Automatic Exudate Detection from Retinal Fundus Images in Diabetic Retinopathy.

Diabetes results in diabetic retinopathy. which occurs due to a change in the vasculature structure. Diabetic abnormalities such as microaneurysms, haemorrhages, and "exudates" describe the stage of diabetic retinopathy. Early-stage detection gains prime importance to avoid permanent sight loss. Exudates are one of the most noticeable signs of diabetic retinopathy. This work presents the automatic and non-invasive approach to detect exudates using requisitely adjusted morphological operators. The use of Standard deviation of histogram equalized image after optic disc detection and deletion is the work's key contribution. The standard deviation offers the tendency of distributing the favourable exudate pixels towards the mean value. The proposed method is implemented on the "IDRiD (Indian Diabetic Retinopathy Image Dataset)". The detected exudates are compared with provided ground-truths reported accuracy of 99.13%. The outcomes prove its effectiveness of the proposed method and can aid ophthalmologists in diagnosis. [ABSTRACT FROM AUTHOR]