{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,27]],"date-time":"2026-09-27T23:56:17Z","timestamp":1790553377618,"version":"4.1.0"},"reference-count":44,"publisher":"Springer Science and Business Media LLC","issue":"1","license":[{"start":{"date-parts":[[2026,6,30]],"date-time":"2026-06-30T00:00:00Z","timestamp":1782777600000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2026,9,28]],"date-time":"2026-09-28T00:00:00Z","timestamp":1790553600000},"content-version":"vor","delay-in-days":90,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"funder":[{"name":"NIH","award":["U24CA279629"],"award-info":[{"award-number":["U24CA279629"]}]}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["Sci Rep"],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>Generative AI models are increasingly used to address data scarcity and class imbalance in biomedical image analysis, yet their practical value depends on downstream clinical task improvements rather than visual fidelity alone. To gain insight into their downstream utility, we conduct a systematic evaluation of class-conditional AI-generated images and their impact when integrated into a breast cancer classification pipeline. We develop a classifier-free guided denoising diffusion probabilistic model (DDPM) to generate benign and malignant full-image mammograms under varying inference configurations. Model training is based on digital breast tomosynthesis and 2D digital mammograms from the Emory Breast Imaging Dataset. DDPM guidance scales for inference are chosen based on image fidelity and class separability, as evaluated by Fr\u00e9chet Inception Distance and a pre-trained Oracle classifier, respectively. We also introduce a two-phase training strategy, where models get pretrained on real data augmented with AI-generated data followed by fine-tuning on real data only. Quantitative performance evaluation for guidance strength, AI-generated-to-real data proportion, and training strategy impact, is based on Balanced Accuracy, Sensitivity, and Specificity. Our findings indicate lack of consistent performance gains for AI-generated images used to mitigate data scarcity. Any gains are sensitive to guidance scale, liable to degrade when AI-generated data replaces real beyond a proportion. However, our two-phase training strategy yields consistent improvements, when compared with real-data-only baseline and fixed guidance scales. This suggests that AI-generated data cannot replace real clinical data, but can serve as a useful transfer learning signal when carefully integrated into the training pipelines.<\/jats:p>","DOI":"10.1038\/s41598-026-58260-0","type":"journal-article","created":{"date-parts":[[2026,6,30]],"date-time":"2026-06-30T06:05:56Z","timestamp":1782799556000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Evaluating the utility of AI-generated mammography images in breast cancer analysis"],"prefix":"10.1038","volume":"16","author":[{"given":"Akis","family":"Linardos","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Siddhesh","family":"Thakur","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Aimilia","family":"Gastounioti","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Dimitrios","family":"Makris","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Spyridon","family":"Bakas","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"297","published-online":{"date-parts":[[2026,6,30]]},"reference":[{"key":"58260_CR1","doi-asserted-by":"crossref","unstructured":"Voigt, P. & Von dem Bussche, A. The eu general data protection regulation (gdpr). A Practical Guide, 1st Ed., Cham: Springer International Publishing 10, 10\u20135555 (2017).","DOI":"10.1007\/978-3-319-57959-7_1"},{"key":"58260_CR2","doi-asserted-by":"publisher","first-page":"1486","DOI":"10.1056\/NEJMlim035027","volume":"348","author":"GJ Annas","year":"2003","unstructured":"Annas, G. J. Hipaa regulations: A new era of medical-record privacy?. New England J. Med. 348, 1486 (2003).","journal-title":"New England J. Med."