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

摘要: Diabetic macular edema (DME) is a common vision threatening complication of diabetic retinopathy. In large scale screening environment DME can be assessed by detecting exudates (a type bright lesions) in fundus images. this work, we introduce new methodology for diagnosis using novel set features based on colour, wavelet decomposition and automatic lesion segmentation. These are employed to train classifier able automatically diagnose through the presence exudation. We present publicly available dataset with ground-truth data containing 169 patients from various ethnic groups levels DME. This other two datasets evaluate our algorithm. achieve performance comparable retina experts MESSIDOR (an independently labelled 1200 images) cross-dataset testing (e.g., was trained an independent tested MESSIDOR). Our algorithm obtained AUC between 0.88 0.94 depending dataset/features used. Additionally, it does not need ground truth at level reject false positives computationally efficient, as generates average 4.4 s (9.3 s, considering optic nerve localisation) per image 2.6 GHz platform unoptimised Matlab implementation.