Leveraging Regular Fundus Images for Training UWF Fundus Diagnosis Models via Adversarial Learning and Pseudo-Labeling
This work is significant for ophthalmologists and researchers, as it provides a method to improve the training of UWF fundus diagnosis models, which are crucial for detecting peripheral retinal diseases, by utilizing existing regular fundus image datasets and reducing the need for labor-intensive UWF image annotation.
The paper addresses the challenge of training diagnostic models for ultra-widefield (UWF) fundus images, which have limited annotated data, by leveraging abundant regular fundus images. They propose a modified CycleGAN to bridge the domain gap and generate synthetic UWF images, combined with pseudo-labeling, achieving superior generalizability and performance improvements across multiple fundus disease tasks.
Recently, ultra-widefield (UWF) 200\degree~fundus imaging by Optos cameras has gradually been introduced because of its broader insights for detecting more information on the fundus than regular 30 degree - 60 degree fundus cameras. Compared with UWF fundus images, regular fundus images contain a large amount of high-quality and well-annotated data. Due to the domain gap, models trained by regular fundus images to recognize UWF fundus images perform poorly. Hence, given that annotating medical data is labor intensive and time consuming, in this paper, we explore how to leverage regular fundus images to improve the limited UWF fundus data and annotations for more efficient training. We propose the use of a modified cycle generative adversarial network (CycleGAN) model to bridge the gap between regular and UWF fundus and generate additional UWF fundus images for training. A consistency regularization term is proposed in the loss of the GAN to improve and regulate the quality of the generated data. Our method does not require that images from the two domains be paired or even that the semantic labels be the same, which provides great convenience for data collection. Furthermore, we show that our method is robust to noise and errors introduced by the generated unlabeled data with the pseudo-labeling technique. We evaluated the effectiveness of our methods on several common fundus diseases and tasks, such as diabetic retinopathy (DR) classification, lesion detection and tessellated fundus segmentation. The experimental results demonstrate that our proposed method simultaneously achieves superior generalizability of the learned representations and performance improvements in multiple tasks.