Home Search Collections Journals About Contact us My IOPscience

Fine resolution calculations of SAR in the human body for frequencies up to 3 GHz

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2002 Phys. Med. Biol. 47 2835

(http://iopscience.iop.org/0031-9155/47/16/301)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 132.203.227.61
This content was downloaded on 21/06/2014 at 19:24

Please note that terms and conditions apply.

Fine resolution calculations of SAR in the human body

P J Dimbylow
National Radiological Protection Board, Chilton Didcot, Oxon, OX11 0RQ, UK

Received 5 April 2002

Abstract
Finite-difference time-domain (FDTD) calculations of whole-body averaged
specific energy absorption rate (SAR) have been performed from 100 MHz to
3 GHz at the basic 2 mm resolution of the voxel (volume pixel) model
NORMAN without any rescaling to larger cell sizes. The reduction in the voxel
size from previous work allows SAR to be calculated at higher frequencies.
Additionally, the calculations have been extended down to 10 MHz, covering
the whole-body resonance regions at a resolution of 4 mm. As well as for the
adult phantom, SAR values are calculated for scaled versions representing 10-,
5- and 1-year-old children for both grounded and isolated conditions. External
electric field levels are derived from limits of whole-body averaged SAR and
localized SAR in the ankle, and compared with NRPB investigation levels and
ICNIRP reference levels. The ICNIRP field reference levels alone would not
provide a conservative estimate of the localized SAR exposure in the leg for
grounded conditions. It would be necessary to invoke the secondary reference
level on limb current to provide compliance with basic restrictions on localized
SAR averaged over 10 g.

1. Introduction

The prevention of excessive heating of the body is the basis for the restriction of
electromagnetic field exposures for frequencies higher than ∼100 kHz. Restrictions are
expressed in terms of the specific energy absorption rate (SAR) in a unit of W kg−1. However,
SAR is a difficult quantity to measure and calculations must be performed to provide a
link between the internal dose quantity SAR and external fields, which can be more readily
measured or characterized. An anatomically realistic, voxel (volume pixel) model, NORMAN,
of the human body has been developed (Dimbylow 1996, 1997a). The phantom structure is
a 3D array of ∼2 mm cells or voxels, each of which has a tag denoting the discrete tissue
type or the surrounding air. There are 38 different tissues in the phantom. The phantom was
normalized to be 1.76 m tall and to have a mass of 73 kg, the new values for reference man in
0031-9155/02/162835+12$30.00 © 2002 IOP Publishing Ltd Printed in the UK 2835
2836 P J Dimbylow

ICRP 66 (1994). Hence, the name was derived from NORmalized MAN and the phantom is
meant to represent the anatomy and size of an average or reference man.
Previously, finite-difference time-domain (FDTD) (Taflove 1995) calculations had been
performed (Dimbylow 1997a) from 1 MHz to 1 GHz at rescaled resolutions because it was
not then computationally tractable to perform calculations directly at a cell size of 2 mm.
Therefore, the phantom was rescaled to produce 6 mm, 1 cm and 2 cm models with the
properties of the rescaled cells being taken as the volume average of the basic component
voxels. As the frequency increases, smaller cell sizes are required so that an adequate sampling
of the waveform can be performed. The computational effort required is proportional to the
reciprocal of the fourth power of the cell size. Computing technology has now advanced to
the point where the FDTD calculations can be performed at the basic 2 mm resolution of
NORMAN. This enables the SAR values to be performed over discrete tissue types avoiding
the smearing out of tissue properties in calculations with larger rescaled voxels. The reduction
in the voxel size also allows SAR to be calculated at higher frequencies. Of particular interest
is the whole-body irradiation from mobile phone base stations at 900 and 1800 MHz. This
paper extends calculations of the whole-body averaged SAR up to 3 GHz by using the basic
∼2 mm resolution of NORMAN. The time step is related to the resolution of the model,
which means that the number of steps required to reach equilibrium after a certain number of
periods of the wave is inversely proportional to the frequency. This means that calculations at
2 mm resolution for frequencies below 100 MHz become very time consuming. Therefore,
NORMAN was rescaled to 4 mm resolution and calculations were continued down to 10 MHz
covering the whole-body resonance regions. A comparison of whole-body average SAR is
made with the calculations from Tinniswood et al (1998) and Mason et al (2000). As well as
for the adult phantom, SAR values are calculated for scaled versions representing 10-, 5- and
1-year-old children for both grounded and isolated conditions. External electric field levels
are derived from limits of whole-body averaged SAR and localized SAR in the ankle and
compared with NRPB (McKinlay et al 1993) and ICNIRP (1998) reference levels.

