EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Synthetic Scientific Image Generation with VAE, GAN, and Diffusion Model Architectures

Zineb Sordo iD, Eric Chagnon iD, Zixi Hu iD, Jeffrey J Donatelli iD, Peter Andeer iD, Peter S Nico iD, Trent Northen iD, Daniela Ushizima iD

Journal of Imaging · 26.07.2025 · 11 (8) · 252

DOI: 10.3390/jimaging11080252 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Imaging, online erschienen am 26.07.2025. Verfasst von Zineb Sordo und 7 weiteren Autor:innen. Beteiligte Einrichtung: Applied Math and Computational Research Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA. Die Arbeit wurde bisher 43-mal zitiert (Crossref) und nennt 90 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: US Department of Energy (DOE) Office of Science Advanced Scientific Computing Research (ASCR), Office of Basic Energy Sciences (BES).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Generative AI (genAI) has emerged as a powerful tool for synthesizing diverse and complex image data, offering new possibilities for scientific imaging applications. This review presents a comprehensive comparative analysis of leading generative architectures, ranging from Variational Autoencoders (VAEs) to Generative Adversarial Networks (GANs) on through to Diffusion Models, in the context of scientific image synthesis. We examine each model’s foundational principles, recent architectural advancements, and practical trade-offs. Our evaluation, conducted on domain-specific datasets including microCT scans of rocks and composite fibers, as well as high-resolution images of plant roots, integrates both quantitative metrics (SSIM, LPIPS, FID, CLIPScore) and expert-driven qualitative assessments. Results show that GANs, particularly StyleGAN, produce images with high perceptual quality and structural coherence. Diffusion-based models for inpainting and image variation, such as DALL-E 2, delivered high realism and semantic alignment but generally struggled in balancing visual fidelity with scientific accuracy. Importantly, our findings reveal limitations of standard quantitative metrics in capturing scientific relevance, underscoring the need for domain-expert validation. We conclude by discussing key challenges such as model interpretability, computational cost, and verification protocols, and discuss future directions where generative AI can drive innovation in data augmentation, simulation, and hypothesis generation in scientific research.

Schlagwörter

Stichwörter der Autor:innen

Generative Adversarial Networksdiffusiongenerative AIimage generationsynthetic data

Autor:innen und Institutionen

  • Applied Math and Computational Research Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA.
  • Environmental Genomics and Systems Biology Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA.
  • Earth Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA.
  • Bakar Computational Health Sciences Institute, University of California, San Francisco, CA 94158, USA.
  • Berkeley Institute for Data Science, University of California, Berkeley, CA 94720, USA.

ORCID-Profile: Zineb Sordo, Eric Chagnon, Zixi Hu, Jeffrey J Donatelli, Peter Andeer, Peter S Nico, Trent Northen, Daniela Ushizima

Förderung

  • US Department of Energy (DOE) Office of Science Advanced Scientific Computing Research (ASCR)
  • Office of Basic Energy Sciences (BES)

Zitieren

APA

Sordo, Z., Chagnon, E., Hu, Z., Donatelli, J. J., Andeer, P., Nico, P. S., Northen, T., & Ushizima, D. (2025). Synthetic Scientific Image Generation with VAE, GAN, and Diffusion Model Architectures. Journal of Imaging, 11(8), 252. https://doi.org/10.3390/jimaging11080252

Vancouver

Sordo Z, Chagnon E, Hu Z, Donatelli JJ, Andeer P, Nico PS, et al. Synthetic Scientific Image Generation with VAE, GAN, and Diffusion Model Architectures. Journal of Imaging. 2025;11(8):252. doi:10.3390/jimaging11080252

BibTeX
@article{Sordo2025,
  author = {Sordo, Zineb and Chagnon, Eric and Hu, Zixi and Donatelli, Jeffrey J and Andeer, Peter and Nico, Peter S and Northen, Trent and Ushizima, Daniela},
  title = {Synthetic Scientific Image Generation with VAE, GAN, and Diffusion Model Architectures},
  journal = {Journal of Imaging},
  year = {2025},
  volume = {11},
  number = {8},
  pages = {252},
  doi = {10.3390/jimaging11080252}
}
Kostenlos starten →