Quasistatic zooming of FDTD (E)over-right-arrow-field computations: the impact of down-scaling techniques

Due to current computer limitations, regional hyperthermia treatment planni ng (HTP) is practically limited to a resolution of 1 cm, whereas a millimet re resolution is desired. Using the centimetre resolution (E) over right ar row -field distribution, computed with, for example, the finite-difference time-domain (FDTD) method and the millimetre resolution patient anatomy it is possible to obtain a millimetre resolution SAR distribution in a volume of interest (VOI) by means of quasistatic zooming. To compute the required low-resolution (E) over right arrow -field distribution, a low-resolution d ielectric geometry is needed which is constructed by down-scaling the milli metre resolution dielectric geometry. In this study we have investigated wh ich down-scaling technique results in a dielectric geometry that yields the best low-resolution E-field distribution as input for quasistatic zooming. A segmented 2 mm resolution CT data set of a patient has been down-scaled to 1 cm resolution using three different techniques: 'winner-takes-all', 'v olumetric averaging' and 'anisotropic volumetric averaging'. The E-field di stributions computed for those low-resolution dielectric geometries have be en used as input for quasistatic zooming. The resulting zoomed-resolution S AR distributions were compared with a reference: the 2 mm resolution SAR di stribution computed with the FDTD method. The (E) over right arrow -field d istribution for both a simple phantom and the complex partial patient geome try down-scaled using 'anisotropic volumetric averaging' resulted in zoomed -resolution SAR distributions that best approximate the corresponding high- resolution SAR distribution (correlation 97, 96% and absolute averaged diff erence 6, 14% respectively).