2. FDTD application to NORMAN

The NRPB phantom, NORMAN, was derived from a series of continuous partial body MRI
scans of a single subject. The greyscale data were interpreted into 38 tissue types. The phantom
structure is a 3D array of cells or voxels, each of which has a tag denoting the discrete tissue
type or the surrounding air. A complete description of the acquisition of the medical imaging
data and its segmentation into tissue types can be found in Dimbylow (1996, 1997a). The
phantom was normalized to be 1.76 m tall and to have a mass of 73 kg, the new values for
reference man in ICRP 66 (1994). Hence, the name was derived from NORmalized MAN.
The height fixes the vertical voxel dimension, 2.021 mm, and the horizontal dimension, 2.077
mm, are then fixed by the mass. The total domain of the phantom and adjacent air is a 3D
array of 148 voxels from the front to back, 277 from side to side and 871 voxels high. There
are 8.3 million voxels in the body. An evaluated review of the dielectric properties of all
the tissue types in NORMAN was performed by Gabriel (Gabriel 1995, Gabriel et al 1996a,
1996b, 1996c). A 4-Cole–Cole dispersion model was fitted to the data for each tissue type to
parameterize the conductivity and permittivity as a function of frequency.
The perfectly matcher layer (pml) based boundary conditions (Berenger 1994) were used.
A Huygens surface (Merewether 1980) was implemented in the FDTD code to allow the
description of arbitrary incident fields, to separate the scattered field that is required for the
boundary conditions from the total field required for the FDTD formulation and also to connect
Fine resolution calculations of SAR in the human body for frequencies up to 3 GHz 2837

the pml layers to the inner region of the domain. The original version of NORMAN has a
vertical voxel dimension, 2.021 mm, which is slightly smaller than the horizontal dimension,
2.077 mm. This was converted to a uniform steplength of 2.077 mm by interpolation in the
vertical direction to yield 847 new planes. The domain describing the phantom comprises
277 × 148 × 847 voxels. A 2-cell layer of white space surrounds the voxel domain and is
enclosed by a 6-cell thick pml layer.
This fine resolution model provides a challenge in terms of the computational effort
required for such a large domain. One way to reduce the size of the program was not to
specify the multipliers in the time update of the magnetic field at every edge on the Yee lattice.
The multipliers are functions of the average material properties of the four surrounding cells
of an edge. As the FDTD calculations are being performed on discrete voxels and not on
rescaled composite cells, one only needs to calculate multipliers for a relatively small number
(1553) of possible combinations of tissue types. Each combination can be given an index and
this index can be stored at each edge in an integer∗2 array (in Fortran). This scheme replaces
six real∗4 multiplier arrays by three integer∗2 index arrays, a saving of ∼625 Mbytes, and
keeps the size of the program below 2 Gbytes. The calculations were run on an SGI Origin
200 server with four 225 MHz R10000 processors. The FDTD code has been modified to run
in parallel mode. The elapsed run time depends on the system load and how efficiently the
code can be partitioned amongst the four processors. When the system is empty, the speed-up
is approximately a factor of 3 when using the parallel option on the processors over just using
a single processor. Hundred iterations take about 280 min of processor time, which, using the
factor of 3 for parallel speed-up, results in a lapsed time of about 1 h and 30 min. At 900 MHz,
five periods of the wave were followed, which corresponds to 1604 time steps. This took ∼24 h.
A similar calculation at 100 MHz would take 9 days.
The FDTD implementation using a 2-cell layer of white space surrounding the voxel
domain and enclosed by a 6-cell thick pml layer was tested by comparing FDTD calculations
with the analytical Mie series solution (Neuder 1979) for a homogeneous sphere. A 30 cm
diameter sphere was constructed from 2 mm cubical cells with 150 cells across the diameter.
Figure 1 displays the comparison at 900 GHz. The values have been thinned out to be displayed
at a 4 mm separation for greater clarity. The vertically aligned (i.e. in the same direction as
the incident field) Ez-field along the central axis of the sphere from front to back is plotted
for plane-wave exposure. The complex standing wave structure excited in the sphere can be
readily seen. The homogeneous electrical properties were taken from an average of the tissue
properties in NORMAN. A quantitative comparison was sought by comparing the average Ez
along the central axis. The average absolute difference between the FDTD method and the
Mie series expansion was 2.3% at 900 MHz. Similar comparisons at 1800 and 3000 MHz
yield differences of 6.6 and 9.8%. The 4 mm version of the code was also tested at lower
frequencies and the average differences are 2.1% at 100 MHz and 2.5% at 35 MHz.

