Multi-strategy ant colony optimization for multi-level image segmentation: Case study of melanoma

Abstract

Melanoma, which results from the cancerous transformation of melanocytes, is the most dangerous skin cancer in the medical field. Today, image processing technology has been widely used in medical fields, and image segmentation plays an important role. Therefore, this work studied the multi-level image segmentation method based on the swarm intelligence algorithm on melanoma pathological images to improve the disease diagnosis. Firstly, an improved ant colony optimizer is proposed, named LACOR. The proposed algorithm introduces the sine cosine strategy (SC), disperse foraging strategy (DFS), and specular reflection learning strategy (SRL) to the original ant colony optimizer. The role of SC is to improve the global search capability of the algorithm. Moreover, DFS and SRL allow the algorithm to jump out of the local optimum. To prove the LACOR’s performance, this work designs a series of experiments with its counterparts on IEEE CEC2014. Experimental results show that LACOR has better convergence speed and accuracy. Meanwhile, a novel multi-level image segmentation model based on LACOR is proposed by combining the non-local mean strategy and 2D Kapur’s entropy strategy applied to the melanoma pathological image. First, the proposed model performs experiments of multi-level image segmentation based on the standard image of BSDS500. Then, this work designs image segmentation experiments based on pathological images of melanoma. This work uses the feature similarity index, structural similarity index, and peak signal-to-noise ratio as evaluation metrics for image segmentation results. The proposed image segmentation model has a higher image segmentation quality than other counterparts. Therefore, the proposed method has the potential to enhance for helping the diagnosis of melanoma.