DRGen: Domain Generalization in Diabetic Retinopathy Classification
Domain generalization is a difficult challenge in deep learning, particularly in medical image analysis, because of the significant variation across various datasets. The majority of existing publications in the literature maximize performance on a single target domain., without regards to model generalizability on other domains or distributions. High discrepancy in the number of images and major domain shifts, can therefore cause single-source trained models to under-perform during testing. In this paper, we address the problem of domain generalization in Diabetic Retinopathy (DR) classification. The baseline for com- parison is set as joint training on different datasets, followed by testing on each dataset individually. We therefore introduce a method that encourages seeking a flatter minima during training while imposing a regularization. This reduces gradient variance from different domains and therefore yields satisfactory results on out-of-domain DR classification. We show that adopting DR-appropriate augmentations enhances model performance and in-domain generalizability. By performing our evaluation on 4 open-source DR datasets, we show that the proposed domain generalization method outperforms separate and joint training strategies as well as well-established methods.
M.Z.A. Atwany, "DRGen: Domain Generalization in Diabetic Retinopathy Classification", M.S. Thesis, Machine Learning, MBZUAI, Abu Dhabi, UAE, 2022.