Exploring the Effectiveness of Deep Features from Domain-Specific Foundation Models in Retinal Image Synthesis
This addresses the challenge of generating clinically accurate synthetic retinal images for medical imaging applications, but the findings are incremental as they show no advantage over existing methods.
The study investigated whether deep features from domain-specific foundation models improve retinal image synthesis, finding they do not enhance autoencoder generation, while conventional edge detection filters effectively sharpen vascular structures.
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., Fréchet 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.