Swarm optimization and multi-level thresholding of cytological images for breast cancer diagnosis

Abstract

This paper presents a novel approach for multi-level thresholding of cytologic images. Typically, thresholding is applied in order to segment the image into regions of interest or objects, each having a high level of homogeneity in some parameter such as luminance. Homogeneous regions are then used to generate a set of features discriminating categories occurring in a given diagnostic problem. Instead of homogeneity measure, our approach uses a classifier to evaluate the quality of segmentation solution directly. The candidate solutions (sets of threshold values) are generated with use of the stochastic swarm intelligence-based metaheuristics. Experimental results demonstrate the promising performance of the proposed classification-driven segmentation in application to breast cancer diagnostics. © Springer International Publishing Switzerland 2013.