3. SAR calculations

Calculations at 2.077 mm resolution were performed for the adult NORMAN phantom from
100 MHz to 3 GHz for isolated and grounded conditions for a vertically aligned electric field,
plane-wave exposure. At lower frequencies, a 4 mm resolution discrete voxel version was
used with an overlap to 200 MHz to test the dependence of whole-body averaged SAR on
resolution. A rescaled discrete voxel model assigns a single tissue type to the new cell as the
most frequently occurring from the constituent 2.077 × 2.077 × 2.021 mm voxels instead of
assigning averaged tissue properties. Figure 2 presents the whole-body averaged SAR from
2838 P J Dimbylow

Figure 1. A comparison of the vertically aligned electric field in a 30 cm diameter sphere calculated
by the Mie series (line) with the FDTD method (symbols). The plot is along the axis of the sphere
from the front to back.

Figure 2. The whole-body averaged SAR for the adult phantom under isolated conditions. The
incident electric field is 1 V m−1 (rms). The diamonds are the 2 mm calculations, the triangles are
the 4 mm calculations and the ‘+’ signs are the previous Dimbylow (1997a) calculations from 20
to 6 mm resolution.

20 MHz to 3 GHz for isolated conditions. A comparison is made with the previous blended
voxel calculations at resolutions from 2 cm at 20 MHz to 6 mm at 1 GHz. The 2 and 4 mm
Fine resolution calculations of SAR in the human body for frequencies up to 3 GHz 2839

Figure 3. The whole-body averaged SAR for the adult phantom under grounded conditions. The
incident electric field is 1 V m−1 (rms). The diamonds are the 2 mm calculations, the triangles are
the 4 mm calculations and the ‘+’ signs are the previous Dimbylow (1997a) calculations from 20
to 6 mm resolution.

calculations merge smoothly at the cross-over region between 100 and 200 MHz. At 100 MHz,
the difference of the 4 mm result from the 2 mm value is +1.2%, and at 200 MHz, it is
+3.9%. Calculations performed at a coarser resolution agree well at higher frequencies but
are somewhat higher at resonance and below. Figure 3 presents a similar comparison for
grounded conditions. Other irradiation geometries were considered for the isolated adult
above 100 MHz. Postero-anterior, PA (from the back) irradiation produces whole-body SAR
values similar to those for AP. Irradiation from either side produces values that are about half
of those for AP at frequencies above 900 MHz but the difference diminishes with lowering
frequency and are slightly higher at 100 MHz. Tinniswood et al (1998) reported calculations
in the 2 × 2 × 3 mm resolution, University of Utah anatomical phantom. The predicted
resonance frequencies of 35 MHz for grounded and 60 MHz for isolated conditions agree well
with the present calculations. They calculated resonant SAR values of 94.8 and 74.9 µW kg−1
per V m−1 rms for grounded and isolated conditions. The corresponding values here are 93.1
and 81.2 µW kg−1 per V m−1 with differences of +2 and −8%. Mason et al (2000) calculated
whole-body SAR from 70 to 2000 MHz at 3 mm resolution in the Brooks digital anatomical
man model for isolated conditions. At 70 Hz, they predict 71.6 µW kg−1 per V m−1, while at
1800 MHz the value is 14.9 µW kg−1 per V m−1. The corresponding values here are 80.0 and
15.5 µW kg−1 per V m−1 with differences of −10 and −4%.
The adult phantom was resampled to represent the heights and masses of reference 10-, 5-
and 1-year-old children (ICRP 1994) while retaining the 2.077 mm voxel resolution at higher
frequencies and 4 mm at lower frequencies. The reference 10-year-old has a height of 1.38 m
and a mass of 33 kg, the 5-year-old has a height of 1.1 m and a mass of 20 kg and the 1-year-old
has a height of 0.75 m and a mass of 10 kg. Table 1 presents the whole-body averaged SAR
for the adult, 10-, 5- and 1-year-old phantoms at a 2.077 mm resolution for both grounded and
isolated conditions. The incident field has the electric field component vertically aligned and
2840 P J Dimbylow

