Quantification of breast tissue index from MR data using fuzzy clustering

The study objective was to develop a segmentation technique to quantify breast tissue and total breast volume from magnetic resonance imaging (MRI) data to obtain a breast tissue index (BTI) related to breast density. Our goal is to quantify MR breast density to improve breast cancer risk assessment for certain high-risk populations for whom mammography is of limited usefulness due to high breast density. A semi-automatic 3D segmentation technique was implemented based on a fuzzy c-means technique (FCM) to segment fibroglandular tissue from fat in the breast images. After validation on a phantom, our FCM technique was first used to test the breast tissue measures reproducibility in two consecutive MR examinations of the same patients. The technique was then applied to measure the BTI on 10 high-risk patients. Results of BTI obtained with the semi-automated FCM method were compared with BTI results for the same patients using two other techniques, manual delineation and global threshold. BTI measures correlated well with mammographic densities (Pearson coefficients r = 0.78 using MR manual delineation, and r = 0.75 using MR FCM). The breast tissue index could therefore become a common measure for future studies of using noncontrast MRI data.