888 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. MIM’-23, NO.

11, NOVEMBER 1975

Computation of the Electromagnetic Fields and Induced

Temperatures within a Model of the
Microwave-h-radiated Human Eye
ALLEN TAFLOVE AND MORRIS E. BRODWIN, SENIOR MEMBER, IEEE

Absfract—The electromagnetic fields within a detailed model of likely thermal [5]. A standard for human exposure, based
the human eye and its surrounding bony orbit are calculated for upon such animal experimentation, has been published [6].
two different frequencies of plane-wave irradiation: 75o MHz and
The use of animal experimentation to establish a human
1.5 GHz. The computation is performed with a finite-dtierence
algorithm for the time-dependent Maxwell’s equations, carried out exposure standard for microwave radiation implies that
to the sinusoidal steady state. The heating potential, derived from the anatomy, physiology, and electromagnetic environ-
the square of the electric field, is used to calculate the temperatures ment of the test animals can be related to that of humans.
induced within the eyeball of the model. This computation is per- However, several elements of this relation remain unclear.
formed with the implicit alternating-direction (IAD) algorithm for the
In particular, the role of tissue structure in determining
heat conduction equation. Using an order-of-magnitude es@nate of
the heat-sinking capacity of the retinal blood supply, it is determined microwave absorption may be significant. We know that
that a hot spot exceeding 40.4°C occurs at the center of the model electromagnetic wave absorption in a lossy dielectric
eyeball at an incident power level of 100 mW/cmz at 1.5 GHz. scatterer is a function of its shape and dimensions. It is
quite possible that the eye–scatterer of man develops
heating potential contours different in location and mag-
I. INTRODUCTION
nitude from those of the rabbit because of the dimen-
sional and structural differences in tissue anatomy. This
A T PRESENT,
in solving for
little theoretical
the temperature
work has been done
distribution induced
possibility
standard
must be explored if a more precise exposure
for man is to be established. As Guy [7] has
by microwave radiation in complicated biological struc-
tures. The emphasis has been on experimental investiga- stated: “A high priority need in this area is a complete

tion. This has been brought about in part by the expensive thermodynamic study of the eye under microwave ex-
and time-consuming numerical methods required to com- posure.”
pute the electromagnetic fields within arbitrary dielectric Direct experimentation with the living human eye using

scatterers. Indeed, inhomogeneous tissues of great comp- either cataractogenic exposures stillar to those of [3],
lexity may require so much direct computer storage with [4], or usipg implanted thermocouple techniques [8] is

well-known techniques that solution is virtually impos- impermissible because of the tissue damage caused by the
sible. experimental procedure. Therefore, the microwave heat-
The eye has been of special experimental interest be- ing of the human eye must be studied using models of the
cause of evidence of microwave-induced cataracts in actual organ. A theoretical approach would attempt to
humans [1], [2]. Typical experiments involved the ex- solve for the fields and temperatures using some analytical
posure of rabbits to high levels of microwave radiation or numerical method. Such a model would postulate a near-
over a short time interval and the observation of the in- er far-field irradiation, and simulate the tissue geometry
duced lens opacification over a period of several weeks and thermodynamics to the maximum possible extent.
[3], [4]. The relevance of such studies was based upon Tb.is is the approach taken in this paper.
the similarity of the anatomy of human and rabbit eyes. Early theoretical work in the area of the biological
These experiments established time-power density thres- effects of electromagnetic radiation centered on the ir-
hold levels for cataract formation. Results indicated that radiation of models of the entire human body [9]. How-
the dose of microwave radiation required for lens injury ever, because experimental work indicated that harmful
is based upon average rather than peak power. The mech- local tissue temperature rises could occur, interest in
anism of microwave cataract formation is therefore most partial body irradiation was stimulated. Cook proposed
the solution of Maxwell’s equations coupled with the heat
conduction equation to solve the problem of local micro-
Manuscript received March 21, 1975; revised May 20, 1975. wave heating [10]. He developed a theory for the heating
A. Taflove was with the Department of Electrical Engineering,
of a tissue half-space composed of layers of skin, fat, and
Technological Institute, Northwestern University, Evanston, Ill.
60201. He is now with the Research Institute, Illinois Institute of muscle, by incident plane waves. The temperature dis-
Technology, Chicago, Ill. 60616.
tribution predicted by the theory was verified in his re-
M. E. Brodwin is with the Department of Electrical Engineering,
Northwestern University, Evanston, Ill. 60201. ported experimental procedure. Shapiro et al. modeled the
TAFLOVE AND BRODWIN: MICROWAVE-IRRADIATED EYE MODEL -S&k-

