Exudate-based diabetic macular edema detection in fundus images using publicly available datasets

► Presentation of a new dataset (HEI-MED) for retinal diabetic macular edema (DME). ► New method for the automatic detection of DME with fundus images. ► New image based feature vector approach with colour, wavelets and exudate probability. ► Good computational performance. Automatic analysis is completed in ∼4.4 s (∼9.3 s with the automatic ON localization). ► Automatic diagnosis is comparable to manual diagnosis of two retina experts.