field. In this paper, we demonstrate a crucial phenomenon: Deep learning typically yields unstable methods for image reconstruction. The instabilities usually occur in several forms: 1) Certain tiny, almost undetectable perturbations, both in the image and sampling domain, may result in severe artefacts in the reconstruction; 2) a small structural change, for example, a tumor, may not be captured in the reconstructed image; and 3) (a counterintuitive type of instability) more samples may yield poorer performance. Our stability test with algorithms and easy-to-use software detects the instability phenomena. The test is aimed at researchers, to test their networks for instabilities, and for government agencies, such as the Food and Drug Administration (FDA), to secure safe use of deep learning methods.<\/jats:p>","DOI":"10.1073\/pnas.1907377117","type":"journal-article","created":{"date-parts":[[2020,5,12]],"date-time":"2020-05-12T00:25:15Z","timestamp":1589243115000},"page":"30088-30095","update-policy":"https:\/\/doi.org\/10.1073\/pnas.cm10313","source":"Crossref","is-referenced-by-count":571,"title":["On instabilities of deep learning in image reconstruction and the potential costs of AI"],"prefix":"10.1073","volume":"117","author":[{"given":"Vegard","family":"Antun","sequence":"first","affiliation":[{"name":"Department of Mathematics, University of Oslo, 0316 Oslo, Norway;"}]},{"given":"Francesco","family":"Renna","sequence":"additional","affiliation":[{"name":"Instituto de Telecomunica\u00e7\u00f5es, Faculdade de Ci\u00eancias, Universidade do Porto, Porto 4169-007, Portugal;"}]},{"given":"Clarice","family":"Poon","sequence":"additional","affiliation":[{"name":"Department of Mathematical Sciences, University of Bath, Bath BA2 7AY, United Kingdom;"}]},{"given":"Ben","family":"Adcock","sequence":"additional","affiliation":[{"name":"Department of Mathematics, Simon Fraser University, Burnaby, BC V5A 1S6, Canada;"}]},{"given":"Anders C.","family":"Hansen","sequence":"additional","affiliation":[{"name":"Department of Mathematics, University of Oslo, 0316 Oslo, Norway;"},{"name":"Department of Applied Mathematics and Theoretical Physics, University of Cambridge, Cambridge CB3 0WA, United Kingdom"}]}],"member":"341","published-online":{"date-parts":[[2020,5,11]]},"reference":[{"key":"e_1_3_3_1_2","doi-asserted-by":"publisher","DOI":"10.1038\/272686a0"},{"key":"e_1_3_3_2_2","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.1119511109"},{"key":"e_1_3_3_3_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-94-009-1740-8"},{"key":"e_1_3_3_4_2","first-page":"119","volume-title":"Computational Inverse Problems in Electrocardiology","author":"Hansen P. C.","year":"2000","unstructured":"P. C. Hansen, \u201cThe l-curve and its use in the numerical treatment of inverse problems\u201d in Computational Inverse Problems in Electrocardiology, P. R. Johnston, Ed. (Advances in Computational Bioengineering, WIT Press, 2000), pp. 119\u2013142."},{"key":"e_1_3_3_5_2","doi-asserted-by":"publisher","DOI":"10.1038\/nature14539"},{"key":"e_1_3_3_6_2","doi-asserted-by":"publisher","DOI":"10.1038\/nature25988"},{"key":"e_1_3_3_7_2","doi-asserted-by":"publisher","DOI":"10.1038\/nmeth.4678"},{"key":"e_1_3_3_8_2","doi-asserted-by":"publisher","DOI":"10.1109\/MSP.2017.2739299"},{"key":"e_1_3_3_9_2","doi-asserted-by":"publisher","DOI":"10.1109\/TIP.2017.2713099"},{"key":"e_1_3_3_10_2","first-page":"9596","volume-title":"Advances in Neural Information Processing Systems","author":"Mardani M.","year":"2018","unstructured":"M. Mardani , \u201cNeural proximal gradient descent for compressive imaging\u201d in Advances in Neural Information Processing Systems, S. Bengio , Eds. (Curran Associates, Inc., 2018), pp. 9596\u20139606."