New Technique for Processing Tissue Displacements Estimation in Breast Ultrasound Elastography

Abstract

Breast ultrasound elastography allows acquiring a displacements estimation images that reflects the tissue elasticity of the area under ultrasound examination. The displacements analysis of the mammary tissues helps to provide an accurate diagnosis related to tissue stiffness evaluation. However; despite the fact that this imaging modality is promising and effective, it encounters a major problem, B-mode images are altered by speckle noise which comes from the difference in echogenicity of the imaged tissues, which deteriorates the diagnostic information, therefore, an application of displacements estimation methods on contaminated B-mode images, amplifies indirectly the speckle noise, causing a reduction of clinical evaluation capacity and prevention of rigorous clinical assessment. In this perspective, the development of tools for treating this problem is crucial and indispensable for controlling the appearance of speckle noise. In this paper, we propose a new strategy based on the bilinear deformable block matching (BDBM) technique, in order to improve the old bilinear deformable block matching (OBDBM) technique by improving the B-mode images pre-processing step, at this stage, the B-mode images processing before estimating displacements is sufficient to have a good estimate. The proposed approach does not only filter the image against noise, it also improves the estimation of tissue deformations, the developed approach is based on the coupling between the reconstruction by filtered back projection (FBP) technique with shrinkage wavelets. The proposed model was validated using in vivo database containing 20 B-mode images acquired before and after compression. The FBP approach was used to reconstruct the tissue texture, consequently improves the resolution and the contours erased by the noise, after that the shrinkage wavelets have been introduced to completely filter the image against speckle noise. By comparing our proposed method with OBDBM and monogenic techniques (displacement estimation method based monogenic technique), we show that it is the most efficient in term of standard deviation calculation between the pixels (SD), it is better also for calculating the contrast to noise ratio (CNR) and it provides an excellent structural similarity (SSIM). The results of the proposed technique are encouraging and it is ready to be used in mammary tissues displacements enhancement images for ultrasound elastography.