{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,16]],"date-time":"2026-10-16T13:00:59Z","timestamp":1792155659923,"version":"4.4.0"},"reference-count":35,"publisher":"Public Library of Science (PLoS)","issue":"8","license":[{"start":{"date-parts":[[2026,8,20]],"date-time":"2026-08-20T00:00:00Z","timestamp":1787184000000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["www.plosone.org"],"crossmark-restriction":false},"short-container-title":["PLoS One"],"abstract":"<jats:p>Data scarcity is a persistent challenge in medical image analysis. Synthetic data generation using deep generative models has been proposed as a potential approach to address this limitation, yet its performance in small-data settings remains poorly characterized. This study compared three class-conditional generative approaches, a conditional variational autoencoder (cVAE), a conditional shallow-decoder VAE variant (cSD-VAE), and a conditional Wasserstein GAN with gradient penalty (cWGAN-GP), against classical geometric augmentation for emphysema subtype classification on 168 CT patches (three classes: normal tissue, centrilobular emphysema, and paraseptal emphysema). Each method was evaluated at three synthetic-to-real ratios (0.5x, 1.0x, 2.0x) using patient-level 70\/30 splits across 10 random seeds, with an ImageNet-pretrained ResNet18 as the downstream classifier. No individual augmentation strategy produced a statistically significant improvement in balanced accuracy over the unaugmented baseline (0.522\u2009\u00b1\u20090.067). The conditional WGAN-GP at 1.0x achieved the highest individual balanced accuracy (0.548\u2009\u00b1\u20090.068, Cohen\u2019s d\u2009=\u20090.47 versus baseline), but did not reach statistical significance (p\u2009=\u20090.084). A pre-specified ensemble combining all four augmentation methods at the 1.0x multiplier did not significantly improve balanced accuracy over baseline (0.541\u2009\u00b1\u20090.088 versus 0.522\u2009\u00b1\u20090.067; Cohen\u2019s d\u2009=\u20090.32, Wilcoxon p\u2009=\u20090.275). Neither pixel-space nor feature-space distributional fidelity was associated with downstream classification performance. Overall, no benefit was detected from class-conditional generative augmentation in this small-data, texture-driven setting. Future work should focus on improving generative modeling under small-data conditions, including task-aware objectives and pathology-constrained synthesis.<\/jats:p>","DOI":"10.1371\/journal.pone.0355850","type":"journal-article","created":{"date-parts":[[2026,8,20]],"date-time":"2026-08-20T17:35:02Z","timestamp":1787247302000},"page":"e0355850","update-policy":"https:\/\/doi.org\/10.1371\/journal.pone.corrections_policy","source":"Crossref","is-referenced-by-count":1,"title":["Synthetic data augmentation for CT-based emphysema subtype classification: A comparative evaluation of generative and classical approaches"],"prefix":"10.1371","volume":"21","author":[{"ORCID":"https:\/\/orcid.org\/0009-0007-1556-8343","authenticated-orcid":true,"given":"Nicholas","family":"Dietrich","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0000-6321-0193","authenticated-orcid":true,"given":"David","family":"McShannon","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"340","published-online":{"date-parts":[[2026,8,20]]},"reference":[{"issue":"10","key":"pone.0355850.ref001","doi-asserted-by":"crossref","first-page":"979","DOI":"10.3390\/ph14100979","article-title":"Mechanisms, pathophysiology and currently proposed treatments of chronic obstructive pulmonary disease","volume":"14","author":"O Rodrigues","year":"2021","journal-title":"Pharmaceuticals (Basel)"},{"key":"pone.0355850.ref002","doi-asserted-by":"crossref","DOI":"10.1016\/j.amjmed.2013.09.020","article-title":"Pulmonary emphysema subtypes on computed tomography: the MESA COPD study","volume":"127","author":"BM Smith","year":"2014","journal-title":"Am J Med"},{"issue":"7","key":"pone.0355850.ref003","doi-asserted-by":"crossref","first-page":"2113","DOI":"10.1148\/rg.2017170077","article-title":"Deep learning: a primer for radiologists","volume":"37","author":"G Chartrand","year":"2017","journal-title":"Radiographics"},{"issue":"1","key":"pone.0355850.ref004","doi-asserted-by":"crossref","first-page":"53","DOI":"10.1186\/s40537-021-00444-8","article-title":"Review of deep learning: concepts, CNN architectures, challenges, applications, future directions","volume":"8","author":"L Alzubaidi","year":"2021","journal-title":"J Big Data"},{"issue":"5","key":"pone.0355850.ref005","doi-asserted-by":"crossref","first-page":"820","DOI":"10.1109\/JPROC.2021.3054390","article-title":"A review of deep learning in medical imaging: imaging traits, technology