Table 1. Whole-body averaged SAR values, µW kg−1 for the 2 mm resolution calculations. The
incident electric field is 1 V m−1 (rms).

Grounded Isolated
Frequency
(MHz) Adult 10 y 5y 1y Adult 10 y 5y 1y
100 32.2 65.4 87.0 116 40.5 92.5 101 54.1
200 20.4 31.8 43.4 80.1
400 16.7 24.7 30.3 39.3 16.7 23.6 30.7 41.3
600 16.6 22.6 27.0 32.9
900 16.5 22.1 25.1 30.7 17.0 22.7 25.8 30.4
1400 16.8 22.8 25.9 30.1
1800 15.2 21.6 24.3 28.1 15.5 22.0 24.6 29.0
2450 12.6 19.4 21.2 25.0
3000 10.4 16.6 18.1 21.7 10.6 17.0 18.6 22.1

Table 2. Whole-body averaged SAR values, µW kg−1 for the 4 mm resolution calculations. The
incident electric field is 1 V m−1 (rms).

Grounded Isolated
Frequency
(MHz) Adult 10 y 5y 1y Adult 10 y 5y 1y
10 11.9 11.4 9.2 5.2
20 48.2 40.8 32.9 19.0 5.5 4.9 3.8 2.3
30 82.9 74.7 59.5 35.6 14.4
35 93.1
40 92.3 98.7 83.3 55.5 30.3 23.4 17.0 9.8
45 107
50 77.7 109 104 72.1 52.9 39.6 26.8
55 108 112
60 61.4 104 115 87.8 76.5 59.3 40.1 20.0
65 115 81.2
70 49.8 93.2 113 105 80.0 82.3 55.6
80 43.1 80.1 108 115 66.7 99.5 73.5 35.8
85 103
90 69.9 97.7 117 51.7 102 92.0
100 31.2 62.7 86.5 117 41.0 92.5 105 55.6
110 114 108 67.5
120 28.5 47.6 69.9 107 30.5 64.8 103 80.4
130 98.5 92.0 91.7
140 89.3 80.2 100
150 23.5 38.8 50.9 81.1 24.3 42.2 69.4 105
160 74.2 60.8 105
170 101
180 63.6 49.3 95.0
190 87.5
200 21.6 31.1 41.3 55.3 21.2 32.0 43.2 79.9
220 39.3 67.2
250 36.4 55.4
300 34.4 46.5

is incident from front to back (AP). Table 2 presents the SAR values at 4 mm resolution at
sufficient frequencies to cover the whole-body resonance. Figures 4 and 5 plot the envelopes
of the higher SAR values from the 2- and 4-mm calculations over the range of ages for isolated
Fine resolution calculations of SAR in the human body for frequencies up to 3 GHz 2841

Figure 4. The whole-body averaged SAR for the adult, 10-, 5- and 1-year-old phantoms under
isolated conditions. The incident electric field is 1 V m−1 (rms).

