Zhu, Guangran; Oreshkin, Boris; Porter, Emily; Coates, Mark; Popovic, Milica (Department of Electrical and Computer Engineering, McGill University, 3480 University St, Montreal, QC, Canada)
This paper presents the development of numerical human breast models suitable for commercial finite-difference time-domain (FDTD) simulators. The geometry of the breast models is derived from images obtained from Magnetic Resonance Imaging (MRI) scans. To avoid assigning tissue properties to every voxel, we apply regression tree analysis to partition the breast tissue region into rectangular cuboid regions (cells) that exhibit similar voxel intensity (and hence have similar tissue structure). The local spatial averaging performed by the analysis addresses the MRI-inherent noise. Further, we use the dielectric and Debye material models to simulate the dispersiveness of the breast tissue. We confirm that the Debye material model captures more accurately the higher attenuation in the high-frequency region than the dielectric material model. Experiments indicate that assuming a fixed relaxation time-constant in the Debye material model does not significantly affect the computed fields.