EAD-Net: A Novel Lesion Segmentation Method in Diabetic Retinopathy Using Neural Networks

Figure 11

An example of segmentation results. (a) Original image. (b) The corresponding ground truth annotations of different lesions. (c) The predicted segmentation results. Note that the areas marked out in different colors represent different lesions: MAs in red, HEs in green, hard exudates in blue, and soft exudates in yellow. The predicted result in (c) also provides counting statistics of the four lesions. In the rectangular box on the right of (c), the number of lesions is obtained by connected components analysis of the corresponding lesion areas in the left of (c).

(a)

(b)

(c)