Figure 5. The whole-body averaged SAR for the adult, 10-, 5- and 1-year-old phantoms under
grounded conditions. The incident electric field is 1 V m−1 (rms).

and grounded conditions, respectively. The results in figure 4 are similar to those given for
isolated conditions in the RF Dosimetry Handbook (Durney et al 1986). The whole-body
resonance occurs at ∼65, 85, 110 and 155 MHz for the adult, 10-, 5- and 1-year-old phantoms
under isolated conditions. The resonance occurs when the height of the body is approximately
2842 P J Dimbylow

0.38–0.4λ, where λ is the wavelength in air. This is comparable to the findings reported in
Gandhi (1980) where the resonant frequency is defined as 0.36–0.4λ. When the phantom is
grounded, the reflection in the ground plane halves the resonant frequency. However, the body
is not a thin dipole and the exact resonant frequencies depend on the anatomy and values of
the frequency-dependent dielectric properties. Under grounded conditions, the whole-body
resonance occurs at ∼35, 50, 65 and 95 MHz for the adult, 10-, 5- and 1-year-old phantoms.
The linear scaling of the adult phantom does not exactly reproduce the anatomy of children
because amongst other things the ratio of the head length to the body decreases as the child
matures. However, the most important factors determining the frequency and magnitude
of resonance are the height and cross section of the phantoms representing different ages
as well as dielectric properties. Hence, the adult model was rescaled using the values of
height and mass of the reference 10-, 5- and 1-year-old children from ICRP Reference Man.
Some information on age-dependent dielectric properties is becoming available. Peyman et al
(2001) measured the dielectric properties of ten rat tissues at six different ages from 130 MHz to
10 GHz. The results show a general decrease of the dielectric properties with age, the trend
being more apparent for brain, skull and skin tissues and less noticeable for abdominal tissues.
The conductivity and permittivity values for a 10 day rat compared to a 70 day rat range from
∼20 to 100% higher. As an indicator of how this may affect the whole-body averaged SAR,
calculations were performed on the 1-year-old isolated phantom at a selection of frequencies
for all the dielectric properties of the body increased by 50 and 100%. In general, the increase
in dielectric property values results in a decrease in the whole-body averaged SAR except for a
region between 110 and 180 MHz. At 1800 MHz, the SAR is 16 and 29% lower for an increase
of property values of 50 and 100%, respectively. This effect diminishes with decreasing
frequency. At 200 MHz, the differences are −1 and −4%, respectively. At 150 MHz,
the resonant frequency, the increase in property values increases the SAR by +9 and +13%.
At present, the scaled phantoms produce the best available information. A definitive answer
will require the development of actual child phantoms coupled with an evaluated set of age-
dependent dielectric properties over the full range of frequencies for all the tissues of the
human body.
Figures 6 and 7 plot the electric field values that would be required to produce the limiting
condition from either whole-body averaged or partial-body SAR, in the head and torso, and
a comparison is made with the NRPB and ICNIRP occupational reference levels. Similarly,
figure 8 plots the electric field values for a comparison with the ICNIRP public exposure
reference levels. In general, the restriction on whole-body averaged SAR (0.4 W kg−1
for NRPB and ICNIRP occupational, 0.08 W kg−1 for ICNIRP public exposure) is the
critical condition requiring the lowest external field value. However, for the adult the limit of
10 W kg−1 in 10 g of the torso for the ICNIRP occupational exposure provides the critical
condition at 2450 and 3000 MHz whereas the limit of 10 W kg−1 in 10 g of the head for
the NRPB guidelines provides the critical condition at the same frequencies. The averaging
masses of 10 and 100 g for localized SAR represent a greater portion of the body for the
smaller phantoms and so for them the whole-body averaged SAR tends to be the critical
condition. The other region where the whole-body averaged SAR may not be the most
important condition is in the narrow part of the leg and ankle for the grounded adult phantom
near resonance. Table 3 combines ankle currents calculated in this work for the 4 mm
adult model with values of SAR versus current (Dimbylow 1997b) injected into the full
resolution model of the lower leg to predict localized SAR in the leg for 2 mm resolution.
The average over 10 g in the limbs, as used by ICNIRP, results in an electric field level of
46 V m−1 for a 20 W kg−1 restriction at 35 MHz. The average over 100 g, as used by the NRPB
guidelines, produces a corresponding field level of 60.4 V m−1. The electric field values that
Fine resolution calculations of SAR in the human body for frequencies up to 3 GHz 2843

Figure 6. Electric field values required to produce the NRPB restrictions on SAR. The values
derived from the restriction of localized SAR in the leg are shown separately (long dashes). The
thick dashed investigation level applies when there is no possibility of small children being exposed.