plane-wave irradiation of a cranial structure [11]. The The lattice of Fig. 1 is a 19-by-39-by-19 array of unit-
absorbed power density within concentric spherical shells cell cubes, with a fixed unit-cell diameter 8 equal to 1.25
was calculated using an approach parallel to that of mm. A space lattice point is denoted as (i,j,iiI = (ib,j~,k~),
Stratton [12]. It was concluded that calculation of micro- where i, j, and k are integers. The eye–scatterer is assumed
wave heating using semi-infinite slab models is not ac- to have even symmetry about lattice planes z = 19$6
curate for tissue geometries where the ratio of the local and z = 198. This symmetry allows large savings of com-
radius of curvature to the wavelength is between 0.05 puter storage by permitting the solution of the complete
and 5. In effect, hot spots may develop deep within 10SSY, eye model with the programming of only one spatial
curved tissue scatterers of ttis size. quadrant of the model. 6 is chosen large enough so that,
The work of Shapiro et al. may be scaled to model the using this symmetry, the lattice covers a 47.5-mm-by-
eye as a Iossy sphere in free space. Such a direct scaling 48.75-mm-by-47.5-mm volume, which is sulticient to en-
suggests the presence of hot spots within the model eye- close most of the bony orbit. Yet 8 is small enough to
ball because the important ratio parameter is of the order fulfill the accuracy requirement of 0.05 wavelength resolu-
of one at microwave frequencies. Yet such a scaling is tion in all tissue for frequencies up to about 1.5 GHz.
faulty in that it neglects the wave reflection effects of the The incident plane wave is assumed to have the field com-
tissues of the bony orbit which surround the eyeball. ponents E, and Hz and propagate in the +y direction.
Unfortunately, any eye model that accounts for these This wave is generated at lattice plane y = 36 using a
effects is much more difficult to solve. Inclusion of the source condition. At planes y = O, y = 396, x = ~8, and
bony orbit eliminates the possibility of an analytic solu- z = O, the field components are determined using lattice-
tion for the fields because the geometry is no longer amen- truncation conditions. The scatterer symmetry condi-
able to classical separation-of-variables techniques. The tions, plane-wave source condition, and latti cc-truncation
improved eye model compels the use of some numerical conditions, are discussed in Section III of this paper.
method to solve for the electromagnetic fields. The location of the lattice of Fig. 1 relative to the eye-
Several computer techniques that appear relevant to ball and to the front of the skull is shown schematically
this problem have appeared in the recent literature [13]- in Fig. 2, which depicts the location of planes y = O and
[16]. Each of these methods derives a set of linear equa- z = 193. The ‘angles and dimensions of the skull cross
tions for either field variables or for field expansion co- section are based upon available cephalomet Iric data [20].
efficients, and then solves the linear system with a suitable In ordet to realize an evenly symmetric irradiation of the
matrix inversion scheme. However, it seems that none of eye–scatterer, the incident wave is assumed to propagate
these methods has been used to solve the microwave ir- in a direction parallel to the axis of one bony orbit. The
radiation of the eye. Inspection of the problems involved selected orbital axis is assumed to be the intersection of
with each method indicates either difficulty in setting up the lattice symmetry planes z = 19~~ and z = 196.
the linear system, or in finding sufficient fast, direct-access The skull surface of Fig. 2 is seen to be approximately
computer storage to invert the matrix of the linear system, parallel to the z axis of the lattice, within about 1 cm of
In this paper, we report a calculation of the microwave the right edge of the orbit. Similarly, it may be shown
fields within a detailed model of the human eye and its that immediately above and below the rim of the orbit
surrounding bony orbit. The fields are computed using the skull surface is almost parallel to the z axis. To con-
the finite-difference, time-domain solution of Maxwell’s struct a model of the eye–scatterer with the assumed sym-
equations discussed in a previous paper [17]. This tech- metry, we let the skull surface of the model extend to
nique has been found to yield first-order accurate solu- infinity parallel to the x–z plane. Thus the model orbit
tions of the electromagnetic fields within arbitrary di- is an indentation centered at ( 19~,j, 19) in an infinite,
electric scatterers of the order of one wavelength in di- planar, bony layer. The space to the rear of this layer is
ameter, situated in free space or at the surface of a half- assumed to contain only brain tissue. The major micro-
space. We also report a numerical solution of the heat wave reflection effects of the model result from the eye-
conduction equation for the eyeball, using the computed ball-orbit juxtaposition alone,
microwave heating function as the source function. The Using available anatomical data [20], [2 1], we sketch
heat equation is solved using the implicit alternating-
direction (IAD) method [18], [19]. It is determined that a * ,398
‘8 1$8 19+8,398,198
significant hot spot develops at the center of the eyeball of
t
the model at an incident frequency of 1.5 GHz.

