Unsupervised ACO: Applying FCM as a supervisor for ACO in medical image segmentation

Abstract

Ant colony optimization (ACO) inspires the foraging manner of real ants in digital habitat is a developed meta-heuristic algorithm in medical image processing. Magnetic resonance (MR) images usually contain irregular and complex structures. Applying ACO, the spatial information is exploited in image processing; however, ACO requires a supervisor to define reference food. Fuzzy c-means (FCM) is an unsupervised clustering algorithm used in medical image processing. Standard FCM seldom incorporates spatial information in image segmentation. This paper presents an Unsupervised ACO (UACO) developed by FCM which utilizes the benefits of two algorithms and overlaps their defects. UACO is unsupervised like FCM and incorporates spatial information the same way as ACO. Besides, elapsed time to define food source by FCM in UACO is less than that of an operator in ACO. Adding that, another novelty of the paper is to effectively update pheromone fields. Utilizing a Gaussian spatial function, proposed approach handles noise effects properly, and exploits spatial neighborhood information efficiently in image segmentation. Experimental results show that proposed hybrid UACO by the Gaussian spatial function preserves details of image and is less sensitive to noise.