Retinal image analysis aimed at extraction of vascular structure using linear discriminant classifier

Automatic segmentation of the retinal vasculature is considered as a first step in computer assisted medical applications related to diagnosis and treatment planning. This paper describes a pixel classification based method of segmenting retinal blood vessels using linear discriminant analysis. The vessel-ness measure of a pixel is defined by the feature vector comprised of a modified multiscale line operator and Gabor filter responses. The sequential forward feature selection scheme is used to identify the optimal scales for the line operator and Gabor filter. The linear discriminant classifier utilizes only two features for pixel classification. The feature vector encodes information to reliably handle normal vessels in addition to vessels with strong light reflexes along their centerline, which is more apparent on retinal arteriolars than venules. The method is evaluated on the three publicly available DRIVE, STARE and MESSIDOR datasets. The method is computationally fast and its performance approximates the 2^nd human observer as well as other existing methodologies available in the literature, thus making it a suitable tool for automated retinal image analysis.