mlsanityData to 7 Oct 2026

Disentangling representations of retinal images with generative models

arXiv 2402.19186 ↗First posted 29 Feb 2024cs.CV cs.LGcited by 8

Not on the map. Why?

Abstract (arXiv)

Retinal fundus images play a crucial role in the early detection of eye diseases. However, the impact of technical factors on these images can pose challenges for reliable AI applications in ophthalmology. For example, large fundus cohorts are often confounded by factors like camera type, bearing the risk of learning shortcuts rather than the causal relationships behind the image generation process. Here, we introduce a population model for retinal fundus images that effectively disentangles patient attributes from camera effects, enabling controllable and highly realistic image generation. To achieve this, we propose a disentanglement loss based on distance correlation. Through qualitative and quantitative analyses, we show that our models encode desired information in disentangled subspaces and enable controllable image generation based on the learned subspaces, demonstrating the effectiveness of our disentanglement loss. The project's code is publicly available: https://github.com/berenslab/disentangling-retinal-images.

Key facts. Cited by 8 papers in the map (data through 7 Oct 2026). Builds on 48 of its 81 references that are in the map. First posted on arXiv 29 Feb 2024.

Builds on

48 of 81 references.

References · 48

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,267
  3. 2019Analyzing and Improving the Image Quality of StyleGANcited by 3,580
  4. 2015Domain-Adversarial Training of Neural Networkscited by 3,023
  5. 2014Conditional Generative Adversarial Netscited by 2,152
  6. 2020Shortcut Learning in Deep Neural Networkscited by 1,928
  7. 2014Unsupervised Domain Adaptation by Backpropagationcited by 1,906
  8. 2018Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representationscited by 929
  9. 2016Deep Variational Information Bottleneckcited by 915
  10. 2016InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Netscited by 858
  11. 2018MINE: Mutual Information Neural Estimationcited by 680
  12. 2018Which Training Methods for GANs do actually Converge?cited by 672
36 more
  1. 2016Conditional Image Synthesis With Auxiliary Classifier GANscited by 562
  2. 2019On Variational Bounds of Mutual Informationcited by 550
  3. 2021Diffusion Autoencoders: Toward a Meaningful and Decodable Representationcited by 447
  4. 2018Diverse Image-to-Image Translation via Disentangled Representationscited by 296
  5. 2018Towards a Definition of Disentangled Representationscited by 294
  6. 2018cGANs with Projection Discriminatorcited by 253
  7. 2020CLUB: A Contrastive Log-ratio Upper Bound of Mutual Informationcited by 233
  8. 2015The Variational Fair Autoencodercited by 223
  9. 2019Causality matters in medical imagingcited by 193
  10. 2018Recent Advances in Autoencoder-Based Representation Learningcited by 130
  11. 2017Controllable Invariance through Adversarial Feature Learningcited by 127
  12. 2021Explaining in Style: Training a GAN to explain a classifier in StyleSpacecited by 122
  13. 2018Invariant Representations without Adversarial Trainingcited by 111
  14. 2020On Disentangled Representations Learned From Correlated Datacited by 94
  15. 2023Multi-View Causal Representation Learning with Partial Observabilitycited by 72
  16. 2020Measuring Disentanglement: A Review of Metricscited by 69
  17. 2017Predicting Cardiovascular Risk Factors from Retinal Fundus Photographs using Deep Learningcited by 63
  18. 2022Unsupervised Representation Learning from Pre-trained Diffusion Probabilistic Modelscited by 53
  19. 2018Learning Latent Subspaces in Variational Autoencoderscited by 51
  20. 2021Learning Disentangled Representations in the Imaging Domaincited by 46
  21. 2020Semi-Supervised StyleGAN for Disentanglement Learningcited by 43
  22. 2022Disentanglement of Correlated Factors via Hausdorff Factorized Supportcited by 37
  23. 2019PathologyGAN: Learning deep representations of cancer tissuecited by 32
  24. 2023DisDiff: Unsupervised Disentanglement of Diffusion Probabilistic Modelscited by 30
  25. 2020ICAM: Interpretable Classification via Disentangled Representations and Feature Attribution Mappingcited by 20
  26. 2021ICAM-reg: Interpretable Classification and Regression with Feature Attribution for Mapping Neurological Phenotypes in Individual Scanscited by 18
  27. 2022On the Versatile Uses of Partial Distance Correlation in Deep Learningcited by 18
  28. 2020Measuring the Biases and Effectiveness of Content-Style Disentanglementcited by 14
  29. 2020Conditional Generation of Medical Images via Disentangled Adversarial Inferencecited by 13
  30. 2021Adversarial learning of cancer tissue representationscited by 8
  31. 2021Disentanglement and Generalization Under Correlation Shiftscited by 8
  32. 2021InvGAN: Invertible GANscited by 7
  33. 2022Instability and Local Minima in GAN Training with Kernel Discriminatorscited by 7
  34. 2023A2V: A Semi-Supervised Domain Adaptation Framework for Brain Vessel Segmentation via Two-Phase Training Angiography-to-Venography Translationcited by 3
  35. 2023Generating Realistic Counterfactuals for Retinal Fundus and OCT Images using Diffusion Modelscited by 3
  36. 2020Unsupervised Discovery, Control, and Disentanglement of Semantic Attributes with Applications to Anomaly Detectioncited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 1; 2025: 3; 2026: 4.

8 papers.

  1. 2025A Pragmatic Note on Evaluating Generative Models with Fr'echet Inception Distance for Retinal Image Synthesiscited by 8
  2. 2024Benchmarking Dependence Measures to Prevent Shortcut Learning in Medical Imagingcited by 2
  3. 2025Enhancing Retinal Vessel Segmentation Generalization via Layout-Aware Generative Modellingcited by 0
  4. 2025Disentanglement and Assessment of Shortcuts in Ophthalmological Retinal Imaging Examscited by 0
  5. 2026SRA-Seg: Synthetic to Real Alignment for Semi-Supervised Medical Image Segmentationcited by 0
  6. 2026MeDUET: Disentangled Unified Pretraining for 3D Medical Image Synthesis and Analysiscited by 0
  7. 2026Mitigating Shortcut Learning via Feature Disentanglement in Medical Imaging: A Benchmark Studycited by 0
  8. 2026Ordinal Diffusion Models for Color Fundus Imagescited by 0