Automatic Prostate Cancer Detection Using Blood Flow Information in Power Doppler Ultrasonography

Abstract

Compared with the benign prostate, the malignant prostate has higher peak systolic velocity, lower end diastolic velocity and resistance index. In these blood flow information, the end diastolic velocity is more consistent and significant. In this paper, an automatic prostate cancer detection using blood flow information in power Doppler ultrasonography is proposed. The prostate region was segmented semi-automatically by the active contour model. The average velocity, resistance index and end diastolic velocity obtained from left and right peripheral zone were combined to form a feature vector. Accordingly, a support vector machine is used to classify the prostate as malignant or benign. Experimental results demonstrate the effectiveness of the proposed approach.