Preprocessing and Segmentation of Retinal Blood Vessels in Fundus Images using U-Net

Deep Learning plays an important role today in disease detection and prediction. All deep learning models need to be trained to process the input; Extract features and return prediction results. Before classification and prediction, the given input must be preprocessed to perform segmentation using augmentation. Only with the help of preprocessed images each model can make accurate predictions at higher speeds. This proposed work aimed to detect Diabetic Eye Diseases by means of segmenting the augmented images using U-Net. U-Net is familiar with its Encoder-Decoder architecture for sampling. Retinal Blood Vessel is one of the most precise parts of an eye. Based on the nature of this blood vessel one can identify whether it is affected by diabetic retinopathy or not. So, segmenting the blood vessel helps to classify the disease category in early stage and of course U-Net is probably meant for segmenting medical images. In this paper, the discussions will be made on preprocessing eye images from the data set, segmenting those images using U-Net to extract the retinal blood vessel, classification based on segmentation