Table 7 Summary of literature review for feature extraction task using GAN in ophthalmology imaging domains
From: Application of generative adversarial networks (GAN) for ophthalmology image domains: a survey
Publication | Basic technique | Domain | Target | Summary |
|---|---|---|---|---|
Schlegl et al. [25] | f-AnoGAN (Wasserstein GAN + latent space mapping) | Retinal OCT | Intra-retinal fluid detection (OCT anomaly detection) | The GAN based unsupervised learning of healthy training data was trained with fast mapping from images to encodings in the latent space. Anomalies were detected via a combined anomaly score based on an image reconstruction error |
Xie et al. [26] | Conditional GAN (with attention encoder and multi-branch structure) | Ultra-widefield fundus photography (scanning laser ophthalmoscopy) | Features for retinal diseases | The GAN based on the attention encoder and multi-branch structure was used to extract features for retinal disease detection. The discriminator in GAN was modified to build the classifier to detect the disease images |