A Methodology of Segmentation : Defining Adipose and Non-Adipose Tissue Volumes in Female Breast MR Images

Understanding the anatomical and physiological factors influencing lactation is important. This study focuses on breast segmentation to receive ratios of the adipose and non-adipose tissue in the mammary breast tissue. This study uses T2-Dixon Fat and Water scans acquired with a 1.5T MRI scanner from four subjects, one of whom is lactating. The created method to segment adipose and non-adipose tissue in the mammary breast starts with setting ellipses and points on five dispersed slices to remove non-mammary breast tissue. The breast’s boundary is defined ventrally to the pectoralis major muscle, with lateral fat tissue delineated using established landmarks. The intermediate slices are interpolated using a ’spline’ function in MATLAB. The MATLAB Toolbox ’Medical Image Labeler’ and a custom automated algorithm further distinguish adipose and non-adipose tissues. Segmentation results in a maximum variance of 6.69% in adipose tissue and 1.51% in non-adipose tissue, showing methodological consistency. The algorithms can be improved to prevent double-labeled tissue and not-labeled tissue within mammary breast tissue. This method yields realistic volume calculations for adipose and non-adipose tissues together, validated against a comparative segmentation method using the same dataset. Variances range from 0.15% to 14.4%, underscoring the method’s reliability and potential for advancing breast tissue analysis.