II. THE MICROWAVE SCATTERING MODEL , : ,X*

+.<
A8.39K5-i;K&;9&o
In this section we discuss the finite-difference lattice
.“””
used for the eye and bony orbit microwave scattering
problem. The lattice boundary planes and assumed co-
ordinate axes are shown in Fig. 1. The field vector com-
ponents are positioned at distinct half-interval points in
Fig. 1. Boundary planes of the finite-difference latl,ice used for the
the lattice, as shown in [17, fig. 1]. microwave scattering problem.
--866’” IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, NOVEMBER 1975

-
NERVE

M E O~AuL&
z= 198 BITAL FAT

LATE~;:C~:CTUS
LATTICE SUPERIOR OBLIQUE
PI. ANC
MUSCLE
y=o
SUPERIOR RECTUS
MUSCLE

(a)
SUPERIOR OBLIQUE—
MUSCLE
. SUPERIOR RECTUS
MUSCLE
ORBITAL FAT

LATERAL RECTUS ~-- OPTIC NERVE

MuSCLE . .

INFERIOR RECTUS
+-J MUSCLE

B INFERIOR OBLIQUE
MUSCLE

(b)
Fig. 3. The major tissues of the orbit. (a) Top view. (b) Side view.

MEDIAL RECTUS

i BRAIN MUSCLE
I
39
Fig.2. Location of thelattice of Fig. Irelative tothe eyeball and
the front of the human skull.
35

top and side views of the major tissues of the orbit in

Fig. 3(a) and (b). The lattice of Fig. 1 contains an 30

idealized, stepped-surface representation of the tissues

of one quadrant of the orbit. This representation is de- 25

picted in Fig. 4(a) and (b) at the two lattice symmetry

u
ORSITAL FAT
AND BONE
[
planes. Comparing Figs. 3 and 4 we see that the model
20
eye simulates all rectus muscles associated with the eye-
ball. The symmetry assumptions, however, prevent simu-
lation of the oblique muscles. In Fig. 4 no bone–fat inter- SKIN ANO ‘5
MUSCLE
face is shown because of the similar dielectric parameters OF ‘v=, ‘n
“ “’”” P d
of these two tissues. Using published data on the composi- 10
r1 W,LA W&: HUMOR
tion of the eyeball media [21] and the electrical param-
eters of the other tissues found in the orbit [11], [22],
-1
the following values of e and u are assigned to each tissue.

E, at: a (mhos/meter) at:

Tissue Type 750 MHz 1.5 GHz 750 MHz 1.5 GHz (a)
EYESALL EYEBALL
AOUEOUS VITREOUS
Skin, muscle, lens 52 49 1.54 1.77 HUMOR HUMOR
ORBITAL FAT
Fat, bone 5.6 5.6 0.09 0412 AIR OPTIC NERVE
Eyeball humors 80 80 1.90 1.90 i
Brain, nerves 49 46 1.20 1.40 ls- -

INCIOEF4T
15 - P1.,llE
In constructing a model of the eye with the assumed WAVE

symmetry, several simplifications of the actual tissue

geometry are necessary. The symmetric model cannot
take into account the curvature of the skull at the sides
of the orbit, the sinus cavities of the skull, the existence 5 -

