Automatic Breast Tumor Classification in Ultrasound Images Using Morphological Features and New Texture Analysis Based on Image Visibility Graph and Gabor Filters

Lesion segmentation is a challenging task in computer-aided diagnosis (CAD) systems. In this paper, an automatic segmentation and diagnosis algorithm is proposed for Breast Ultrasound (BUS) images. Among imaging methods, ultrasound is used as an appropriate tool in the diagnosis of breast cancer owing to its advantages, including real-timeliness, low cost, no use of ionizing radiation, and high sensitivity in dense tissues. For this task, the main focus is to provide an efficient and automatic method to segment the region of interest (ROI), as well as to use morphological and texture-based features for diagnostic purposes. Two texture-based feature extraction methods, i.e. “estimation of Gabor filter coefficients by an autoregressive model” and “using statistical features in image visibility graph”, have been introduced after the automatic development of ROI. These features together with morphological features are classified by the Support Vector Machine (SVM) classifier to identify lesion type. After the selection of superior features by the recursive feature elimination algorithm, the proposed method is tested on a database with 163 images; the obtained results confirm the image segmentation ability and the feature separation ability.