Automatic extraction of retinal vessels based on gradient orientation analysis

Retinal vessel extraction is important for the diagnosis of numerous eye diseases. It plays an important role in automatic retinal disease screening systems. This paper presents an efficient method for the automated analysis of retinal images. Fine anatomical features, such as blood vessels, are detected by analyzing the gradient orientation of the retinal images. The method is independent of image intensity and gradient magnitude; therefore, it performs accurately despite the common problems inherent to the retinal images, such as low contrast and non-uniform illumination. Blood vessels with varying diameters are detected by applying this method at multiple scales. The blood vessel network is then extracted from the detected features by manual thresholding followed by a few simple morphological operations. Based on the binary vessel map obtained, we attempt to evaluate the performance of the proposed algorithm on two publicly available databases (DRIVE and STARE database) of manually labeled images. The receiver operating characteristics (ROC), area under ROC and segmentation accuracy is taken as the performance criteria. The results demonstrate that the proposed method outperforms other unsupervised methods in respect of maximum average accuracy (MAA). The proposed method results in the area under ROC and the accuracy of 0.9037, 0.9358 for DRIVE database 0.9117, 0.9423 for STARE database respectively.