Mapping of Breast Tissue Dielectric Properties Using T1-Weighted MRI Data

This study introduces a method for developing realistic numerical two-dimensional (2D) microwave breast models from T1-weighted MRI data, integrating anatomical features, such as: shape, size, and density. This method contributes a new perspective in enhancing noise reduction, tissue segmentation, nonlinear mapping of electromagnetic attributes, and the authentic depiction of breast contours for precise tissue categorization. It follows a comprehensive four-phase procedure: denoising MRI data by estimating and mitigating bias field in each slice, after which the breast tissue is segmented in two main categories: adipose (fat) and fibroglandular. Furthermore, using the distribution of MRI pixel intensities, a nonlinear mapping functions are generated, which allow each MRI pixel to map corresponding electromagnetic properties. The efficiency of proposed method is assessed through Dice-Sorensen Coefficient statistic, juxtaposing the categorized images against reference standards set by radiologists. The method achieved an average Dice Similarity Coefficient of 74.66 % for class A, 85.12% for class B, 78.77% for class C, and 85.56% for class D breast densities.