%
of the oblique muscles, and the exact shape and thickness
,
I
of the orbital wall. The accuracy of the model may be 5

increased by omitting the assumed symmetry and using SKIN ANO MUSCLE ORBITAL FAT
OF EYELIO AND BONE
a lattice to cover the entire eye–scatterer. Further, the
(b)
maximum frequency of irradiation may be increased by
Fig. 4. Cross sections of the orbital tissues of the model at the
decreasing the lattice spacing. These elaborations are de- lattice symmetry planes. (a) Plane z = 198. (b) Plane z = 19*8.
TAFLOVE AND BRODWIN: MICROWAVE-IRRADIATED EYE MODEL 891

pendent upon the acquisition of sufficient computer time

on a machine with more fast storage capacity than the
Northwestern University CDC 6400 which was used for
+ C,(m) [H=n+’/2(i + *,j + +,k)

the present research. However, the symmetric model – Hzn+ll’(i + +,j – +,,~)
simulates the basic geometry of the eye-scatterer, that of
a water-like sphere encased in a low-loss dielectric cavity. + Hvn+’/2(i + *, j,ii – *)
This model should be accurate enough to locate and de-
termine the magnitude of any high concentrations of – Hg”+’/z(i + +,j,lc + ~) ]. (4b)
microwave energy, with an acceptable level of uncertainty.
This modification eliminates the need for computer storage
III. ELEMENTS OF THE MICROWAVE
of separate c and a lattices. Now, only a MEDIA lattice,
SCATTERING ALGORITHM1
which specifies the tissue type at each lattice point, need be
In this section we discuss the modifications of the al- stored. In addition, the e, and u of each tissue can now be
gorithm of [17] necessary for the present research. These changed without having to repunch a large data card deck.
modifications include a time-stepping algorithm with Such a change involves only the recalculaticm of the five
fewer multiplications, simpler symmetry conditions, adap- values of C.(m) and the five values of Cb(m). Finally, we
tive lattice-truncation conditions, and increased stability rewrite [17, eq. (10)] as
considerations. In the discussion, any function of space
fi.”(i,3,k + ~) e-- 1000R~ sin (27r fn M) + ~i’a(-i,3,1c
-1-~).
and time is denoted as F“ (i,j,k) = F(i8,j8,1c8,n8t).
Assuming that the quantity 6t/P(i,j,k) 6 is constant for (5)
all (i,j,k) of the lattice, the algorithm of (6a) – (6f ) of
[17] requires nine multiplications per unit cell per time This modification is needed to provide a plane-wave source
step. The number of required multiplications can be re- condition at lattice plane y = 38 that agrwes with the
duced to six and the algorithm considerably simplified definition of ~,
in the following manner. We define the constants Symmetry condition [17, eq. (12)] requires extension
of the lattice 0.58 beyond the plane of symmetry. A set of
R = 8t/eo (la)
conditions useful for the assumed even symmetry of the
l?. = M’/(IYpl)eo) (lb) eye-scatterer, and requiring no lattice points beyond the
symmetry planes, is given as
Rb = 8t/’/.u@ (lC)
Hv”(19*,j,k + +) = If=”(19*,j + +,k) = o (6a)
c.(m) = 1.0 – RIT(n’z)/6,(7n) (id)

Cb(m) = Ra/er(?n) (le) flm”(i + *,j,19) = I?.”(i,j + *,19) ,= O. (6b)

where m is a tissue-type integer from 1 to 5 assigned in the We next consider a set of simple, approximate lattice-
following way: 1, air; 2, skin, muscle, lens tissue; 3, fat, truncation conditions analogous to [17, eq [9a)-(9d)].
From the basic time-step relation of this algorithm
bone tissue; 4, eyeball humors; 5, brain, nerve tissue. We
also define the proportional electric-field vector
2cfst = a (7)

it is seen that a wave requires two time steps to propagate

Using the definitions of ( la)–( le) and (2), we rewrite across a single unit cell, in air. We define the integer con-
[17, eq. (6a) -(6c) ] in a manner similar to the following: stant

Hz”+’/2(i,j + +,k + +) = Hz”-1/2(i,j + *,k + +)

l(m) = 2X(l) /X(m), x(m) : wavelength in
+ -a”(isi + +,~+ 1) tissue-type m (8)

– E.”(ijj + *J) as the number of time steps required for a wave to propa-
gate across a single unit cell, in tissue-type m. We also
+ I?z”(i,j,k + ;)
define the stored field vectors
– 17z”(i,j + !l,k + +)0 (3)
i!(i,j,lc) = &z(MEDIA(’s~-@
(i, j,k) (9a)
This modification eliminates the three multiplications
previously needed in the H part of the algorithm. Further,
we rewrite [17, eq. (6d) –(6f ) ] in a manner similar to the
following: Then the truncation condition for lattice plane x = 48 is
given by
m = MEDIA(i + ~,j,k) (4a)
Hy”(~,j,k +’$) = [fiy(%,j,k – ~) + ~J$,j,?: + +)
1 The listing of the 362 card Fortran IV source deck is available
from the authors. + l?u(~,j,k + ~) ]/3 (lOa)
892 IEEE TRANSACTIONS ON MICROWAVE THEORY ND TECHNIQUES, NOVEMRER 1975

