Texture-based simultaneous registration and segmentation of breast DCE-MRI

We present a registration method for breast dynamic contrast-enhanced (DCE) MRI data based on texture information. The algorithm combines feature and spatial information to propose an image segmentation based on a Hidden Markov Random Measure Field (HMRMF) model using expectation-maximisation (EM) iteration. It can be sued to simultaneously estimate parameters in order to segment and register the images iteratively. Global motions are modeled by an affine transformation, while local breast motions are described using free-form deformations (FFD) based on B-splines. Experimental results on real DCE-MRI data are presented to demonstrate the performance of the algorithm.