Retinal blood vessel segmentation using graph cut analysis

The automated segmentation of blood vessels helps the ophthalmologist for early detection and possible treatment of retinal diseases. This paper presents a novel method for automatic segmentation of blood vessels using graph cut method. Initially, we applied mean filter, convolution by Gaussian kernel, shade correction and top-hat transformation as preprocessing steps for enhancement of blood vessels. It significantly enhance retinal image while suppress the noise and non-vessel structures keeping vessel information. Then vascular structure is extracted using graph cut segmentation. The proposed approach is tested on publicly available DRIVE dataset. Performance analysis is carried out and compared with other methods. The values achieved with our novel method for area under curve, accuracy, sensitivity and specificity are 0.9605, 0.9626, 0.7261 and 0.9806 respectively. This performance parameter comparison shows the effectiveness of our method for improving the segmentation results and hence detects blood vessels accurately.