Background: Classification of breast ultrasound (BUS) images is an important step in\nthe computer-aided diagnosis (CAD) system for breast cancer. In this paper, a novel\nphase-based texture descriptor is proposed for efficient and robust classifiers to\ndiscriminate benign and malignant tumors in BUS images.\nMethod: The proposed descriptor, namely the phased congruency-based binary\npattern (PCBP) is an oriented local texture descriptor that combines the phase\ncongruency (PC) approach with the local binary pattern (LBP). The support vector\nmachine (SVM) is further applied for the tumor classification. To verify the efficiency\nof the proposed PCBP texture descriptor, we compare the PCBP with other three\nstate-of-art texture descriptors, and experiments are carried out on a BUS image\ndatabase including 138 cases. The receiver operating characteristic (ROC) analysis\nis firstly performed and seven criteria are utilized to evaluate the classification\nperformance using different texture descriptors. Then, in order to verify the\nrobustness of the PCBP against illumination variations, we train the SVM\nclassifier on texture features obtained from the original BUS images, and\nuse this classifier to deal with the texture features extracted from BUS images\nwith different illumination conditions (i.e., contrast-improved, gamma-corrected\nand histogram-equalized). The area under ROC curve (AUC) index is used as the\nfigure of merit to evaluate the classification performances.\nResults and conclusions: The proposed PCBP texture descriptor achieves the\nhighest values (i.e. 0.894) and the least variations in respect of the AUC index,\nregardless of the gray-scale variations. It�s revealed in the experimental results\nthat classifications of BUS images with the proposed PCBP texture descriptor are\nefficient and robust, which may be potentially useful for breast ultrasound CADs.
Loading....