{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,30]],"date-time":"2026-09-30T14:00:55Z","timestamp":1790776855557,"version":"4.1.0"},"reference-count":31,"publisher":"Springer Science and Business Media LLC","issue":"1","license":[{"start":{"date-parts":[[2026,9,30]],"date-time":"2026-09-30T00:00:00Z","timestamp":1790726400000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2026,9,30]],"date-time":"2026-09-30T00:00:00Z","timestamp":1790726400000},"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":["BMC Med Imaging"],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>Assume you encounter an inverse problem that shall be solved for a large number of data, but no ground-truth data is available. To emulate this, in this study we assume it is unknown how to solve the imaging problem of Computed Tomography. We introduce an unsupervised deep learning framework that leverages the inherent similarities between iterative reconstruction, Deep Image Prior (DIP), and unrolled optimization schemes. Our specific contribution is a training framework for amortized reconstruction: After training on a dataset without any image-domain ground truth, reconstruction of an unseen scan reduces to a single network forward pass. We demonstrate the feasibility of reconstructing images from measurement data by pure network inference, without additional gradient steps for unseen samples. Our method is evaluated on the two-dimensional 2DeteCT dataset. Within a controlled, geometry-matched benchmark, our reconstructions are competitive with, or better than, filtered back-projection, maximum-likelihood reconstruction, and a supervised network of identical architecture. Compared to a per-image DIP baseline, our method reaches similar quality while replacing the costly per-instance optimization with a single forward pass, yielding a speed-up of about four orders of magnitude. This makes it a promising candidate for time-critical imaging applications. Future work will address multi-dataset adaptability, counter-measures against over-smoothing, advanced uncertainty quantification, and further medical-imaging inverse problems.<\/jats:p>","DOI":"10.1186\/s12880-026-02794-2","type":"journal-article","created":{"date-parts":[[2026,9,30]],"date-time":"2026-09-30T12:34:20Z","timestamp":1790771660000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Unsupervised deep learning for inverse problems in computed tomography"],"prefix":"10.1186","volume":"26","author":[{"given":"Laura","family":"Hellwege","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Johann Christopher","family":"Engster","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Moritz","family":"Schaar","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Thorsten M.","family":"Buzug","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Maik","family":"Stille","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"297","published-online":{"date-parts":[[2026,9,30]]},"reference":[{"key":"2794_CR1","doi-asserted-by":"publisher","unstructured":"Bertero M, Boccacci P, Mol DC. Introduction to inverse problems in imaging. In: Chap. Iterative methods. 2nd. CRC Press, 2021. p. 105\u201330. https:\/\/doi.org\/10.1201\/9780367806941","DOI":"10.1201\/9780367806941"},{"issue":"9","key":"2794_CR2","doi-asserted-by":"publisher","first-page":"4509","DOI":"10.1109\/TIP.2017.2713099","volume":"26","author":"KH Jin","year":"2017","unstructured":"Jin KH, McCann MT, Froustey E, Unser M. Deep convolutional neural network for inverse problems in imaging. IEEE Trans Image Process. 2017;26(9):4509\u201322. https:\/\/doi.org\/10.1109\/TIP.2017.2713099.","journal-title":"IEEE Trans Image Process"},{"key":"2794_CR3","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1808.03015","volume-title":"Radon inversion via deep learning. arXiv","author":"J He","year":"2018","unstructured":"He J, Ma J. Radon inversion via deep learning. arXiv. 2018. https:\/\/doi.org\/10.48550\/ARXIV.1808.03015."