},{"key":"e_1_3_3_11_2","doi-asserted-by":"crossref","first-page":"647","DOI":"10.1007\/978-3-319-59050-9_51","volume-title":"International Conference on Information Processing in Medical Imaging","author":"Schlemper J.","year":"2017","unstructured":"J. Schlemper, J. Caballero, J. V. Hajnal, A. Price, D. Rueckert, \u201cA deep cascade of convolutional neural networks for MR image reconstruction\u201d in International Conference on Information Processing in Medical Imaging, M. Niethammer , Eds. (Springer, 2017), pp. 647\u2013658."},{"key":"e_1_3_3_12_2","doi-asserted-by":"publisher","DOI":"10.1109\/TMI.2017.2785879"},{"key":"e_1_3_3_13_2","unstructured":"C. McCollough Data from \u201cLow Dose CT Grand Challenge dataset.\u201d Mayo Clinic. https:\/\/www.aapm.org\/GrandChallenge\/LowDoseCT\/. Accessed 19 April 2018."},{"key":"e_1_3_3_14_2","doi-asserted-by":"publisher","DOI":"10.1002\/mrm.26977"},{"key":"e_1_3_3_15_2","doi-asserted-by":"publisher","DOI":"10.1109\/MSP.2017.2760358"},{"key":"e_1_3_3_16_2","volume-title":"Deep, Deep Trouble - Deep Learning\u2019s Impact on Image Processing, Mathematics, and Humanity","author":"Elad M.","year":"2017","unstructured":"M. Elad, Deep, Deep Trouble - Deep Learning\u2019s Impact on Image Processing, Mathematics, and Humanity (SIAM News, 2017)."},{"key":"e_1_3_3_17_2","first-page":"580","volume-title":"IEEE Conference on Computer Vision and Pattern Recognition","author":"Girshick R.","year":"2014","unstructured":"R. Girshick, J. Donahue, T. Darrell, J. Malik, \u201cRich feature hierarchies for accurate object detection and semantic segmentation\u201d in IEEE Conference on Computer Vision and Pattern Recognition (Institute of Electrical and Electronic Engineers, 2014), pp. 580\u2013587."},{"key":"e_1_3_3_18_2","first-page":"1097","volume-title":"Advances in Neural Information Processing Systems","author":"Krizhevsky A.","year":"2012","unstructured":"A. Krizhevsky, I. Sutskever, G. E. Hinton, \u201cImagenet classification with deep convolutional neural networks\u201d in Advances in Neural Information Processing Systems, F. Pereira, C. J. C. Burges, L. Bottou, K. Q. Weinberger, Eds. (Curran Associates, Inc., 2012), pp. 1097\u20131105."},{"key":"e_1_3_3_19_2","first-page":"487","volume-title":"Advances in Neural Information Processing Systems","author":"Zhou B.","year":"2014","unstructured":"B. Zhou, A. Lapedriza, J. Xiao, A. Torralba, A. Oliva, \u201cLearning deep features for scene recognition using places database\u201d in Advances in Neural Information Processing Systems, Z. Ghahramani, M. Welling, C. Cortes, N. D. Lawrence, K. Q. Weinberger, Eds. (Curran Associates, Inc., 2014), pp. 487\u2013495."},{"key":"e_1_3_3_20_2","first-page":"1026","volume-title":"IEEE International Conference on Computer Vision","author":"He K.","year":"2015","unstructured":"K. He, X. Zhang, S. Ren, J. Sun, \u201cDelving deep into rectifiers: Surpassing human-level performance on imagenet classification\u201d in IEEE International Conference on Computer Vision (Institute of Electrical and Electronic Engineers, 2015), pp. 1026\u20131034."},{"key":"e_1_3_3_21_2","first-page":"4441","volume-title":"IEEE Conference on Computer Vision and Pattern Recognition","author":"Kanbak C.","year":"2018","unstructured":"C. Kanbak, S.-M. Moosavi-Dezfooli, P. Frossard, \u201cGeometric robustness of deep networks: Analysis and improvement\u201d in IEEE Conference on Computer Vision and Pattern Recognition (Institute of Electrical and Electronic Engineers, 2018), pp. 4441\u20134449."},{"key":"e_1_3_3_22_2","volume-title":"IEEE Conference on Computer Vision and Pattern Recognition","author":"Moosavi-Dezfooli S.-M.","year":"2016","unstructured":"S.