Generating Brain MRI with StyleGAN2-ADA: The Effect of the Training Set Size on the Quality of Synthetic Images.
Saved in:
| Title: | Generating Brain MRI with StyleGAN2-ADA: The Effect of the Training Set Size on the Quality of Synthetic Images. |
|---|---|
| Authors: | Lai, Matteo1 (AUTHOR) matteo.lai3@unibo.it, Mascalchi, Mario2 (AUTHOR) mario.mascalchi@unifi.it, Tessa, Carlo3 (AUTHOR) carlo.tessa@uslnordovest.toscana.it, Diciotti, Stefano1,4 (AUTHOR) stefano.diciotti@unibo.it |
| Source: | Journal of Imaging Informatics in Medicine. Jun2026, Vol. 39 Issue 3, p2319-2329. 11p. |
| Subjects: | Generative adversarial networks, Diagnostic imaging, Prediction models, Research funding, Brain, Privacy, Magnetic resonance imaging, Descriptive statistics, Deep learning, Digital image processing, Data quality, Sensitivity & specificity (Statistics), Medical ethics |
| Abstract: | The potential of deep learning for medical imaging is often constrained by limited data availability. Generative models can unlock this potential by generating synthetic data that reproduces the statistical properties of real data while being more accessible for sharing. In this study, we investigated the influence of training set size on the performance of a state-of-the-art generative adversarial network, the StyleGAN2-ADA, trained on a cohort of 3,227 subjects from the OpenBHB dataset to generate 2D slices of brain MR images from healthy subjects. The quality of the synthetic images was assessed through qualitative evaluations and state-of-the-art quantitative metrics, which are provided in a publicly accessible repository. Our results demonstrate that StyleGAN2-ADA generates realistic and high-quality images, deceiving even expert radiologists while preserving privacy, as it did not memorize training images. Notably, increasing the training set size led to slight improvements in fidelity metrics. However, training set size had no noticeable impact on diversity metrics, highlighting the persistent limitation of mode collapse. Furthermore, we observed that diversity metrics, such as coverage and β-recall, are highly sensitive to the number of synthetic images used in their computation, leading to inflated values when synthetic data significantly outnumber real ones. These findings underscore the need to carefully interpret diversity metrics and the importance of employing complementary evaluation strategies for robust assessment. Overall, while StyleGAN2-ADA shows promise as a tool for generating privacy-preserving synthetic medical images, overcoming diversity limitations will require exploring alternative generative architectures or incorporating additional regularization techniques. [ABSTRACT FROM AUTHOR] |
| Copyright of Journal of Imaging Informatics in Medicine is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
| FullText | Text: Availability: 0 |
|---|---|
| Header | DbId: egs DbLabel: Engineering Source An: 194225505 AccessLevel: 6 PubType: Academic Journal PubTypeId: academicJournal PreciseRelevancyScore: 0 |
| IllustrationInfo | |
| Items | – Name: Title Label: Title Group: Ti Data: Generating Brain MRI with StyleGAN2-ADA: The Effect of the Training Set Size on the Quality of Synthetic Images. – Name: Author Label: Authors Group: Au Data: <searchLink fieldCode="AR" term="%22Lai%2C+Matteo%22">Lai, Matteo</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> matteo.lai3@unibo.it</i><br /><searchLink fieldCode="AR" term="%22Mascalchi%2C+Mario%22">Mascalchi, Mario</searchLink><relatesTo>2</relatesTo> (AUTHOR)<i> mario.mascalchi@unifi.it</i><br /><searchLink fieldCode="AR" term="%22Tessa%2C+Carlo%22">Tessa, Carlo</searchLink><relatesTo>3</relatesTo> (AUTHOR)<i> carlo.tessa@uslnordovest.toscana.it</i><br /><searchLink fieldCode="AR" term="%22Diciotti%2C+Stefano%22">Diciotti, Stefano</searchLink><relatesTo>1,4</relatesTo> (AUTHOR)<i> stefano.diciotti@unibo.it</i> – Name: TitleSource Label: Source Group: Src Data: <searchLink fieldCode="JN" term="%22Journal+of+Imaging+Informatics+in+Medicine%22">Journal of Imaging Informatics in Medicine</searchLink>. Jun2026, Vol. 39 Issue 3, p2319-2329. 