Optimization of Vessel Segmentation Using Genetic Algorithms

Abstract

Vessel segmentation is an important task to extract helpful information from retinal images that can help make a retinopathy diagnosis. A good segmentation perfectly represents the structure and obtains patterns that diagnose retinal diseases. Most of the current methods require many parameters, and the final quality of vessel segmentation depends on these parameters, which increases the complexity of the methods. We propose a new Vessel segmentation algorithm to address these issues using genetic algorithms. The method uses several steps to segment the retinal images. However, each of the parameters used in the steps is optimized by the genetic algorithm. To evaluate the performance of the proposed method, we achieved experiments with two freely accessible datasets for vessel segmentation, digital retinal images for vessel extraction (Drive) and the Child Heart Health Study in England (Chase-db1). Experimental results show an acceptable performance of the proposed method using sensitivity (0.7941), specificity (0.9451), and accuracy (0.9578) performance metrics.