Segmentation of retinal fundus vasculature in nonmydriatic camera images using wavelets

Diabetes and its associated complications, including diabetic retinopathy (DR), have been identified as a significant growing global public health problem [1]. Within this context, an important step toward reducing the number of individuals seriously affected by diabetic retinopathy is to simplify the procedure used to diagnose the condition and ensure that early eye examinations become routine for diabetic patients [1-3]. However, large population screening, especially in rural and remote areas, is difficult because of the geographical and economic isolation associated with the cost of visits to specialists and the lack of ophthalmologists available [4, 5]. Studies that trialed nonspecialist health workers in identifying diabetic eye disease have shown that correct identification is no better than random in all cases tested (50) and for identifying earlier stages of proliferation [6]. Introducing a simple and economical means to quantitatively assess the health status of the optic fundus while visiting rural and remote communities would provide support to health workers not specialized in identification of optic fundus pathology and timely, on-the-spot feedback to the individuals being assessed.