Using histopathology latent diffusion models as privacy-preserving dataset augmenters improves downstream classification performance

Jan Moritz Niehues, Gustav Müller-Franzes, Yoni Schirris, Sophia J. Wagner, Michael Jendrusch, Matthias Kloor, Alexander T. Pearson, Hannah Sophie Muti, Katherine Jane Hewitt, Gregory Patrick Veldhuizen, Laura Zigutyte, Daniel Truhn, Jakob Nikolas Kather

Abstract

Highlights•Latent Diffusion Models (LDMs) generate realistic images of histopathological tissue.•LDMs beat generative adversarial networks (GANs) across several evaluation metrics.•Our best LDM shows no data memorization.•Deployment of LDMs improves downstream classification tasks.
BibTeX
@article{niehues2024using,
  author = {Jan Moritz Niehues and Gustav Müller-Franzes and Yoni Schirris and Sophia J. Wagner and Michael Jendrusch and Matthias Kloor and Alexander T. Pearson and Hannah Sophie Muti and Katherine Jane Hewitt and Gregory Patrick Veldhuizen and Laura Zigutyte and Daniel Truhn and Jakob Nikolas Kather},
  title = {{Using histopathology latent diffusion models as privacy-preserving dataset augmenters improves downstream classification performance}},
  journal = {Computers in Biology and Medicine},
  year = {2024},
  doi = {10.1016/j.compbiomed.2024.108410},
}

Loading related content...

Figure