A Swarm Optimization-Based Kmedoids Clustering Technique for Extracting Melanoma Cancer Features

Abstract

Melanoma is a dangerous type of skin cancers. It is alarming to see the increase of this noxious disease in modern societies, however, it can be cured by surgical excision if it is detected early. In this paper, a swarm-based clustering technique for detecting melanoma is developed. Meaningful colour features from images are extracted, and a new objective function is introduced by applying an efficient and fast linear transformation to detect Melanoma. Specifically, the proposed technique consists of three main phases. The first phase is a pre-processing stage to organize data into proper attributes, while the subsequent two phases comprise iterative swarm optimisation procedures. The iterative swarm optimisation procedures involve a linear transformation to convert the existing colour components into a new colour space, formulation of the Kmedoids objective function, and error minimisation of the particle swarm optimisation (PSO) solutions. The Otsu threshold technique is utilised to provide binary images. The proposed technique is efficient and effective due to its linearity and simplicity.