Visualisation of Exudates in Fundus Images using Radar Chart and Color Auto Correlogram Technique

Fundus images provide an opportunity for early detection of diabetes. Generally, retina fundus images of diabetic patients exhibit exudates, which are lesions indicative of Diabetic Retinopathy (DR). Therefore, computational tools can be considered to be used in assisting ophthalmologists and medical doctor for the early screening of the disease. Hence in this paper, we proposed visualisation of exudates in fundus images using radar chart and Color Auto Correlogram (CAC) technique. The proposed technique requires that the Optic Disc (OD) from the fundus image be removed. Next, image normalisation was performed to standardise the colors in the fundus images. The exudates from the modified image are then extracted using Artificial Neural Network (ANN) and visualised using radar chart and CAC technique. The proposed technique was tested on 149 images of the publicly available MESSIDOR database. Experimental results suggest that the method has potential to be used for early indication of DR, by visualising the overlap between CAC features of the fundus images.