Table 1. Quantitative results of different methods in PEDs segmentation.

Method	U-Net++	Att-U-Net	ResNet	RetfluidNet	Swin-Unet	Our method
Dice	0.675 ± 0.234	0.650 ± 0.233	0.556 ± 0.262	0.713 ± 0.154	0.785 ± 0.183	0.850 ± 0.137
Precision	0.913 ± 0.222	0.868 ± 0.277	0.692 + 0.296	0.861 + 0.206	0.892 ± 0.173	0.835 ± 0.147
Accuracy	0.986 ± 0.007	0.986 ± 0.007	0.980 ± 0.010	0.991 ± 0.004	0.986 ± 0.008	0.992 ± 0.003
Recall	0.599 ± 0.210	0.638 ± 0.202	0.555 ± 0.242	0.873 ± 0.083	0.616 ± 0.185	0.931 ± 0.128
IOU	0.535 ± 0.204	0.562 ± 0.202	0.448 ± 0.206	0.767 ± 0.181	0.562 ± 0.172	0.790 ± 0.171
F1-score	0.526 ± 0.225	0.548 ± 0.225	0.426 ± 0.247	0.741 ± 0.151	0.615 ± 0.178	0.759 ± 0.136