JOINEDTrans: Prior guided multi-task transformer for joint optic disc/cup segmentation and fovea detection

Huaqing He, Jiaming Qiu, Li Lin, Zhiyuan Cai, Pujin Cheng, Xiaoying Tang

Abstract: Highlights•Vessels can guide optic cup/disc segmentation, macula localization.•Simultaneously performing both tasks enhances robustness.•Vessel pretraining encoder proving superior to a normal retinal encoder.•Star-shape loss expedites convergence, correcting erroneous segmentation results.