Fully automated computer aided diagnosis system for classification of breast mass from ultrasound images
Poulami Raha, Radhika V. Menon, Indrajit Chakrabarti · 2017
This paper presents a fully automated computer aided diagnosis (CAD) system for breast cancer detection and classification from breast ultrasound (BUS) images. Detection of breast mass from ultrasound images is inherently difficult due to the presence of speckle noise. The preprocessing algorithm used here efficiently removes noise and enhances the contrast of BUS images. Region of interest (ROI) is accurately detected by marker based watershed algorithm without manual selection of ROI in 13.29 s on Intel(R) i7 - 4770SCPU (3.10 GHz) and 8 GB RAM. Classification of BUS images is achieved using k-NN classifier based on extracted feature set. The proposed fully automated CAD system has acquired an accuracy of 96.4%.