Transfer Learning from the Domain of Diabetic Retinopathy to Aid in the Detection of Age-Related Macular Degeneration

Age-related macular degeneration (AMD) is the leading cause of vision loss in the elderly population. Transfer learning has proven useful in fundus image analysis for early diagnosis. Previous models were pre-trained on the ImageNet database. However, a source domain related to retinal diagnosis would facilitate model learning. Our objective was to apply transfer learning from the domain of diabetic retinopathy (DR) to aid in the detection of AMD (binary classification). The proposed model was based on the ResNet-RS architecture. Pre-training aimed at DR diagnosis was conducted using the Kaggle database. Then, fine-tuning was performed using the Automatic Detection challenge on Age-related Macular degeneration (ADAM) dataset. We carried out 3 experiments with different number of images used for fine-tuning. As the main result, our method showed a much faster convergence than the corresponding models pre-trained on ImageNet. Additionally, the proposed source domain was proven especially useful when scarce data in the destination domain was available.