A Comparison Study Between Otsu’s Thresholding, Fuzzy C-Means, and K-Means for Breast Tumor Segmentation in Mammograms

Abstract

Currently, the technique of digital mammography is considered the first screening tool used for detecting breast cancer in early stages. The shape and margin of masses in mammograms play a crucial role in the classification of benign and malignant tumors. Several machine learning (ML) algorithms have demonstrated high performance in breast mass edge segmentation from mammograms. The main objective of this paper is to present a comparative study of three algorithms of ML where: C-means clustering, K-means clustering, and Otsu’s thresholding for automatically detecting the lesion edges on mammogram images. Two groups (C = 2) are assigned to clustering as output for each method. In order to achieve high accuracy in contour extraction, only relevant pixels are selected as inputs. This study will highlight the advantages and disadvantages of these algorithms, which are expected to improve the detection of breast masses contours in mammographic images. The mini-MIAS database is used to evaluate the obtained results which show a higher performance in terms of segmentation accuracy when using fuzzy C-means algorithm. However, a faster lesion extraction process can be achieved by utilizing the K-means clustering and Otsu’s threshold.