An Artificial Neural Network for the Segmentation of Dynamic MR Mammographic Image Series

Introduction The large number of images acquired in a dynamic MR breast examination demand a comprehensive presentation of the kinetics of contrast enhancement and lesion architecture. Data reduction can be obtained by 1) pharmacokinetic analysis combined with color-coded presentation of the pharmacokinetic parameters, or 2) a pixel-by-pixel characterization of breast tissue based on neural networks. Both approaches will be compared.