Computer-Aided Detection and Classification of Mass from BUS Images

Computer-Aided Detection (CAD) of breast cancer helps the radiologists in early diagnosis of breast cancer with good accuracy in a very cost-effective way. There is a growing interest for breast ultrasound (BUS) diagnosis owing to its efficiency and portability. However, the presence of speckle noise, low contrast and blurred boundary of mass in a BUS image make it challenging to determine the mass. In the current work, a CAD system is proposed for the diagnosis of breast cancer from BUS images. The methodology includes the phases of preprocessing, segmentation, feature extraction and classification of BUS images. The preprocessing algorithm used in this work efficiently removes noise and enhances the contrast of BUS images. Segmenting an accurate region of interest in turn results in efficient feature extraction and classification of BUS images into benign and malignant ones. The proposed CAD system has been tested on Matlab platform with several images to obtain reasonably good accuracy, specificity, and sensitivity. Moreover, hardware/software co-simulation of preprocessing and active contour based segmentation algorithm on Xilinx Zynq has been performed using Vivado HLS.