Textural Feature Analysis for Ultrasound Breast Tumor Images

Abstract

Texture is one of the important characteristics used in identifying objects or regions of interest in an image. This paper describes some textural features based on integrated spatial gray level co-occurrence matrix, and illustrates the effectiveness of four textural features in categorizing ultrasound breast tumor images by means of Fuzzy C-means and K-medoid clustering algorithms respectively. The experimental identification accuracy is 72.6415 percent. These results indicate that textural features probably have a general applicability for classification of breast tumors.