Cross-Dataset Generalization For Retinal Lesions Segmentation
This work addresses the challenge of dataset generalization for automated retinal disease diagnosis, but it is incremental as it compares existing techniques without introducing new methods.
The study tackled the problem of generalizing retinal lesion segmentation models across datasets with annotation discrepancies by comparing techniques like stochastic weight averaging, model soups, and ensembles, resulting in insights on combining coarsely and finely labeled data to improve segmentation.
Identifying lesions in fundus images is an important milestone toward an automated and interpretable diagnosis of retinal diseases. To support research in this direction, multiple datasets have been released, proposing groundtruth maps for different lesions. However, important discrepancies exist between the annotations and raise the question of generalization across datasets. This study characterizes several known datasets and compares different techniques that have been proposed to enhance the generalisation performance of a model, such as stochastic weight averaging, model soups and ensembles. Our results provide insights into how to combine coarsely labelled data with a finely-grained dataset in order to improve the lesions segmentation.