},{"issue":"10","key":"2794_CR4","doi-asserted-by":"publisher","first-page":"2469","DOI":"10.1109\/tmi.2019.2910760","volume":"38","author":"Y Li","year":"2019","unstructured":"Li Y, Li K, Zhang C, Montoya J, Chen G-H. Learning to reconstruct computed tomography images directly from sinogram data under a variety of data acquisition conditions. IEEE Trans Med Imag. 2019;38(10):2469\u201381. https:\/\/doi.org\/10.1109\/tmi.2019.2910760.","journal-title":"IEEE Trans Med Imag"},{"issue":"2","key":"2794_CR5","doi-asserted-by":"publisher","first-page":"415","DOI":"10.21037\/qims.2019.12.12","volume":"10","author":"Y Ge","year":"2020","unstructured":"Ge Y, Su T, Zhu J, Deng X, Zhang Q, Chen J, et al. ADAPTIVE-NET: deep computed tomography reconstruction network with analytical domain transformation knowledge. Quant Imag Med Surg. 2020;10(2):415\u201327. https:\/\/doi.org\/10.21037\/qims.2019.12.12.","journal-title":"Quant Imag Med Surg"},{"issue":"6","key":"2794_CR6","doi-asserted-by":"publisher","first-page":"1322","DOI":"10.1109\/TMI.2018.2799231","volume":"37","author":"J Adler","year":"2018","unstructured":"Adler J, \u00d6ktem O. Learned primal-dual reconstruction. IEEE Trans Med Imag. 2018;37(6):1322\u201332. https:\/\/doi.org\/10.1109\/TMI.2018.2799231.","journal-title":"IEEE Trans Med Imag"},{"key":"2794_CR7","doi-asserted-by":"crossref","unstructured":"Ulyanov D, Vedaldi A, Lempitsky VS. Deep image prior. Int J Comput Vis. 2020;7:1867\u201388. http:\/\/dx.doi.org\/10.1007\/s11263-020-01303-4.","DOI":"10.1007\/s11263-020-01303-4"},{"key":"2794_CR8","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.2012.06983","volume-title":"LEARN++: recurrent dual-domain reconstruction network for compressed sensing CT. arXiv","author":"Y Zhang","year":"2020","unstructured":"Zhang Y, Chen H, Xia W, Chen Y, Liu B, Liu Y, et al. LEARN++: recurrent dual-domain reconstruction network for compressed sensing CT. arXiv. 2020. https:\/\/doi.org\/10.48550\/ARXIV.2012.06983."},{"key":"2794_CR9","doi-asserted-by":"publisher","unstructured":"Li Y, Li W, Xiong J, Xia J, Xie Y. Comparison of supervised and unsupervised deep learning methods for medical image synthesis between computed tomography and magnetic resonance images. Biomed Res Int. 2020;2020(1). https:\/\/doi.org\/10.1155\/2020\/5193707.","DOI":"10.1155\/2020\/5193707"},{"key":"2794_CR10","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.2401.00135","volume-title":"Deep Radon prior: a fully unsupervised framework for sparse-view CT reconstruction.","author":"S Xu","year":"2024","unstructured":"Xu S, Zhang Y, Chen G, Xiang X, Cong P, Sun Y. Deep Radon prior: a fully unsupervised framework for sparse-view CT reconstruction. arXiv. 2024. https:\/\/doi.org\/10.48550\/ARXIV.2401.00135."},{"key":"2794_CR11","doi-asserted-by":"crossref","unstructured":"Mildenhall B, Srinivasan PP, Tancik M, Barron JT, Ramamoorthi R, Ng R. Nerf: representing scenes as neural radiance fields for view synthesis. CoRR abs\/ 2020;2003(08934). https:\/\/arxiv.org\/abs\/2003.08934..","DOI":"10.1007\/978-3-030-58452-8_24"},{"key":"2794_CR12","doi-asserted-by":"publisher","first-page":"1320","DOI":"10.1109\/TCI.2020.3019647","volume":"6","author":"AA Hendriksen","year":"2020","unstructured":"Hendriksen AA, Pelt DM, Batenburg KJ. Noise2Inverse: self-supervised deep convolutional denoising for tomography. IEEE Trans Comput Imag. 2020;6:1320\u201335. https:\/\/doi.org\/10.1109\/TCI.2020.3019647.","journal-title":"IEEE Trans Comput Imag"},{"issue":"6","key":"2794_CR13","doi-asserted-by":"publisher","first-page":"3172","DOI":"10.1002\/mrm.28378","volume":"84","author":"B Yaman","year":"2020","unstructured":"Yaman B, Hosseini SAH, Moeller S, Ellermann J, U\u011furbil K, Ak\u00e7akaya M. Self-supervised learning of physics-guided reconstruction neural networks without fully sampled reference data. Magn Reson Med. 2020;84(6):3172\u201391.","journal-title":"Magn Reson Med"},{"issue":"2","key":"2794_CR14","doi-asserted-by":"publisher","first-page":"394","DOI":"10.1109\/TMI.2018.2865356","volume":"38","author":"HK