Minimal Training Time in Supervised Retinal Vessel Segmentation

In this paper, we perform comparative analysis between different classifiers using the same experimental setup for supervised retinal vessel segmentation. The aim of this paper is to find supervised classifier that can obtain good segmentation accuracy with minimal training time. Minimizing the training time is essential when dealing with biomedical images. The more samples introduced to a learning model, the better it can adapt to the unseen data. The results indicate a trade-off between accuracy and training time can be obtained in a classifier trained by a Neural Network. When tested with a publicly available database, the learning model only requires less than 2 minutes in the learning phase and achieves overall accuracy of 94.54%.