Chapter 14 - An automatic level set segmentation of breast tumor from mammogram images using optimized fuzzy c-means clustering

Abstract

In the medical field, segmentation plays a vital role in which the required area can be easily extracted via automatic methods. It can split the entire image into sections based on the requirement like body tissues can be segmented for tumor detection and mass detection. There are many applications, which can be made out of this segmentation method. One among them is breast tumor segmentation in which the accurate shape of the tumor and area spread can be detected for an easy diagnosis. Two difficult problems in real-world images that present a significant challenge to image segmentation are intensity inhomogeneity and noise in medical images. The most popular segmentation algorithms do not deliver accurate segmentation results because of noise and intensity inhomogeneity. In this chapter, an automatic level set method for detecting breast tumors using optimized k-means (OKM) and optimized fuzzy c-means (OFCM) is proposed. Preprocessing and postprocessing of mammogram images are the two stages that have been added to the proposed method to allow for more robust tumor segmentation. Therefore, k-means clustering and fuzzy c-means (FCM) clustering techniques are implemented with cuckoo search optimization algorithm in the image preprocessing stage; this integrating method is known as optimized k-means and OFCM clustering algorithms. Although k means and fuzzy clustering have excellent performance, they have a few drawbacks, such as randomly selected centroids and inconsistent results across executions. Consider the cuckoo search optimization algorithm used in the preprocessing step to obtain optimal pixel values, and these pixels serve as cluster centers for the conventional clustering methods. As a result, the existing methods’ limitations are overcome by employing optimized k-means and optimized fuzzy c-means.