# Exploring the Effectiveness of Deep Features from Domain-Specific Foundation Models in Retinal Image Synthesis

arXiv: https://arxiv.org/abs/2506.11753 · HTML: https://mlsanity.com/paper/2506.11753/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 20 of its 32 references that are in the map.
- First posted on arXiv 13 Jun 2025.

## Builds on (most cited first, 20 of 20)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2006.11239/) (2020)
- [Very Deep Convolutional Networks for Large-Scale Image Recognition](https://mlsanity.com/paper/1409.1556/) (2014)
- [The Unreasonable Effectiveness of Deep Features as a Perceptual Metric](https://mlsanity.com/paper/1801.03924/) (2018)
- [Denoising Diffusion Implicit Models](https://mlsanity.com/paper/2010.02502/) (2020)
- [Rethinking the Inception Architecture for Computer Vision](https://mlsanity.com/paper/1512.00567/) (2015)
- [Neural Discrete Representation Learning](https://mlsanity.com/paper/1711.00937/) (2017)
- [Taming Transformers for High-Resolution Image Synthesis](https://mlsanity.com/paper/2012.09841/) (2020)
- [PeCo: Perceptual Codebook for BERT Pre-training of Vision Transformers](https://mlsanity.com/paper/2111.12710/) (2021)
- [Synthetic Data -- what, why and how?](https://mlsanity.com/paper/2205.03257/) (2022)
- [TADM: Temporally-Aware Diffusion Model for Neurodegenerative Progression on Brain MRI](https://mlsanity.com/paper/2406.12411/) (2024)
- [Morphology-preserving Autoregressive 3D Generative Modelling of the Brain](https://mlsanity.com/paper/2209.03177/) (2022)
- [LeFusion: Controllable Pathology Synthesis via Lesion-Focused Diffusion Models](https://mlsanity.com/paper/2403.14066/) (2024)
- [When Diffusion MRI Meets Diffusion Model: A Novel Deep Generative Model for Diffusion MRI Generation](https://mlsanity.com/paper/2408.12897/) (2024)
- [SynCellFactory: Generative Data Augmentation for Cell Tracking](https://mlsanity.com/paper/2404.16421/) (2024)
- [Self-supervised Vision Transformer are Scalable Generative Models for Domain Generalization](https://mlsanity.com/paper/2407.02900/) (2024)
- [On Biases in a UK Biobank-based Retinal Image Classification Model](https://mlsanity.com/paper/2408.02676/) (2024)
- [Paired Diffusion: Generation of related, synthetic PET-CT-Segmentation scans using Linked Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2403.17734/) (2024)
- [FundusGAN: A Hierarchical Feature-Aware Generative Framework for High-Fidelity Fundus Image Generation](https://mlsanity.com/paper/2503.17831/) (2025)

## Built on by (most cited first, 1 of 1)

- [Evaluation of Clinically Steerable Retinal Image Generation from Foundation Model Latent Spaces](https://mlsanity.com/paper/2608.13455/) (2026)

## Abstract (arXiv)

The adoption of neural network models in medical imaging has been constrained by strict privacy regulations, limited data availability, high acquisition costs, and demographic biases. Deep generative models offer a promising solution by generating synthetic data that bypasses privacy concerns and addresses fairness by producing samples for under-represented groups. However, unlike natural images, medical imaging requires validation not only for fidelity (e.g., Fr'echet Inception Score) but also for morphological and clinical accuracy. This is particularly true for colour fundus retinal imaging, which requires precise replication of the retinal vascular network, including vessel topology, continuity, and thickness. In this study, we in-vestigated whether a distance-based loss function based on deep activation layers of a large foundational model trained on large corpus of domain data, colour fundus imaging, offers advantages over a perceptual loss and edge-detection based loss functions. Our extensive validation pipeline, based on both domain-free and domain specific tasks, suggests that domain-specific deep features do not improve autoen-coder image generation. Conversely, our findings highlight the effectiveness of con-ventional edge detection filters in improving the sharpness of vascular structures in synthetic samples.