Strain Compounding: Improvement in Contour Extraction of Ultrasonic Breast Imaging

Abstract

Computer-aided diagnosis for breast diseases has gained wide interest. In ultrasound, the automatic diagnosis typically requires knowledge of the shape of a lesion, thus making contour extraction a critical step. Unfortunately, the speckle noise inherent in an ultrasonic image not only lowers the contrast resolution but also limits the accuracy of contour extraction. Thus, speckle noise reduction is desired prior to contour extraction. Contour extraction with a speckle noise reduction technique called strain compounding is presented in this paper. Strain compounding reduces speckle noise by incoherently combing partially correlated measurements of the same image object. Efficacy of speckle reduction using strain compounding has been demonstrated on both simulated and clinical images under different strain conditions. It was shown that speckle noise can be effectively reduced without significant degradation in spatial resolution. In this paper, we further demonstrate its ability to improve accuracy of contour extraction. Simulations and clinical images are both included and the improvement is also quantified and discussed.