Research Article
EAD-Net: A Novel Lesion Segmentation Method in Diabetic Retinopathy Using Neural Networks
Table 2
Comparison with top 10 teams in the lesion segmentation competition on IDRiD dataset.
| | Model (team) | MAs | HEs | Hard exudates | Soft exudates |
| | VRT (1st) | 0.4951 | 0.6804 | 0.7127 | 0.6995 | | PATech (2nd) | 0.4740 | 0.6490 | 0.8850 | — | | iFLYTEK-MIG (3rd) | 0.5017 | 0.5588 | 0.8741 | 0.6588 | | SOONER (4th) | 0.4003 | 0.5395 | 0.7390 | 0.5369 | | SHAIST (5th) | — | — | 0.8582 | — | | lzyuncc_fusion (6th) | — | — | 0.8202 | 0.6259 | | SDNU (7th) | 0.4111 | 0.4572 | 0.5018 | 0.5374 | | CIL (8th) | 0.3920 | 0.4886 | 0.7554 | 0.5024 | | MedLabs (9th) | 0.3397 | 0.3705 | 0.7863 | 0.2637 | | AIMIA (10th) | 0.3792 | 0.3283 | 0.7662 | 0.2733 | | Proposed EAD-Net | 0.2408 | 0.5649 | 0.7818 | 0.6083 |
|
|
The results are based on AUPR (Area under Precision-Recall curve).
|
