Texture Feature Analysis in Fundus Image in Screening Diabetic Retinopathy

Abstract

Diabetic retinopathy is an eye ailment caused due to the impairment of nerve fibres in retina, which leads to the risk of blindness.Early diagnosis and treatment help to considerably lower the impact of vision loss. Retinal digital image captured by the fundus camera plays a vital role in supporting cost effective method of analysis and for the subsequent treatment of diabetic retinopathy.  The proposed system identifies and extracts the yellow bright lesion called exudateswhich is one of the clinical factorsassociated with diabetic retinopathy. The Haralick texture features which facilitates to accurately classify the data as presence of exudates or absence of exudates are extracted. The extent of the lesions present in the retina designates the severity of the disease. A screening simulation system has been designed to extract the exudates from the retinal image automatically.The proposed system was evaluated on publically available dataset and the performance was substantial.