3-D ultrasound strain images for breast cancer diagnosis

Strain imaging with ultrasound is a recent technique to provide additional information for tissue characterization by imaging differences in tissue stiffness. The conventional 2-D strain method is applied for only two US images with and without compression. In this paper, the strain imaging technique is extended for the 3-D ultrasound dataset for classify breast masses. The proposed strain image analysis is based on quantifying the turnout's shape change after compression. A level set segmentation method is used to extract the lesion contours from the 3-D US images. Some strain features are proposed to compare the lesion contours of the pre- and post-compressed 3-D US images. A support vector machine (SVM) is also used to classify lesions. According to the experimental results, the 3-D stain method is better than the 2-D strain method. (c) 2005 CARS & Elsevier B.V. All rights reserved.