Aggarwal","year":"2019","unstructured":"Aggarwal HK, Mani MP, Jacob M. MoDL: model-based deep learning architecture for inverse problems. IEEE Trans Med Imag. 2019;38(2):394\u2013405. https:\/\/doi.org\/10.1109\/TMI.2018.2865356.","journal-title":"IEEE Trans Med Imag"},{"issue":"6","key":"2794_CR15","doi-asserted-by":"publisher","first-page":"3055","DOI":"10.1002\/mrm.26977","volume":"79","author":"K Hammernik","year":"2018","unstructured":"Hammernik K, Klatzer T, Kobler E, Recht MP, Sodickson DK, Pock T, et al. Learning a variational network for reconstruction of accelerated MRI data. Magn Reson Med. 2018;79(6):3055\u201371. https:\/\/doi.org\/10.1002\/mrm.26977.","journal-title":"Magn Reson Med"},{"key":"2794_CR16","doi-asserted-by":"crossref","unstructured":"Sriram A, Zbontar J, Murrell T, Defazio A, Zitnick CL, Yakubova N, et al. End-to-end variational networks for accelerated MRI reconstruction. CoRR abs\/ 2020;2004.06688. https:\/\/arxiv.org\/abs\/2004.06688.","DOI":"10.1007\/978-3-030-59713-9_7"},{"key":"2794_CR17","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.2103.14756","volume-title":"Equivariant imaging: learning beyond the range space","author":"D Chen","year":"2021","unstructured":"Chen D, Tachella J, Davies ME. Equivariant imaging: learning beyond the range space. 2021. https:\/\/doi.org\/10.48550\/ARXIV.2103.14756."},{"key":"2794_CR18","doi-asserted-by":"publisher","unstructured":"Paszke A, Gross S, Massa F, Lerer A, Bradbury J, Chanan G, et al. PyTorch: An imperative style, high-performance deep learning library. CoRR abs\/ 2019;1912.01703. https:\/\/doi.org\/10.48550\/arXiv.1912.01703..","DOI":"10.48550\/arXiv.1912.01703"},{"key":"2794_CR19","doi-asserted-by":"publisher","unstructured":"Schramm G, Thielemans K. PARALLELPROJ - an open-source framework for fast calculation of projections in tomography. Front Nucl Med. 2023;3, 1324562. https:\/\/doi.org\/10.3389\/fnume.2023.1324562.","DOI":"10.3389\/fnume.2023.1324562"},{"key":"2794_CR20","doi-asserted-by":"publisher","unstructured":"Zhou Z, Siddiquee MMR, Tajbakhsh N, Liang J. Unet++: Aa nested U-Net architecture for medical image segmentation. Deep Learn Med Image Anal Multimodal Learn Clin Decis Support. 2018;11045:3\u201311. https:\/\/doi.org\/10.1007\/978-3-030-00889-5_1.","DOI":"10.1007\/978-3-030-00889-5_1"},{"key":"2794_CR21","unstructured":"Xie S, Girshick RB, Doll\u00e1r P, Tu Z, He K. Aggregated residual transformations for deep neural networks. CoRR abs\/ (2016;1611(05431)."},{"key":"2794_CR22","unstructured":"Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation. CoRR abs\/ (2015;1505(04597). http:\/\/arxiv.org\/abs\/1505.04597."},{"key":"2794_CR23","unstructured":"Iakubovskii P. Segmentation models PyTorch. GitHub. 2019. https:\/\/github.com\/qubvel-org\/segmentation_models.pytorch."},{"issue":"1","key":"2794_CR24","doi-asserted-by":"publisher","first-page":"576","DOI":"10.1038\/s41597-023-02484-6","volume":"10","author":"MB Kiss","year":"2023","unstructured":"Kiss MB, Coban SB, Batenburg KJ, Leeuwen T, Lucka F. 2DeteCT - A large 2D expandable, trainable, experimental computed tomography dataset for machine learning. Sci Data. 2023;10(1):576. https:\/\/doi.org\/10.1038\/s41597-023-02484-6.","journal-title":"Sci Data"},{"key":"2794_CR25","doi-asserted-by":"publisher","unstructured":"Rit S, Vila Oliva M, Brousmiche S, Labarbe R, Sarrut D, Sharp GC. The Reconstruction toolkit (RTK), an open-source cone-beam CT reconstruction toolkit based on the insight toolkit (ITK). J Phys Conf Ser. 2014;489, 012079. https:\/\/doi.org\/10.1088\/1742-6596\/489\/1\/012079.","DOI":"10.1088\/1742-6596\/489\/1\/012079"},{"key":"2794_CR26","doi-asserted-by":"publisher","first-page":"261","DOI":"10.1038\/s41592-019-0686-2","volume":"17","author":"P Virtanen","year":"2020","unstructured":"Virtanen P, Gommers R, Oliphant TE, Haberland M, Reddy T, Cournapeau D, et al. SciPy 1.0 contributors: SciPy 1.0: fundamental algorithms for scientific computing in Python. Nat Methods. 