Constraint Split Bregman Method Combined With Total Variation For Structural Decomposition Of Breast Dce-Mri

Dynamic contrast enhanced magnetic imaging (DCE-MRI) is one of the most valuable methods for detecting breast lesions. The time intensity curve (TIC) acquired from a series of dynamic images is critical for distinguishing benign lesions and cancer. However, the TIC is often distorted by the original intensity of fine scale-details and the boundary of the normal tissue. In this paper, the constraint Split Bregman method combined with total variation (CSB-TV) is proposed to decompose the original image into texture part and structure part. A convex constraint item is introduced to ensure the consistence of texture images and convergence of algorithm. To accelerate the computational process, we employ the Split Bregman iteration algorithm. On a total of 60 breast DCE-MRI studies, the TICs obtained from the structure images acquired by CSB-TV are more consistent with three-time-point (3TP) method and pathological result. The main improvement of this algorithm is a significantly more accurate lesion classification (i.e., higher AUC) which turns out into better quality of TICs.