},{"key":"58260_CR3","doi-asserted-by":"publisher","first-page":"2113","DOI":"10.1148\/rg.2017170077","volume":"37","author":"G Chartrand","year":"2017","unstructured":"Chartrand, G. et al. Deep learning: A primer for radiologists. Radiographics 37, 2113\u20132131 (2017).","journal-title":"Radiographics"},{"key":"58260_CR4","unstructured":"Marcus, G. Deep learning: A critical appraisal (2018) arXiv:1801.00631 arXiv preprint."},{"key":"58260_CR5","doi-asserted-by":"crossref","unstructured":"El Jiani, L. & El Filali, S. Overcome medical image data scarcity by data augmentation techniques: A review. In 2022 International conference on microelectronics (ICM) 21\u201324 (IEEE, 2022).","DOI":"10.1109\/ICM56065.2022.10005544"},{"key":"58260_CR6","doi-asserted-by":"publisher","first-page":"81","DOI":"10.3390\/jimaging9040081","volume":"9","author":"A Kebaili","year":"2023","unstructured":"Kebaili, A., Lapuyade-Lahorgue, J. & Ruan, S. Deep learning approaches for data augmentation in medical imaging: A review. J. Imaging 9, 81 (2023).","journal-title":"J. Imaging"},{"key":"58260_CR7","doi-asserted-by":"publisher","DOI":"10.1016\/j.compbiomed.2022.106391","volume":"152","author":"F Garcea","year":"2023","unstructured":"Garcea, F., Serra, A., Lamberti, F. & Morra, L. Data augmentation for medical imaging: A systematic literature review. Comput. Biol. Med. 152, 106391 (2023).","journal-title":"Comput. Biol. Med."},{"key":"58260_CR8","first-page":"3551","volume":"12","author":"A Linardos","year":"2022","unstructured":"Linardos, A., Kushibar, K., Walsh, S., Gkontra, P. & Lekadir, K. Federated learning for multi-center imaging diagnostics: A simulation study in cardiovascular disease. Scientif. Rep. 12, 3551 (2022).","journal-title":"Scientif. Rep."},{"key":"58260_CR9","doi-asserted-by":"crossref","unstructured":"DECAHNG, D. Semi-supervised medical image segmentation with multimodal data augmentation and self-supervised learning. In 2024 21st international computer conference on wavelet active media technology and information processing (ICCWAMTIP), 1\u20134 (IEEE, 2024).","DOI":"10.1109\/ICCWAMTIP64812.2024.10873760"},{"key":"58260_CR10","doi-asserted-by":"crossref","unstructured":"Xiang, Z., Wang, Z., Xie, J., Wang, W. & Xi, L. Research on medical image data enhancement based on variational autoencoders. In 2024 International conference on virtual reality and visualization (ICVRV), 78\u201381 (IEEE, 2024).","DOI":"10.1109\/ICVRV62410.2024.00023"},{"key":"58260_CR11","doi-asserted-by":"publisher","first-page":"1044496","DOI":"10.3389\/fonc.2022.1044496","volume":"12","author":"L Garrucho","year":"2023","unstructured":"Garrucho, L. et al. High-resolution synthesis of high-density breast mammograms: Application to improved fairness in deep learning based mass detection. Front. Oncol. 12, 1044496 (2023).","journal-title":"Front. Oncol."},{"key":"58260_CR12","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2022.102704","volume":"84","author":"R Osuala","year":"2023","unstructured":"Osuala, R. et al. Data synthesis and adversarial networks: A review and meta-analysis in cancer imaging. Med. Image Anal. 84, 102704 (2023).","journal-title":"Med. Image Anal."