A Comparative Study of Automatic Vessel Segmentation Algorithms

Vessels appearance and their morphological features play a vital part in timely treatment of numerous diseases, such as vein occlusions and diabetic retinopathy. This paper presents a detailed comparison of three recently developed vessel segmentation algorithms in terms of Accuracy (Acc), Sensitivity (Se), and Specificity (Sp) on two publicly available DRIVE and STARE datasets. Our simulations indicate that for high image resolution of 400×500 pixels or above and on DRIVE dataset the frangi and Otsu thresholding based vessel segmentation algorithm yields the highest Accuracy. Whereas on STARE dataset, Unet based convolutional neural network based vessel segmentation algorithm outperforms the compared algorithms at the cost of higher computational time.