2020;17:261\u201372. https:\/\/doi.org\/10.1038\/s41592-019-0686-2.","journal-title":"Nat Methods"},{"key":"2794_CR27","doi-asserted-by":"publisher","unstructured":"He K, Zhang X, Ren S, Sun J. Deep residual learning for image recognition. CoRR abs\/(2015;1512(03385). https:\/\/doi.org\/10.48550\/arXiv.1512.03385","DOI":"10.48550\/arXiv.1512.03385"},{"issue":"3","key":"2794_CR28","doi-asserted-by":"publisher","first-page":"456","DOI":"10.1007\/s10851-019-00923-x","volume":"62","author":"S Dittmer","year":"2019","unstructured":"Dittmer S, Kluth T, Maass P, Otero Baguer D. Regularization by architecture: a deep prior approach for inverse problems. J Math Imag Vis. 2019;62(3):456\u201370. https:\/\/doi.org\/10.1007\/s10851-019-00923-x.","journal-title":"J Math Imag Vis"},{"key":"2794_CR29","doi-asserted-by":"publisher","first-page":"156","DOI":"10.1007\/978-3-658-51100-5_33","volume-title":"Bildverarbeitung F\u00fcr die Medizin","author":"L Hellwege","year":"2026","unstructured":"Hellwege L, Engster JC, Schaar M, Buzug TM, Stille M. Reversible image augmentations for unsupervised deep learning in computed tomography. In: Handels H, Breininger K, Deserno T, Maier A, Maier-Hein K, Palm C, Tolxdorff T, editors. Bildverarbeitung F\u00fcr die Medizin. Wiesbaden: Springer; 2026. p. 156\u201361. https:\/\/doi.org\/10.1007\/978-3-658-51100-5_33."},{"key":"2794_CR30","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1603.08155","volume-title":"Perceptual losses for real-time style transfer and super-resolution. arXiv","author":"J Johnson","year":"2016","unstructured":"Johnson J, Alahi A, Fei-Fei L. Perceptual losses for real-time style transfer and super-resolution. arXiv. 2016. https:\/\/doi.org\/10.48550\/ARXIV.1603.08155."},{"issue":"12","key":"2794_CR31","doi-asserted-by":"publisher","first-page":"2536","DOI":"10.1109\/tmi.2017.2708987","volume":"36","author":"JM Wolterink","year":"2017","unstructured":"Wolterink JM, Leiner T, Viergever MA, Isgum I. Generative adversarial networks for noise reduction in low-dose ct. IEEE Trans Med Imag. 2017;36(12):2536\u201345. https:\/\/doi.org\/10.1109\/tmi.2017.2708987.","journal-title":"IEEE Trans Med Imag"}],"container-title":["BMC Medical Imaging"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/link.springer.com\/article\/10.1186\/s12880-026-02794-2","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1186\/s12880-026-02794-2.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1186\/s12880-026-02794-2.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,30]],"date-time":"2026-09-30T13:33:16Z","timestamp":1790775196000},"score":1,"resource":{"primary":{"URL":"https:\/\/link.springer.com\/10.1186\/s12880-026-02794-2"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,30]]},"references-count":31,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2026,12]]}},"alternative-id":["2794"],"URL":"https:\/\/doi.org\/10.1186\/s12880-026-02794-2","relation":{},"ISSN":["1471-2342"],"issn-type":[{"value":"1471-2342","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,30]]},"assertion":[{"value":"23 March 2026","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"8 September 2026","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"30 September 2026","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"Not applicable.","order":1,"name":"Ethics","label":"Ethical approval","group":{"name":"EthicsHeading","label":"Declaration"}},{"value":"During the preparation of this work, the authors used Claude Opus 4.8\/Anthropic in order to improve structure, language and readability of the manuscript text. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.","order":2,"name":"Ethics","label":"Generative AI statement","group":{"name":"EthicsHeading","label":"Declaration"}},{"value":"The authors declare no competing interests.","order":3,"name":"Ethics","label":"Competing interests","group":{"name":"EthicsHeading","label":"Declaration"}}],"article-number":"482"}}