},{"key":"58260_CR13","doi-asserted-by":"publisher","first-page":"061403","DOI":"10.1117\/1.JMI.10.6.061403","volume":"10","author":"R Osuala","year":"2023","unstructured":"Osuala, R. et al. Medigan: A python library of pretrained generative models for medical image synthesis. J. Med. Imaging 10, 061403\u2013061403 (2023).","journal-title":"J. Med. Imaging"},{"key":"58260_CR14","doi-asserted-by":"crossref","unstructured":"Singla, C., Bhardwaj, R., Shelke, N. & Singh, G. Data augmentation: Synthetic image generation for medical images using vector quantized variational autoencoders. In 2025 3rd International conference on disruptive technologies (ICDT), 1502\u20131507 (IEEE, 2025).","DOI":"10.1109\/ICDT63985.2025.10986686"},{"key":"58260_CR15","doi-asserted-by":"publisher","first-page":"11","DOI":"10.1002\/acm2.13121","volume":"22","author":"T Wang","year":"2021","unstructured":"Wang, T. et al. A review on medical imaging synthesis using deep learning and its clinical applications. J. Appl. Clin. Med. Phys. 22, 11\u201336 (2021).","journal-title":"J. Appl. Clin. Med. Phys."},{"key":"58260_CR16","doi-asserted-by":"crossref","unstructured":"Kazerouni, A. Diffusion models for medical image analysis: A comprehensive survey (2022) arXiv:2211.07804 arXiv preprint.","DOI":"10.1016\/j.media.2023.102846"},{"key":"58260_CR17","unstructured":"Xiao, Z., Kreis, K. & Vahdat, A. Tackling the generative learning trilemma with denoising diffusion gans. arXiv preprint arXiv:2112.07804 (2021)."},{"key":"58260_CR18","first-page":"7303","volume":"13","author":"F Khader","year":"2023","unstructured":"Khader, F. et al. Denoising diffusion probabilistic models for 3d medical image generation. Scientif. Rep. 13, 7303 (2023).","journal-title":"Scientif. Rep."},{"key":"58260_CR19","unstructured":"Pati, S., Mazurek, S. & Bakas, S. Gandlf-synth: A framework to democratize generative ai for (bio) medical imaging. arXiv preprint arXiv:2410.00173 (2024)."},{"key":"58260_CR20","doi-asserted-by":"crossref","unstructured":"Osuala, R. Enhancing the utility of privacy-preserving cancer classification using synthetic data. In Deep breast workshop on AI and imaging for diagnostic and treatment challenges in breast care 54\u201364 (Springer, 2024).","DOI":"10.1007\/978-3-031-77789-9_6"},{"key":"58260_CR21","doi-asserted-by":"crossref","unstructured":"Gheshlaghi, S.\u00a0H., Kan, C. N.\u00a0E. & Ye, D.\u00a0H. Breast cancer histopathological image classification with adversarial image synthesis. In 2021 43rd annual international conference of the IEEE engineering in medicine & biology society (EMBC), 3387\u20133390 (IEEE, 2021).","DOI":"10.1109\/EMBC46164.2021.9630678"},{"key":"58260_CR22","doi-asserted-by":"publisher","DOI":"10.1016\/j.cmpb.2021.106019","volume":"202","author":"T Shen","year":"2021","unstructured":"Shen, T., Hao, K., Gou, C. & Wang, F.-Y. Mass image synthesis in mammogram with contextual information based on gans. Comput. Method. Progr. Biomed. 202, 106019 (2021).","journal-title":"Comput. Method. Progr. Biomed."},{"key":"58260_CR23","doi-asserted-by":"publisher","first-page":"176","DOI":"10.3390\/diagnostics9040176","volume":"9","author":"T Fujioka","year":"2019","unstructured":"Fujioka, T. et al. Breast ultrasound image synthesis using deep convolutional generative adversarial networks. Diagnostics 9, 176 (2019).","journal-title":"Diagnostics"},{"key":"58260_CR24","doi-asserted-by":"crossref","unstructured":"Korkinof, D. et al. Perceived realism of high-resolution generative adversarial network\u2013derived synthetic mammograms. Radiology: Artificial Intelligence 3, e190181 (2020).","DOI":"10.1148\/ryai.2020190181"},{"key":"58260_CR25","doi-asserted-by":"publisher","DOI":"10.1016\/j.bspc.2023.105456","volume":"87","author":"AJ Joseph","year":"2024","unstructured":"Joseph, A. J. et al. Prior-guided generative adversarial network for mammogram synthesis. Biomed. Signal Process. Control 87, 105456 (2024).","journal-title":"Biomed. Signal Process. Control"},{"key":"58260_CR26","first-page":"6166","volume":"12","author":"ON Oyelade","year":"2022","unstructured":"Oyelade, O. N. et al. A generative adversarial network for synthetization of regions of interest based on digital mammograms. Scientif. Rep. 12, 6166 (2022).","journal-title":"Scientif. Rep."