A Novel Medical Image Edge Detection Method Based on Reinforcement Learning and Ant Colony Optimization

Edge detection is one of the most essential steps and research focuses in medical imaging. In recent years, ant colony optimization has been widely used in medical image edge detection due to its robustness and accuracy. To further improve the performance of ant colony optimization based medical image edge detection methods, in this paper we proposed a novel strategy combining ant colony optimization and machine learning. At first, instead of using a constant number of neighborhood pixels to calculate the heuristic information for each pixel, we integrate multi-agent reinforcement learning into the movement of artificial ants to select variable perceived radius to calculate heuristic information. Additionally, another adaptive parameter is presented to control the moving direction of artificial ants in order for jumping from local optima. The proposed method is evaluated on typical medical images, and the experimental results show that the proposed method can perform high-precision edge detection for medical images.