Staircase-Net: a deep learning based architecture for retinal blood vessel segmentation

The retina is one of the most metabolically active tissues in the human body. The retinal vessels provide blood to the inner retinal neurons. The retinal blood vessels are affected by diseases such as Hypertensive retinopathy and Diabetic retinopathy. The early diagnosis prevents the patients from blindness and fatality in some cases. Thus, examining the retinal blood vessels becomes an important work of an ophthalmologist. Thus, automated retinal blood vessel segmentation aids the ophthalmologist and makes their work easier. In this paper, a supervised Convolutional Neural Network (CNN) is suggested that enhances the performance of retinal blood vessel segmentation. Three publicly available datasets are used: STARE, DRIVE, and CHASE_DB1. A novel model, 'Staircase-Net,' is proposed, which has a series of up-sampling and down-sampling processes for feature extraction (extracting the thick and thin blood vessel features, respectively). The images in the datasets undergo a series of transformations in the preprocessing steps. The evaluation metrics considered are specificity, accuracy, sensitivity, and area under the curve. Finally, the proposed model results are compared with the state-of-the-art techniques.