Testbeds for 2-D Microwave Breast Imaging Sample Clauses

Testbeds for 2-D Microwave Breast Imaging. ‌ A 2-D microwave breast imaging simulation scenario that has been used in previous work is considered for evaluation of imaging performance with different inversion approaches [131, 135]. Simulation data is produced by the FDTD method with a CPML boundary condition. The tests include all four types of numerical breast phantoms taken from the UWCEM repository [136]. In particular, 2-D axial slices representative of the phantoms classified as ‘mostly fatty’ (ID:071904), ‘scattered fibroglandular’ (ID:010204), ‘heterogeneously dense’ (ID:062204) and ‘very dense’ (ID:012304) have been considered. The single-pole Debye model is employed to describe the frequency-dependence for all breast tissues in the computation model, 1 + jωτ ϵ r (ω) = ϵ∞ + ϵ s − ϵ∞ σs j − ωϵ 0 where τ is assumed constant for all tissues (with a value of 17.125 ps). As in previous work [118, 131, 132], a lossless background medium is assumed with ϵ r = 2.6 in the simulations. However, the impact of losses is examined for some realistic background coupling media in Section 5.4. Our setup considers sixteen antennas surrounding the 2-D breast phantom, representing point sources excited with a wideband Gaussian pulse in a TM configuration (i.e. the electric field is perpendicular to the breast phantom). Six sampling frequencies are selected at 1.0, 1.5, 2.0, 2.5, 3.0 and 3.5 GHz. It is noted that the choice of the number of antennas is based on the analysis in [137]. In the 2-D scalar case, the essential number of the antennas is defined as, M = 2 βα (4.3) where α is the radius of the reconstruction domain and β is the wave number. Consider- ing our first operating frequency of 1 GHz, M is approximately equal to 15. The shape of the breast model is the only prior information assumed known for the reconstruction, while the relative dielectric permittivity of the skin and its thickness are unknown. To compare image reconstruction quality, a relative reconstruction error has been defined in (3.16) where ϵ∞ is chosen as the representative of reconstructed Debye pa- rameter, but similar metrics can be calculated for any of the parameters of the Debye (or an Ohmic) model. As the true ϵ∞ cannot be known in a realistic application, and then a ‘Residual’ error must also be defined as, t t Residual = M E − M S (4.4) where M E and M S denote complex vectors of the ‘experimental’ and ‘model’ data at the t t ttℎ iteration respectively, recorded at the antenna locations. The ‘Residual’ difference can be comp...