Boundary extraction of breast ultrasonic images.

The accurate boundary extraction is an essential preprocessing step for computerized analysis of a breast ultrasonic image. In this study, a novel approach based on the wavelet transform and the dynamic programming is proposed to extract tumor boundaries from breast ultrasonic images. Firstly, a rectangular region-of-interest (ROI) is manually selected from the ultrasonic image, followed by the ROI-based preprocessing for the noise reduction and image enhancement. Then an initial boundary of the tumor is obtained using the wavelet transform. To achieve a more accurate result, a dynamic programming algorithm based on the active contour model is applied to adjust the initial boundary. Experiments on 45 breast ultrasonic images have shown that this proposed method is a promising technique for the boundary extraction of breast tumor on ultrasonic images.