Disentangling representations of retinal images with generative models

Abstract

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.

Final journal paper version for Medical Image Analysis (MedIA)

References in corpus (31)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Conditional Generative Adversarial Nets
• GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
• Unsupervised Domain Adaptation by Backpropagation
• Measuring and testing dependence by correlation of distances
• Shortcut Learning in Deep Neural Networks
• Conditional Image Synthesis With Auxiliary Classifier GANs
• Predicting Cardiovascular Risk Factors from Retinal Fundus Photographs using Deep Learning
• InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets
• Which Training Methods for GANs do actually Converge?
• Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations
• Equivalence of distance-based and RKHS-based statistics in hypothesis testing
• Recent Advances in Autoencoder-Based Representation Learning
• Causality matters in medical imaging
• Towards a Definition of Disentangled Representations
• Invariant Representations without Adversarial Training
• Learning Disentangled Representations in the Imaging Domain
• Learning Latent Subspaces in Variational Autoencoders
• Controllable Invariance through Adversarial Feature Learning
• CLUB: A Contrastive Log-ratio Upper Bound of Mutual Information
• On Disentangled Representations Learned From Correlated Data
• Conditional Generation of Medical Images via Disentangled Adversarial Inference
• Unsupervised Representation Learning from Pre-trained Diffusion Probabilistic Models
• Semi-Supervised StyleGAN for Disentanglement Learning
• ICAM: Interpretable Classification via Disentangled Representations and Feature Attribution Mapping
• DisDiff: Unsupervised Disentanglement of Diffusion Probabilistic Models
• Multi-View Causal Representation Learning with Partial Observability
• Disentanglement of Correlated Factors via Hausdorff Factorized Support
• Generating Realistic Counterfactuals for Retinal Fundus and OCT Images using Diffusion Models
• Disentanglement and Generalization Under Correlation Shifts
• Unifying Causal Representation Learning with the Invariance Principle