Comparative analysis on supervised classification techniques for segmentation and detecting abnormal blood vessels in retinal images

The development of new vessels on the retina of people with diabetes is rare., but is likely to lead to severe visual impairment. The technique implements a supervised classification method for blood vessel detection as well as new vessels on the optic disc in digital retinal images. Blood vessel segmentation is performed through various stages: Preprocessing., Feature Extraction by using Gray-level and Moment Invariants-based., Classification and Post processing. For new vessel detection., the fourteen features are chosen based on their discrimination capability and absence of correlation with other features. Classification is performed using a Support Vector Machine. The system is trained and tested by cross-validation using 25 images with new vessels and 25 normal images without new vessels.