Detection of changes in color fundus images due to diabetic retinopathy

Condition of the retina can be known from the color fundus image. Appearance and location of different features such as exudates, hemorrhages and drusen in color fundus image are used for the assessment of diabetic retinopathy (DR). DR affects the structure of blood vessels and optic disc (OD). In this paper we propose a novel method for detecting the changes appeared in the color fundus image due to the progression of diabetics by using independent component analysis (ICA) on wavelet subbands. Wavelet subband decomposition helps highlighting different information such as blood vessels, optic disc, exudates, hemorrhages, macula and drusen in different subband. ICA is applied on the wavelet subbands to find the different changes. This method is tested on 50 number of color fundus images. The changes appeared in the fundus image are detected correctly.