11p. – Name: Subject Label: Subjects Group: Su Data: <searchLink fieldCode="DE" term="%22Generative+adversarial+networks%22">Generative adversarial networks</searchLink><br /><searchLink fieldCode="DE" term="%22Diagnostic+imaging%22">Diagnostic imaging</searchLink><br /><searchLink fieldCode="DE" term="%22Prediction+models%22">Prediction models</searchLink><br /><searchLink fieldCode="DE" term="%22Research+funding%22">Research funding</searchLink><br /><searchLink fieldCode="DE" term="%22Brain%22">Brain</searchLink><br /><searchLink fieldCode="DE" term="%22Privacy%22">Privacy</searchLink><br /><searchLink fieldCode="DE" term="%22Magnetic+resonance+imaging%22">Magnetic resonance imaging</searchLink><br /><searchLink fieldCode="DE" term="%22Descriptive+statistics%22">Descriptive statistics</searchLink><br /><searchLink fieldCode="DE" term="%22Deep+learning%22">Deep learning</searchLink><br /><searchLink fieldCode="DE" term="%22Digital+image+processing%22">Digital image processing</searchLink><br /><searchLink fieldCode="DE" term="%22Data+quality%22">Data quality</searchLink><br /><searchLink fieldCode="DE" term="%22Sensitivity+%26+specificity+%28Statistics%29%22">Sensitivity & specificity (Statistics)</searchLink><br /><searchLink fieldCode="DE" term="%22Medical+ethics%22">Medical ethics</searchLink> – Name: Abstract Label: Abstract Group: Ab Data: The potential of deep learning for medical imaging is often constrained by limited data availability. Generative models can unlock this potential by generating synthetic data that reproduces the statistical properties of real data while being more accessible for sharing. In this study, we investigated the influence of training set size on the performance of a state-of-the-art generative adversarial network, the StyleGAN2-ADA, trained on a cohort of 3,227 subjects from the OpenBHB dataset to generate 2D slices of brain MR images from healthy subjects. The quality of the synthetic images was assessed through qualitative evaluations and state-of-the-art quantitative metrics, which are provided in a publicly accessible repository. Our results demonstrate that StyleGAN2-ADA generates realistic and high-quality images, deceiving even expert radiologists while preserving privacy, as it did not memorize training images. Notably, increasing the training set size led to slight improvements in fidelity metrics. However, training set size had no noticeable impact on diversity metrics, highlighting the persistent limitation of mode collapse. Furthermore, we observed that diversity metrics, such as coverage and β-recall, are highly sensitive to the number of synthetic images used in their computation, leading to inflated values when synthetic data significantly outnumber real ones. These findings underscore the need to carefully interpret diversity metrics and the importance of employing complementary evaluation strategies for robust assessment. Overall, while StyleGAN2-ADA shows promise as a tool for generating privacy-preserving synthetic medical images, overcoming diversity limitations will require exploring alternative generative architectures or incorporating additional regularization techniques. [ABSTRACT FROM AUTHOR] – Name: AbstractSuppliedCopyright Label: Group: Ab Data: <i>Copyright of Journal of Imaging Informatics in Medicine is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract.</i> (Copyright applies to all Abstracts.) |
| PLink | https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=egs&AN=194225505 |
| RecordInfo | BibRecord: BibEntity: Identifiers: – Type: doi Value: 10.1007/s10278-025-01536-0 Languages: – Code: eng Text: English PhysicalDescription: Pagination: PageCount: 11 StartPage: 2319 Subjects: – SubjectFull: Generative adversarial networks Type: general – SubjectFull: Diagnostic imaging Type: general – SubjectFull: Prediction models Type: general – SubjectFull: Research funding Type: general – SubjectFull: Brain Type: general – SubjectFull: Privacy Type: general – SubjectFull: Magnetic resonance imaging Type: general – SubjectFull: Descriptive statistics Type: general – SubjectFull: Deep learning Type: general – SubjectFull: Digital image processing Type: general – SubjectFull: Data quality Type: general – SubjectFull: Sensitivity & specificity (Statistics) Type: general – SubjectFull: Medical ethics Type: general Titles: – TitleFull: Generating Brain MRI with StyleGAN2-ADA: The Effect of the Training Set Size on the Quality of Synthetic Images. Type: main BibRelationships: HasContributorRelationships: – PersonEntity: Name: NameFull: Lai, Matteo – PersonEntity: Name: NameFull: Mascalchi, Mario – PersonEntity: Name: NameFull: Tessa, Carlo – PersonEntity: Name: NameFull: Diciotti, Stefano IsPartOfRelationships: – BibEntity: Dates: – D: 01 M: 06 Text: Jun2026 Type: published Y: 2026 Identifiers: – Type: issn-print Value: 29482925 Numbering: – Type: volume Value: 39 – Type: issue Value: 3 Titles: – TitleFull: Journal of Imaging Informatics in Medicine Type: main |
| ResultId | 1 |