H*”(*,j+*,k) = [a.(#,j+*,k–1) +Il=(;,j++,ii) fields are reduced less than 10 percent in propagating from
the wave source at y = 36 to the surface of the scatterer;
+ fiz(%j + M + 1) 1/3. (lOb)
yet these values are large enough to suppress the instability y
Forplaney= O for more than 600 time steps of the algorithm. This is
sufficient time for the fields to reach the sinusoidal steady
E=”(i+*,o,k) =1%”-’(i+wi) (1OC)
state.
l?.”(i,o,k ++) =E,”-’(i,l,k +*). (lOd)
IV. RESULTS OF THE MICROWAVE
Forplaney = 39$ SCATTERING PROGRAMS

flz’(i+~,39,k) =&(i+~,38,k) (lOe) In thk section we compare the results of the microwave
scattering programs for two different frequencies of ir-
~=*(i,39,k +3) =&(i,38,k+ ~). (lOf) radiation, 750 MHz and 1.5 GHz. The storage and ex-
For plane z = O ecution time requirements of each program are noted.
Next, the computed results are presented as contour maps
fi.”(i + +,j,o) = [I%(; – ~,j,l) + 2(;+ +,j,l) of the normalized heating potential Q. at the two symm-
+ F.(i + 3,3-,1)]/3 (%) etry planes of the lattice. Last, the accuracy of the loca-
tions and magnitudes of the contours of Q. is estimated.
Each run of the microwave scattering program required
about 16000 words of central memory and 213280 words
+ -i?V(i + l,j + +,1)1/3. (W
of extended core storage for compilation. A high-optimiza-
To illustrate the use of the lattice-truncation conditions, tion Fortran compiler was used to minimize the execution
we consider the
programming implementation of ( 10e) time of each run. The 750-MHz program was run 640
at point (i + *,39,k)A The values ~=”-’(i + ~,38,1c), time steps (with 8t = ~/2c = 2.083 ps), or exactly one
&-2(i + ~,38,k) , “ “ oJL (i + ~,38,1c) ,i%n(i + ~,39,k) are cycle of the incident wave. Over the last half-cycle of the
stored in an array which is manipulated like a shift register, incident wave, or 320 time steps, the program determined
with one shift for each time step. This simulates the propa- the envelope of the electric field at each unit cell of the
gation de~ay of a scattered wave between the lattice planes lattice. A total of 2300-s central processor time was re-
Y = and Y = 3913. The propagation
386 delays of several quired to execute this run. The 1.5-GHz program was run
types of tissue can be simulated by providing a sufficiently 600 time steps (with M = 2.083 ps). This reduction of 40
large register to account for the maximum delay case, time steps from the previous run was necessary to avoid
and by performing the shifting operation through only the onset of instability which was hastened by the use
Z(MEDIA(i + ~,38,k) ) + 1 words of the array. In tfis of a smaller value of u(1) for this run. The envelope of the
way the truncation condition can be made to adapt to electric field was determined over the last half-cycle of
the type of tissue at each lattice-truncation plane. the incident wave, or 160 time steps. A total of 2100 s was
We now consider the problem of the stability of the required for program execution.
overall algorithm. A comparison of the value of 8t of (7) Fig. 5(a) and (b) gives contour maps of the computed,
with the maximum value of N allowed by stability condi- normalized heating potential
tion [17, eq. (14b) ]
Q~(i,j,k) = ~ff(i,jjk) I E(i,j,k) 12/E,~j