},{"key":"58260_CR27","doi-asserted-by":"publisher","first-page":"2076","DOI":"10.3390\/s24072076","volume":"24","author":"R Montoya-del Angel","year":"2024","unstructured":"Montoya-del Angel, R., Sam-Millan, K., Vilanova, J. C. & Mart\u00ed, R. Mam-e: Mammographic synthetic image generation with diffusion models. Sensors 24, 2076 (2024).","journal-title":"Sensors"},{"key":"58260_CR28","doi-asserted-by":"crossref","unstructured":"Wang, C.-r., Zhang, F., Yu, Y. & Wang, Y. Br-gan: bilateral residual generating adversarial network for mammogram classification. In International conference on medical image computing and computer-assisted intervention, 657\u2013666 (Springer, 2020).","DOI":"10.1007\/978-3-030-59713-9_63"},{"key":"58260_CR29","doi-asserted-by":"crossref","unstructured":"Yang, Y., Fu, H., Aviles-Rivero, A.\u00a0I., Sch\u00f6nlieb, C.-B. & Zhu, L. Diffmic: Dual-guidance diffusion network for medical image classification. In International conference on medical image computing and computer-assisted intervention, 95\u2013105 (Springer, 2023).","DOI":"10.1007\/978-3-031-43987-2_10"},{"key":"58260_CR30","doi-asserted-by":"crossref","unstructured":"Wang, J. & Chen, L. Hybrid diffusion architecture for high-fidelity medical image synthesis. In 2025 IEEE International conference on pattern recognition, machine vision and artificial intelligence (PRMVAI), 1\u20135 (IEEE, 2025).","DOI":"10.1109\/PRMVAI65741.2025.11108327"},{"key":"58260_CR31","doi-asserted-by":"crossref","unstructured":"Wang, X. Conditional diffusion model for abdominal ct image synthesis. In 2025 IEEE 22nd international symposium on biomedical imaging (ISBI) 1\u20135 (IEEE, 2025).","DOI":"10.1109\/ISBI60581.2025.10980773"},{"key":"58260_CR32","unstructured":"Li, H. B. The brain tumor segmentation (brats) challenge 2023: Brain MR image synthesis for tumor segmentation (brasyn). ArXiv (2024) arXiv-2305."},{"key":"58260_CR33","doi-asserted-by":"crossref","unstructured":"Jeong, J.\u00a0J. et al. The emory breast imaging dataset (embed): A racially diverse, granular dataset of 3.4 million screening and diagnostic mammographic images. Radiology: Artificial Intelligence 5, e220047 (2023).","DOI":"10.1148\/ryai.220047"},{"key":"58260_CR34","doi-asserted-by":"crossref","unstructured":"Liu, Z. A convnet for the 2020s. In Proceedings of the IEEE\/CVF conference on computer vision and pattern recognition 11976\u201311986 (2022).","DOI":"10.1109\/CVPR52688.2022.01167"},{"key":"58260_CR35","doi-asserted-by":"crossref","unstructured":"Setiawan, D., Karnyoto, A.\u00a0S., Intan, I. & Pardamean, B. Convnext model for breast cancer image classification. In 2024 6th international conference on cybernetics and intelligent system (ICORIS), 1\u20135 (IEEE, 2024).","DOI":"10.1109\/ICORIS63540.2024.10903832"},{"key":"58260_CR36","first-page":"6840","volume":"33","author":"J Ho","year":"2020","unstructured":"Ho, J., Jain, A. & Abbeel, P. Denoising diffusion probabilistic models. Adv. Neural Inform. Process. Sys. 33, 6840\u20136851 (2020).","journal-title":"Adv. Neural Inform. Process. Sys."