Study Group Learning: Improving Retinal Vessel Segmentation Trained with Noisy Labels

Retinal vessel segmentation from retinal images is an essential task for developing the computer-aided diagnosis system diseases. Efforts have been made on high-performance deep learning-based approaches to segment in end-to-end manner. However, acquisition of and labels requires onerous work professional clinicians, which results smaller training dataset with incomplete labels. As known, data-driven methods suffer data insufficiency, models will easily over-fit small-scale data. Such a situation becomes more severe when are or incorrect. In this paper, we propose Study Group Learning (SGL) scheme improve robustness model trained noisy Besides, learned enhancement map provides better visualization than conventional as auxiliary tool clinicians. Experiments demonstrate that proposed method further improves performance DRIVE CHASE\(\_\)DB1 datasets, especially noisy. Our code available at https://github.com/SHI-Labs/SGL-Retinal-Vessel-Segmentation.