(11) at lattice symmetry plane z = 19~, for incident frequencies

of 750 MHz and 1.5 GHz, respectively. Fig. 6(a) and (b)
indicates that the time-stepping algorithm should be are the corresponding maps at symmetry plane z = 19~&
absolutely stable. However, the inclusion of the lattice- In all of the maps the dashed lines indicate the boundaries
truncation conditions perturbs the basic algorithm to the of the tissues.
extent that this is not correct. Equations (10a) – ( 10h) Wlthm the model eyeball, the heating potential due to ,
have been found to induce a slowly growing instability. irradiation at 750 MHz peaks at the muscle interface at
No simple reformulation of the lattice truncations has the front of the eye and gradually decreases with depth.
yet been found that eliminates this problem while still Some peaking is noted at the muscle interface directly
holding wave reflections at the truncations to an accept- below the eye. Assuming an incident power level Pi.. of
able level. The growth of the instability has been found 100 mW/cm’, the total power dissipated within the model
to be greatly slowed by increasing the size of the lattice, eye is 0.13 W. The average density of absorbed power for
or by increasing the losses of the dielectric media of the this condition is 0.022 W/cm’ (assuming a model eyeball
lattice. The modeling of the irradiation of the eye-scatterer volume of 5.9 cm’). The peak absorbed power density at
is made possible by assuming a small value of u for the air the front of the eye is 0.16 W/cms, a level 7.3 times the
in front of the eye. At 750 MHz, u(1) = 0.04 mho/m; average. For comparison, we note the measured data ob-
at 1.5 GHz, u(1) = 0.025 mho/m. These values of con- tained by Guy et al. [8] for the rabbit eye irradiated at
ductivity are chosen small enough so that the incident 915 MHz at the level PinC. Here, absorption was almost
TA3’LOVE AND BRODWIN : MICROWAVE-IRRADIATED EYE MODEL 893

i j
mW/cm2, the total power dissipated within the model
eye is 0.19 W. The average absorbed power density for
thk condition is 0.033 W/cm8. The peak absorbed power
density at the center of the eyeball is 0.11 W/cm8, a level
3.3 times the average. For comparison, we again note the
data of Guy et d., but consider the results for the rabbit
irradiated at 2.45 GHz at the level Pin.. Here a dktinct
peak was observed in the vitreous body near the retina
with a density of 0.092 W/cm3.
The frequency-dependent position of the heating po-
tential peak reported in this paper and observed by Guy
et al. in the rabbit might be explained on th( ?basis of dis-
tinct resonances of the eye–scatterer and t Ihe head as a
whole. At the lower microwave frequent y, the dimerlsions
of the eye-orbit combination are too small to support a
wave concentration effect. Any peaking is l!mobably due

I I I [
to
I
15’
H
,925
.bi ,
10
1
Is
1 *i
to a resonance

However,
of the whole

at the higher microwave frequency,

head
occur behind the orbit in the brain as Guy et al. observed.
the eye–
and should therefore

25 19
scatterer can support a resonant behavior independent
(a) (b)
of possible concentration effects of the whole head. There-
Fig. 5. Maps of the computed, normalized heating potential Q* at fore, heating potential peaks are expected within the
lattice symmetry plane z = 19& (a) Incident frequency = 750
MHz. (b) Incident frequency = 1.5 GHz. tissues of the orbit above 1.5 GHz.
Any estimate of the accuracy of the microwave scatter-
ing model of the eye must take into account both the ac-

19
! .010 .074 .040 .005 curacy of the tissue model and the accuracy of the com-
---------------- . puter algorithm. The sirnplihcations of the tissue model
Is
--------- have been discussed in Section II of thk paper. Assuming
that this model conforms to the physical reality, we esti-
mate that the algorithm locates heating potential peaks
10
and contours with a maximum error of *26, or about +10
percent of the diameter of the eyeball. The uncertain y
5I If 11:020
I
.LILJl
.—
5
,:______-,------r;”---”---;
10 15 20 25
.---w

J
30 35
I
38
+J
of the magnitudes of the heating potential

prior runs of the algorithm for geometries

peaks is esti-
mated as &10 percent. These estimates arc based upon
solvable by
(a)
analytic methods. The main sources of error have been
i .005 .040 .090 “’?5 .090 .040 .005 found to be the stepped-surface approximation of the
19
scatterer and the residual wave reflections at the lattice
15
.- —______ truncations. With the present model, the error contributed
by the stepped surfaces should be of greater significance
because the effect of the imperfect lattice truncation is
10
reduced by the loss of the tissue media.