},{"key":"58260_CR37","doi-asserted-by":"crossref","unstructured":"Liu, J. Residual denoising diffusion models. In Proceedings of the IEEE\/CVF Conference on computer vision and pattern recognition 2773\u20132783 (2024).","DOI":"10.1109\/CVPR52733.2024.00268"},{"key":"58260_CR38","doi-asserted-by":"crossref","unstructured":"Zhang, J., Liu, D., Park, E., Zhang, S. & Xu, C. Residual learning in diffusion models. In Proceedings of the IEEE\/CVF conference on computer vision and pattern recognition, 7289\u20137299 (2024).","DOI":"10.1109\/CVPR52733.2024.00696"},{"key":"58260_CR39","doi-asserted-by":"crossref","unstructured":"Shabrina, N.\u00a0H., Gunawan, D., Ham, M.\u00a0F. & Harahap, A.\u00a0S. Papillary thyroid cancer histopathological image classification using pretrained convnext tiny and grad-cam interpretation. In 2023 IEEE 11th joint international information technology and artificial intelligence conference (ITAIC), vol.\u00a011, 1836\u20131842 (IEEE, 2023).","DOI":"10.1109\/ITAIC58329.2023.10409019"},{"key":"58260_CR40","doi-asserted-by":"crossref","unstructured":"Selvaraju, R. R. Grad-cam: Visual explanations from deep networks via gradient-based localization. In Proceedings of the IEEE international conference on computer vision 618\u2013626 (2017).","DOI":"10.1109\/ICCV.2017.74"},{"key":"58260_CR41","doi-asserted-by":"publisher","DOI":"10.1016\/j.bspc.2025.108915","volume":"113","author":"S Yan","year":"2026","unstructured":"Yan, S., Feng, M. & Cai, Y. An integrated deep learning model with enhanced EfficientNetB0 and MobileNetV1 for diabetic retinopathy grading. Biomed. Signal Process. Control 113, 108915 (2026).","journal-title":"Biomed. Signal Process. Control"},{"key":"58260_CR42","doi-asserted-by":"publisher","first-page":"614","DOI":"10.3390\/fractalfract9090614","volume":"9","author":"J Lin","year":"2025","unstructured":"Lin, J., Huang, L., Ding, L. & Yan, S. Deep brain tumor lesion classification network: A hybrid method optimizing ResNet50 and EfficientNetB0 for enhanced feature extraction. Fractal Fract. 9, 614 (2025).","journal-title":"Fractal Fract."},{"key":"58260_CR43","doi-asserted-by":"publisher","first-page":"1557","DOI":"10.3390\/math13101557","volume":"13","author":"M Feng","year":"2025","unstructured":"Feng, M., Cai, Y. & Yan, S. Enhanced ResNet50 for diabetic retinopathy classification: External attention and modified residual branch. Mathematics 13, 1557 (2025).","journal-title":"Mathematics"},{"key":"58260_CR44","doi-asserted-by":"publisher","first-page":"236","DOI":"10.1016\/j.acra.2011.09.014","volume":"19","author":"IC Moreira","year":"2012","unstructured":"Moreira, I. C. et al. INbreast: Toward a full-field digital mammographic database. Academic Radiol. 19, 236\u2013248 (2012).","journal-title":"Academic Radiol."}],"container-title":["Scientific Reports"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.nature.com\/articles\/s41598-026-58260-0","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s41598-026-58260-0.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s41598-026-58260-0.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,27]],"date-time":"2026-09-27T23:13:41Z","timestamp":1790550821000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.nature.com\/articles\/s41598-026-58260-0"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6,30]]},"references-count":44,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2026,12]]}},"alternative-id":["58260"],"URL":"https:\/\/doi.org\/10.1038\/s41598-026-58260-0","relation":{},"ISSN":["2045-2322"],"issn-type":[{"value":"2045-2322","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,6,30]]},"assertion":[{"value":"20 February 2026","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"12 June 2026","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"30 June 2026","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","label":"Competing interests","group":{"name":"EthicsHeading","label":"Declarations"}}],"article-number":"29981"}}