-M. Moosavi-Dezfooli, A. Fawzi, P. Frossard, \u201cDeepfool: A simple and accurate method to fool deep neural networks\u201d in IEEE Conference on Computer Vision and Pattern Recognition (Institute of Electrical and Electronic Engineers, 2016)."},{"key":"e_1_3_3_23_2","first-page":"86","volume-title":"IEEE Conference on Computer Vision and Pattern Recognition","author":"Moosavi-Dezfooli S.","year":"2017","unstructured":"S. Moosavi-Dezfooli, A. Fawzi, O. Fawzi, P. Frossard, \u201cUniversal adversarial perturbations\u201d in IEEE Conference on Computer Vision and Pattern Recognition (Institute of Electrical and Electronic Engineers, 2017), pp. 86\u201394."},{"key":"e_1_3_3_24_2","volume-title":"International Conference on Learning Representations","author":"Szegedy C.","year":"2014","unstructured":"C. Szegedy, W. Zaremba, I. Sutskever, J. Bruna, D. Erhan, I. J. Goodfellow, R. Fergus, \u201cIntriguing properties of neural networks\u201d in International Conference on Learning Representations (2014). https:\/\/openreview.net\/forum?id=kklr_MTHMRQjG. Accessed 28 April 2020."},{"key":"e_1_3_3_25_2","doi-asserted-by":"publisher","DOI":"10.1109\/MSP.2017.2740965"},{"key":"e_1_3_3_26_2","doi-asserted-by":"publisher","DOI":"10.1109\/TIT.2005.862083"},{"key":"e_1_3_3_27_2","doi-asserted-by":"publisher","DOI":"10.1109\/TIT.2006.871582"},{"key":"e_1_3_3_28_2","doi-asserted-by":"publisher","DOI":"10.1016\/0167-2789(92)90242-F"},{"key":"e_1_3_3_29_2","doi-asserted-by":"publisher","DOI":"10.1017\/fms.2016.32"},{"key":"e_1_3_3_30_2","doi-asserted-by":"publisher","DOI":"10.1109\/TMI.2018.2832656"},{"key":"e_1_3_3_31_2","doi-asserted-by":"publisher","DOI":"10.1088\/1361-6420\/aa9581"},{"key":"e_1_3_3_32_2","doi-asserted-by":"publisher","DOI":"10.1109\/TMI.2018.2887072"},{"key":"e_1_3_3_33_2","first-page":"763","volume-title":"Advances in Neural Information Processing Systems","author":"Bigdeli S. A.","year":"2017","unstructured":"S. A. Bigdeli, M. Zwicker, P. Favaro, M. Jin, \u201cDeep mean-shift priors for image restoration\u201d in Advances in Neural Information Processing Systems, I. Guyon , Eds. (Curran Associates, Inc., 2017), pp. 763\u2013772."},{"key":"e_1_3_3_34_2","first-page":"5888","volume-title":"Proceedings of the IEEE International Conference on Computer Vision","author":"Chang J. H. R.","year":"2017","unstructured":"J. H. R. Chang, C.-L. Li, B. Poczos, B. V. K. Vijaya Kumar, A. C. Sankaranarayanan, \u201cOne network to solve them all\u2013solving linear inverse problems using deep projection models\u201d in Proceedings of the IEEE International Conference on Computer Vision (Institute of Electrical and Electronic Engineers, 2017), pp. 5888\u20135897."}],"container-title":["Proceedings of the National Academy of Sciences"],"original-title":[],"language":"en","link":[{"URL":"http:\/\/www.pnas.org\/syndication\/doi\/10.1073\/pnas.1907377117","content-type":"unspecified","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/pnas.org\/doi\/pdf\/10.1073\/pnas.1907377117","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2022,4,13]],"date-time":"2022-04-13T09:22:12Z","timestamp":1649841732000},"score":1,"resource":{"primary":{"URL":"https:\/\/pnas.org\/doi\/full\/10.1073\/pnas.1907377117"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2020,5,11]]},"references-count":34,"journal-issue":{"issue":"48","published-print":{"date-parts":[[2020,12]]}},"alternative-id":["10.1073\/pnas.1907377117"],"URL":"https:\/\/doi.org\/10.1073\/pnas.1907377117","relation":{},"ISSN":["0027-8424","1091-6490"],"issn-type":[{"value":"0027-8424","type":"print"},{"value":"1091-6490","type":"electronic"}],"subject":[],"published":{"date-parts":[[2020,5,11]]},"assertion":[{"value":"2020-05-11","order":2,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}