V. THE “HEAT CONDUCTION MODEL

2- +j
1 5 10 15 20 25 30 35 S* In this section we discuss the assumptions on which the
(b) heat conduction model of the microwave-irradiated eye
is based, These assumptions involve the mechanism of heat
Fig. 6. Maps of the computed, normalized heating potential Q. at
lattice symmetry plane x = 19i&. (a) Incident frequency = 750 transfer from the eye and the selection of the thermal pa-
MHz. (b) Incident frequency = 1.5 GHz.
rameters of the media of the eye. The implementation of
these assumptions in the solution of the heat conduction
constant within the eyeball at a density of 0.024 W/cma, equation is then discussed. Throughout, the tissue geom-
but peaked ~to 0.036 W/cm’, 1 cm behind the orbit in the etry and lattice point notation is consistent with that used
brain. for the microwave scattering problem.
The heating potential due to 1.5-GHz irradiation shows The basic assumption of the heat conduction model of
a pronounced peak near the center of the model’ eyeball. the irradiated eye is that the chief cooling mechanism is
The potential also peaks strongly at the muscle interfaces localized at the surface of the eyeball in the form of the
below and in front of the eye. Assuming Pi.. = 100 retinal blood supply. This is supported by l!~wo observa-
894 IEEE TRANSACTIONS ON MICROWAVE THEORY AND~CHNIQUES, NOVEMBER1975

tions. First, the interior of the eyeball is largely without meaningful for this model. The average metabolic rate of
blood supply [20]. Second, 70 percent of the overall blood body tissues is given as 1 mW/cm3 [24]. We assume that
stream supplied to the eye flows through the capillary layer the heat transfer mechanism of ( 12d), working over the
of the choroid, which is adjacent to the retina and mainly entire surface of the eyeball, is capable of dealing with
responsible for its cooling [23]. This capillary layer ex- this magnitude of heat generation over the volume of the
tends over much of the surface of the eyeball. Over the eye. Therefore, h is estimated as
remainder, surface cooling is accomplished through the metabolic rate. volume of model eyeball
tearing and wiping mechanism of the eyelids. Therefore, h%
(eye surface temperature – 37°C) -surface area of
in formulating the heat conduction model of the eye, cool-
model eyeball.
ing of the eyeball through contact with the fat and muscle
tissues of the orbit is ignored. The heat conduction equa- (13a)
tion is solved only for lattice points within the interior of
On an order-of-magnitude basis, the temperature differ-
the eyeball, subject to a boundary condition approximating
ence at the interface of the eye media and the retinal
the cooling of the retinal blood supply and the eyelids.
blood supply is estimated to be in the range 0.05°C to
For simplicity in solving the heat conduction equation,
0.5°C under normal conditions with no irradiation. With
the eye-tissue parameters p (density), C. (specific heat),
a model eyeball volume of 5.9 cm3 and a surface area of
and K (thermal conductivity) are taken to be independent
24.2 cm2, his thus estimated to satisfy the inequality
of position, temperature, and time. Further, these param-
eters are assumed to be those of water: p = 1000 kg/m3, 5 W/mz. OC -< h <50 W/m2.0C. (13b)
CP = 4178 J/kg* ‘C, K = 0.627 W/m.°C. This assump-
During microwave irradiation, h is assumed to remain
ton is made because the aqueous and vitreous humors,
fixed at its estimated, nominal value.
which comprise the bulk of the eyeball, have a water con-
There is one major simplification implicit in the heat
tent of about 99 percent by weight [21].
conduction model of (12a) – ( 12d) : the blood-flow network
The partial differential equation of heat conduction
at the eye surface is assumed to have no detailed structure.
that is solved over the volume of the eyeball is
A more sophkticated model would consider the effects of
dT(x,t) the location of blood vessels, the velocity and temperature
pcp ~ = Kv’T(x,t) + Q(x,t) (12a)
of blood flow, and the variation of blood-flow parameters
with temperature. Guy et al. [25] have recently developed
where T is the unknown temperature function and Q is several elements of the improved model for the rabbit eye.
the heating potential due to microwave irradiation. Q (x,t) The further elaboration of this method and its applica-
is related to the incident power level Pi.c and the com- tion to the human eye is a subject of future work.
puted normalized microwave heating potential Q. by The boundary and initial value problem posed by (12a) -
(12d) is solved numerically using an IAD algorithm.’ The
lattice points used are those points of the lattice of Fig. 1
that lie within the model eyeball. In thk way, a value of
(12b) the computed microwave heating potential Q. is available
at each lattice point. Symmetry is again used to reduce the
where TO = (po/eo) 1/2, Pine is given in watts/m2, and U(t) is storage and execution-time requirements of the program.
the unit step function at t = O. The hitial conditions are The even symmetry of the eye geometry and the potential
given by Q. assures that T(x,t) must also possess even symmetry.
This can be programmed as
T(x,O) = 37°C (nominal body temperature). ( 12c)
Tfi(20,j,k) = T’(19,j,k) (14a)
The boundary condition at the surface of the eyeball,
approximating the cooling of the retinal blood supply, is T“ (i,j,19i) = Tn(i,j,18;) . (14b)
given by
Boundary condition ( 12d) may be translated into a form
useful for the IAD algorithm. For example, in calculating
h. (T(r,t) .37) .Ky (12d)
the intermediate temperature value T* (i,j,k) along each
lattice line parallel to the z axis, we have
where r is a point on the eyeball surface, n is the unit vector
37ha Kff’*(~i;~al,j,~) p5)
normal to the surface at r, and h is a heat transfer param- T“(imi.,j,k) = _ +
eter in units of W/m2. ‘C. h is assumed constant over the K+h~
. .
entire surface of the eye and may range in value from zero as the necessary boundary condltlon at point (kin,j} “k).
to infinity. For h = O, 8T/8n = O at the eye surface and In (M), it is assumed that points (&in + l,j,~),
the eye is insulated. For h = co, T = 37°C at the eye
surface, the nominal body temperature.
z The listing of the 195 card Fortran IV source deck is available
We now estimate a range of values of h that is physically from the authors.
TAFLOVE AND BRODWIN: MICBOWAVE-IRRADIATED EYE MODEL 895

