Automated localisation of the optic disc, fovea, and retinal blood vesselsfrom digital colour fundus images

Aim-To recognise automatically the main components of the fundus on digital colour images. Methods-The main features of a fundus retinal image were defined as the opt ic disc, fovea, and blood vessels. Methods are described for their automati c recognition and location. 112 retinal images were preprocessed via adapti ve, local, contrast enhancement. The optic discs were located by identifyin g the area with the highest variation in intensity of adjacent pixels. Bloo d vessels were identified by means of a multilayer perceptron neural net, f or which the inputs were derived from a principal component analysis (PCA) of the image and edge detection of the first component of PCA. The foveas w ere identified using matching correlation together with characteristics typ ical of a fovea-for example, darkest area in the neighbourhood of the optic disc. The main components of the image were identified by an experienced o phthalmologist for comparison with computerised methods. Results-The sensitivity and specificity of the recognition of each retinal main component was as follows: 99.1% and 99.1% for the optic disc; 83.3% an d 91.0% for blood vessels; 80.4% and 99.1% for the fovea. Conclusions-In this study the optic disc, blood vessels, and fovea were acc urately detected. The identification of the normal components of the retina l image will aid the future detection of diseases in these regions. In diab etic retinopathy, for example, an image could be analysed for retinopathy w ith reference to sight threatening complications such as disc neovascularis ation, vascular changes, or foveal exudation.