Figure 7. Electric field values required to produce the ICNIRP occupational restrictions on SAR.
The values derived from the restriction of localized SAR in the leg are shown separately (long
dashes).

would be required to produce SAR values of 10 W kg−1 in the leg averaged over 100 g for
the NRPB guidelines and over 10 g for the ICNIRP guidelines are also plotted separately in
figures 6–8.
2844 P J Dimbylow

Figure 8. Electric field values required to produce the ICNIRP public exposure restrictions on
SAR. The values derived from the restriction of localized SAR in the leg are shown separately
(long dashes).

Table 3. The limiting electric fields, E for SARs averaged over 10 and 100 g in the leg for the
grounded adult phantom. The ankle current is calculated from the 4 mm version of NORMAN.
The SAR values per unit injected current are taken from Dimbylow (1997b) for a 2 mm model of
the lower leg. These two quantities are combined to give an SAR value for a unit applied field
(Note: SAR is proportional to the square of the current). The limiting electric field can then be
found for a restriction of 20 W kg−1. The currents and electric field values are rms quantities.

Ankle
current SAR (10 g) SAR (10 g) SAR (100 g) SAR(100 g)
Frequency (mA per (W kg−1 (mW kg−1 E (10 g) (W kg−1 (W kg−1 E (100 g)
(MHz) V m−1) per A) per V m−1) (V m−1) per A) per V m−1) (V m−1)
10 1.17 973 1.34 122 555 0.766 162
20 2.47 844 5.16 62.3 488 2.98 81.9
30 3.27 791 8.46 48.6 460 4.92 63.8
35 3.51 767 9.43 46.1 446 5.48 60.4
40 3.53 720 8.96 47.2 422 5.25 61.7
50 3.27 677 7.23 52.6 402 4.29 68.3
60 2.94 607 5.26 61.7 364 3.15 79.7
70 2.65 577 4.06 70.2 347 2.44 90.5
80 2.42 531 3.10 80.3 327 1.91 102

Over the resonance region, the NRPB levels provide a conservative estimate of the
calculated SAR values. However, above 1 GHz they underestimate exposure to SAR for all
four phantoms, from the adult to the 1-year-old. A more appropriate value in the region above
1 GHz may be 110 V m−1. The ICNIRP occupational level is conservative at these higher
frequencies above 1 GHz but the public level for 1-year-old children is not. The ICNIRP
localized average over 10 g rather than 100 g in the lower leg results in field levels that are
Fine resolution calculations of SAR in the human body for frequencies up to 3 GHz 2845

not conservative. Levels of 46 and 21 V m−1 would be more appropriate for the occupational
and public exposure, respectively. However, ICNIRP has a secondary reference level on limb
current of 100 mA, and 45 mA for occupational and public exposure between 10 and 100 MHz.
The corresponding external field value for a limb current of 100 mA at 35 MHz (from table 3)
would be 28 V m−1 and so the limb current reference levels would provide compliance with
basic restrictions on localized SAR averaged over 10 g in the leg.

4. Conclusions

FDTD calculations have been performed at the basic 2 mm resolution of the voxel model,
NORMAN, without any rescaling to larger cell sizes. The reduction in the voxel size allows
SAR to be calculated at higher frequencies. This paper extends calculations of the whole-body
averaged SAR up to 3 GHz from 100 MHz by using the basic ∼2 mm resolution of NORMAN.
Calculations at 2 mm resolution for frequencies below 100 MHz become very time
consuming because the period of the wave is inversely proportional to its frequency. NORMAN
was rescaled to 4 mm resolution and calculations were continued down to 10 MHz covering
the whole-body resonance regions. SAR values are calculated for scaled versions representing
10-, 5- and 1-year-old children for both grounded and isolated conditions as well as for the
adult phantom. Comparison of the 2 and 4 mm results in the overlap frequency range of the two
sets of calculations shows that the whole-body averaged SAR is a robust quantity with respect
to model resolution. The SAR values were converted to external fields that would produce a
whole-body average of 0.4 W kg−1 for NRPB guidelines and ICNIRP occupational guidelines,
and 0.08 W kg−1 for ICNIRP public-exposure guidelines. Comparison with reference levels
reveals that they may not be sufficiently conservative in the region approaching 3 GHz.
Ankle currents were also calculated in the adult resonance region and combined with
values of SAR versus current (Dimbylow 1997b) injected into the full resolution, 2 mm
NORMAN model of the lower leg to calculate localized SAR in the leg for 2 mm resolution.
Comparison of derived external electric fields shows that the ICNIRP field reference levels
alone would not provide a conservative estimate of the localized SAR exposure. It would be
necessary to invoke the secondary reference level on limb current to provide compliance with
basic restrictions on localized SAR averaged over 10 g in the leg.

