Image Segmentation of Breast Cancer Histopathology Images Using PSO-Based Clustering Technique

Abstract

Image segmentation has key influence in numerous medical imaging uses. An image segmentation model that is based on the particle swarm optimizer (PSO) is developed in this paper for breast cancer histopathology images of different magnification levels (40X, 100X, 200X and 400X), thus simplifying image representation and making it meaningful and easier for future analysis. As lower the magnification level, the bigger is the field of view and as it can provide greater detail, thus more time and care must be taken to use such images. Thus, finding a segmentation method that works equally well for all zoom levels of images is a big challenge. To explicate the better performance of the proposed method and its applicability on breast cancer images, results of applications and tests are augmented which shows PSO image clustering approach using intra-cluster distance as an optimization function, performs superior to cutting edge strategies, namely K-means and genetic algorithm (GA). The algorithms, when given specified number of clusters, find the centroids, thus grouping similar image primitives. The influence of different estimations of PSO control parameters on execution is additionally outlined.