trends, case studies with progress highlights, and future promises","volume":"109","author":"SK Zhou","year":"2021","journal-title":"Proc IEEE Inst Electr Electron Eng"},{"issue":"11","key":"pone.0355850.ref006","doi-asserted-by":"crossref","first-page":"375","DOI":"10.1016\/j.diii.2025.05.006","article-title":"Adversarial artificial intelligence in radiology: attacks, defenses, and future considerations","volume":"106","author":"N Dietrich","year":"2025","journal-title":"Diagn Interv Imaging"},{"key":"pone.0355850.ref007","first-page":"1","article-title":"Medical image data augmentation: techniques, comparisons and interpretations","author":"E Goceri","year":"2023","journal-title":"Artif Intell Rev"},{"issue":"3","key":"pone.0355850.ref008","article-title":"Tackling the small data problem in medical image classification with artificial intelligence: a systematic review","volume":"6","author":"S Piffer","year":"2024","journal-title":"Prog Biomed Eng (Bristol)"},{"key":"pone.0355850.ref009","doi-asserted-by":"crossref","first-page":"100258","DOI":"10.1016\/j.array.2022.100258","article-title":"Data augmentation: a comprehensive survey of modern approaches","volume":"16","author":"A Mumuni","year":"2022","journal-title":"Array"},{"key":"pone.0355850.ref010","doi-asserted-by":"crossref","first-page":"109834","DOI":"10.1016\/j.compbiomed.2025.109834","article-title":"Generative AI for synthetic data across multiple medical modalities: a systematic review of recent developments and challenges","volume":"189","author":"M Ibrahim","year":"2025","journal-title":"Comput Biol Med"},{"key":"pone.0355850.ref011","doi-asserted-by":"crossref","DOI":"10.1561\/9781680836233","article-title":"An introduction to variational autoencoders","author":"DP Kingma","year":"2019"},{"issue":"3","key":"pone.0355850.ref012","doi-asserted-by":"crossref","DOI":"10.1148\/radiol.232471","article-title":"Generating synthetic data for medical imaging","volume":"312","author":"LR Koetzier","year":"2024","journal-title":"Radiology"},{"issue":"1","key":"pone.0355850.ref013","article-title":"Synthetic data in radiological imaging: current state and future outlook","volume":"1","author":"E Sizikova","year":"2024","journal-title":"BJR Artif Intell"},{"key":"pone.0355850.ref014","doi-asserted-by":"crossref","first-page":"106391","DOI":"10.1016\/j.compbiomed.2022.106391","article-title":"Data augmentation for medical imaging: a systematic literature review","volume":"152","author":"F Garcea","year":"2023","journal-title":"Comput Biol Med"},{"key":"pone.0355850.ref015","doi-asserted-by":"crossref","first-page":"107","DOI":"10.22630\/MGV.2023.32.3.6","article-title":"Additional look into GAN-based augmentation for deep learning COVID-19 image classification","volume":"32","author":"O Fedoruk","year":"2023","journal-title":"MG&V"},{"issue":"2","key":"pone.0355850.ref016","doi-asserted-by":"crossref","first-page":"559","DOI":"10.1109\/TMI.2009.2038575","article-title":"Quantitative analysis of pulmonary emphysema using local binary patterns","volume":"29","author":"L S\u00f8rensen","year":"2010","journal-title":"IEEE Trans Med Imaging"},{"key":"pone.0355850.ref017","unstructured":"Computed Tomography Emphysema Database. https:\/\/lauge-soerensen.github.io\/emphysema-database\/. 2010. Accessed 2026 Jan 5."},{"key":"pone.0355850.ref018","doi-asserted-by":"crossref","unstructured":"He K, Zhang X, Ren S, Sun J. Deep residual learning for image recognition. In: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016. 770\u20138. https:\/\/doi.org\/10.1109\/cvpr.2016.90","DOI":"10.1109\/CVPR.2016.90"},{"key":"pone.0355850.ref019","doi-asserted-by":"crossref","unstructured":"Deng J, Dong W, Socher R, Li L-J, Kai Li, Li Fei-Fei. ImageNet: A large-scale hierarchical image database. In: 2009 IEEE Conference on Computer Vision and Pattern Recognition, 2009. 248\u201355. https:\/\/doi.org\/10.1109\/cvpr.2009.5206848","DOI":"10.1109\/CVPR.2009.5206848"},{"key":"pone.0355850.ref020","first-page":"5767","article-title":"Improved training of wasserstein GANs","volume":"30","author":"I Gulrajani","year":"2017","journal-title":"Advances in Neural Information Processing Systems"},{"key":"pone.0355850.ref021","first-page":"723","article-title":"A kernel two-sample test","author":"A Gretton","year":"2012","journal-title":"J Mach Learn Res"},{"key":"pone.0355850.ref022","doi-asserted-by":"crossref","first-page":"106539","DOI":"10.1016\/j.cmpb.2021.106539","article-title":"Domain- and task-specific transfer learning for medical segmentation tasks","volume":"214","author":"R Zoetmulder","year":"2022","journal-title":"Comput Methods Programs