Acknowledgment

I would like to thank Dr Ian Scivill of the NRPB for performing the Mie series calculations.

© 2002 NRPB

References

Berenger J 1994 A perfectly matcher layer for the absorption of electromagnetic waves J. Comp. Phys. 114 185–200
Dimbylow P J 1996 The development of realistic voxel phantoms for electromagnetic field dosimetry Proc. Int.
Workshop on Voxel Phantom Development (National Radiological Protection Board Report) pp 1–7
Dimbylow P J 1997a FDTD calculations of the whole-body averaged SAR in an anatomically realistic voxel model
of the human body from 1 MHz to 1 GHz Phys. Med. Biol. 42 479–90
Dimbylow P J 1997b The calculation of localised SAR in a 2 mm resolution anatomically realistic model of the lower
leg Radiat. Prot. Dosim. 72 321–6
Durney C H, Massoudi H and Iskander M F 1986 Radio Frequency Radiation Dosimetry Handbook 4th edn
USAFSAM-TR-85–73 (Brooks Air Force Base, TX: USAF School of Aerospace Medicine)
2846 P J Dimbylow

Gabriel C 1995 Compilation of the dielectric properties of body tissues at RF and microwave frequencies Report
prepared for the NRPB by Microwave Consultants Ltd
Gabriel C, Gabriel S and Corthout E 1996a The dielectric properties of biological tissues: 1. Literature survey Phys.
Med. Biol. 41 2231–49
Gabriel S, Lau R W and Gabriel C 1996b The dielectric properties of biological tissues: 2. Measurements in the
frequency range 10 Hz to 20 GHz Phys. Med. Biol. 41 2251–69
Gabriel S, Lau R W and Gabriel C 1996c The dielectric properties of biological tissues: 3. Parametric models for the
dielectric spectrum of tissues Phys. Med. Biol. 41 2271–93
Gandhi O P 1980 State of the knowledge for electromagnetic absorbed dose in man and animals Proc. IEEE 68 24–32
ICNIRP 1998 Guidelines for limiting exposure to time-varying electric, magnetic and electromagnetic fields (up to
300 GHz) Health Phys. 74 494–522
ICRP 1994 Human Respiratory Tract Model For Radiological Protection ICRP Publication 66 (table B.6) p 189
Mason P A, Hurt W D, Walters T J, D’Andrea J W, Gajsek P, Ryan K L, Nelson D A, Smith K I and Ziriax J M 2000
Effects of frequency, permittivity and voxel size on predicted specific absorption rate values in biological tissue
during electromagnetic-field exposure IEEE Trans. Microw. Theory Tech. 48 2050–7
McKinlay A F, Allen S G, Dimbylow P J, Muirhead C R and Saunders R D 1993 Restrictions on human exposure to
static and time varying electromagnetic fields and radiation Document of the National Radiological Protection
Board 4 (5)
Merewether D E, Fisher R and Smith F W 1980 On implementing a numeric Huygens source scheme in a finite
differences program to illuminate scattering bodies IEEE Trans. Nucl. Sci. 27 1829–33
Neuder S M 1979 Electromagnetic fields in biological media: Part 2. The SCAT program, multilayered spheres,
theory and applications HEW Publication (FDA) 79–8072
Peyman A, Rezazadeh A A and Gabriel C 2001 Changes in the dielectric properties of rat tissue as a function of age
at microwave frequencies Phys. Med. Biol. 46 1617–29
Taflove A 1995 Computational Electromagnetics—The Finite-Difference Time-Domain Method (London: Artech)
Tinniswood A D, Furse C M and Gandhi O P 1998 Power deposition in the head and neck of an anatomically based
human body model for plane wave exposure Phys. Med. Biol. 43 2361–78