Retinal Blood Vessel Segmentation Algorithm for Diabetic Retinopathy and Abnormality Classification by Supervised Machine Learning

The field of medical imaging gains its importance with increase in the need of accurate and efficient diagnosis over a short period of time. Since manual processes are tedious, time consuming and impractical for large data there is a need for automatic processing that helps community health workers. Retinal blood vessel segmentation in diabetic retinopathy plays an important role in diagnosing the pathologies, which occur as swelling in parts of the vasculature, changing of width along blood vessels, and tortuosity that later on may cause blindness. A method is presented in this work for an automated segmentation of blood vessels in the fundus images using texture, thresholding and morphological operation (combined approach) and classification by artificial neural networks. e distinct invariant features are en features for classification. Our method gives clearer and more accurate output for ophthalmologists. The implementation is observed on various types of normal and abnormal retinal images are used for the prediction of diabetes retinopathy in a given retinal image.