To describe and evaluate an automated system to detect exudate lesions, for early diagnosis of diabetic retinopathy and macula edema.

A total of 305 retinal images from three public available retinal image datasets are employed. An automated retinal exudate detection system was developed. In the first stage, the vessels are separated from the lesions using the morphological component analysis (MCA) algorithm with the appropriate dictionaries. Then, the lesion part of the retinal image is prepared for the detection of exudate regions. The final exudate map is created using dynamic thresholding and mathematical morphologies.

The system achieved an area under receiver operating characteristic (ROC) curve of 96.10 and 94.80 and 93.73 for DiaretDB, HEI-MED and e-ophtha datasets respectively, which is greater than most of the state-of-the-art methods.

Diabetic retinopathy is the major cause of blindness in urban populations. Exudates are prime signs of DR on the account that they cause subsequent visual loss. However, due to the growing number of diabetic patients in the world, the analysis and manual detection of retinal lesions is a time-consuming task that requires remarkable skills. As such, developing an automatic technique for the analysis of images is of prominent importance. This system with high sensitivity can be used in a screening tool in the reading centers to reduce the workload of the ophthalmologists.