Incorporating texture features in a computer-aided breast lesion diagnosis system for automated three-dimensional breast ultrasound

• DOI: 10.1117/1.jmi.1.2.024501
• Publication Date: 2014-07-25T12:38:27Z
• Authors: Haixia Liu; Tao Tan; Jan van Zelst; Ritse Mann; Nico Karssemeijer; Bram Platel
• Keywords: Breast ultrasound; Breast imaging; Texture (cosmology); Computer-Aided Diagnosis; Breast tumor

We investigated the benefits of incorporating texture features into an existing computer-aided diagnosis (CAD) system for classifying benign and malignant lesions in automated three-dimensional breast ultrasound images. The takes account 11 different features, describing lesion properties; however, it does not include features. In this work, we expand by including based on local binary patterns, gray level co-occurrence matrices, Gabor filters computed from each to be diagnosed. To deal with resulting large number proposed a combination feature-oriented classifiers combining group single likelihood, three additional used final classification. classification was performed using support vector machine classifiers, evaluation done 10-fold cross validation dataset containing 424 (239 185 lesions). compared performance CAD without area under receiver operating characteristic curve increased 0.90 0.91 after adding (p<0.001).