Biomed"},{"key":"pone.0355850.ref023","doi-asserted-by":"crossref","first-page":"104115","DOI":"10.1016\/j.compbiomed.2020.104115","article-title":"A scoping review of transfer learning research on medical image analysis using ImageNet","volume":"128","author":"MA Morid","year":"2021","journal-title":"Comput Biol Med"},{"issue":"3","key":"pone.0355850.ref024","doi-asserted-by":"crossref","first-page":"219","DOI":"10.3390\/bioengineering11030219","article-title":"A comprehensive review on synergy of multi-modal data and AI technologies in medical diagnosis","volume":"11","author":"X Xu","year":"2024","journal-title":"Bioengineering (Basel)"},{"key":"pone.0355850.ref025","doi-asserted-by":"crossref","first-page":"1808","DOI":"10.3390\/healthcare11121808","article-title":"Ensemble learning for disease prediction: a review","volume":"11","author":"P Mahajan","year":"2023","journal-title":"Healthcare"},{"key":"pone.0355850.ref026","doi-asserted-by":"crossref","unstructured":"Li H, Mukundan R. Robust Texture Features For Emphysema Classification In CT Images. In: 2020 28th European Signal Processing Conference (EUSIPCO), 2021. 1220\u20134. https:\/\/doi.org\/10.23919\/eusipco47968.2020.9287761","DOI":"10.23919\/Eusipco47968.2020.9287761"},{"key":"pone.0355850.ref027","doi-asserted-by":"crossref","unstructured":"Karabulut EM, Ibrikci T. Emphysema discrimination from raw HRCT images by convolutional neural networks. In: 2015 9th International Conference on Electrical and Electronics Engineering (ELECO), 2015. 705\u20138. https:\/\/doi.org\/10.1109\/eleco.2015.7394441","DOI":"10.1109\/ELECO.2015.7394441"},{"issue":"9","key":"pone.0355850.ref028","doi-asserted-by":"crossref","first-page":"1215","DOI":"10.1109\/LSP.2015.2392623","article-title":"Local diagonal extrema pattern: a new and efficient feature descriptor for CT image retrieval","volume":"22","author":"SR Dubey","year":"2015","journal-title":"IEEE Signal Process Lett"},{"issue":"5","key":"pone.0355850.ref029","doi-asserted-by":"crossref","first-page":"545","DOI":"10.1111\/1754-9485.13261","article-title":"A review of medical image data augmentation techniques for deep learning applications","volume":"65","author":"P Chlap","year":"2021","journal-title":"J Med Imaging Radiat Oncol"},{"key":"pone.0355850.ref030","doi-asserted-by":"crossref","first-page":"102846","DOI":"10.1016\/j.media.2023.102846","article-title":"Diffusion models in medical imaging: A comprehensive survey","volume":"88","author":"A Kazerouni","year":"2023","journal-title":"Med Image Anal"},{"issue":"8","key":"pone.0355850.ref031","doi-asserted-by":"crossref","first-page":"252","DOI":"10.3390\/jimaging11080252","article-title":"Synthetic scientific image generation with VAE, GAN, and diffusion model architectures","volume":"11","author":"Z Sordo","year":"2025","journal-title":"J Imaging"},{"key":"pone.0355850.ref032","first-page":"79","article-title":"Adapting pre-trained generative model to medical image for data augmentation","author":"Z Yuan","year":"2024","journal-title":"Medical Image Computing and Computer-Assisted Intervention (MICCAI)"},{"key":"pone.0355850.ref033","article-title":"Diffusion model-based medical image generation as a potential data augmentation strategy for AI applications","volume":"21","author":"Z Cao","year":"2025","journal-title":"Curr Med Imaging"},{"key":"pone.0355850.ref034","doi-asserted-by":"crossref","first-page":"211","DOI":"10.3390\/diagnostics16020211","article-title":"A systematic review of diffusion models for medical image-based diagnosis: methods, taxonomies, clinical integration, explainability, and future directions","volume":"16","author":"M Azad","year":"2026","journal-title":"Diagnostics"},{"issue":"12","key":"pone.0355850.ref035","doi-asserted-by":"crossref","first-page":"4217","DOI":"10.1109\/TPAMI.2020.2970919","article-title":"A Style-based generator architecture for generative adversarial networks","volume":"43","author":"T Karras","year":"2021","journal-title":"IEEE Trans Pattern Anal Mach Intell"}],"container-title":["PLOS One"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pone.0355850","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,20]],"date-time":"2026-08-20T17:35:09Z","timestamp":1787247309000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pone.0355850"}},"subtitle":[],"editor":[{"given":"Zeheng","family":"Wang","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"editor"}]}],"short-title":[],"issued":{"date-parts":[[2026,8,20]]},"references-count":35,"journal-issue":{"issue":"8","published-online":{"date-parts":[[2026,8,20]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pone.0355850","relation":{},"ISSN":["1932-6203"],"issn-type":[{"value":"1932-6203","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,8,20]]}}}