k
(kin + 2,j7k) j “ “ “, ( 19,j,k) lie in the interior of the model
,9? 39.0” 3.3” 39,0. 38.s. 38.0.
eyeball, and that the values of T* at these points are to -
:1
(6 r
be obtained with a simultaneous solution. Conditions ,-., ~
17
similar to (15) can be formulated or the T** arid T“+l - :.; J-J r-~
16 -
iterations along y- and z-directed lines, respectively, 1,..,
15 -
~.-l

VI. RESULTS OF THE HEAT CONDUCTION 14 -

L-., ~--!
13 -
PROGRAMS ++---, ~_J
12 -
L-- .-, ,----J
In this section, we compare the results of the heat con- II -
--------- ---- 1
duction programs for the two frequencies of irradiation of 10 Y2J J
8 9 10 II 12 13 14 15 IS 17 18 19 20 21 22 2.$ 24 25 26
the model eye. The storage and execution time require-
(a)
ments of each program are noted. Next, the computed
results are presented as contour maps of the steady-state k

,$ ?J90* 395” 40. 404” 40.0” 39.5” ~9.o” 385“

temperature function at the two symmetry planes of the
eyeball. Then the effect of varying parameter h on the [6 ij
-- .-
approach of the temperature function to the steady state 17 - ~- ---- J
16 - r
is examined. Last, the significance of the results is dis- -1
Is -
cussed. vJ
14 -
Each run of the heat conduction program required about .–
13 - r
18000 words of central memory for compilation. All pro- L.-, .-
12
L-––_l ,---
grams were run for 200 time steps, with at = 4 s. This /
II -’ ~~
L--------------i
proved sufficient to reach the thermal steady state for all 10
, !~+j

6 9 10 II 12 13 14 15 16 17 18 19 20 21 22 23 24 23 26
cases except for those with h fixed at the minimum value
in its range. The execution time required for each run (b)

was 88 s. Fig. 8. Maps of the computed steady-state temperatures at lattice

Fig. 7(a) and (b) gives contour maps of the computed symmetry plane z = 19+6 for I’ime = 100 mW/cmz and h = 50
W/mz. “C. (a) incident frequency = 750 MHz. (b) Incident
steady-state temperatures at lattice symmetry plane frequency = 1.5 GHz.
z = 19~ for incident frequencies of 750 MHz and 1.5 GHz,
respectively. Fig. 8(a) and (b) are the corresponding con-
of the maps, parameter h is fixed at 501T’/m2. ‘C, and
tour maps at lattice symmetry plane z = 19*3. In all
power level Pine is fixed at 100 mW/cma. The dashed lines
indicate the boundaries of the tissues of the eyeball.
For the case of 750-MHz irradiation, the peak temper-
i
27r r 8
~------.
ature of 39.3°C k located at the back of the lens. For the
26 -

25 -

24 r

23 -

22 -
~--!
I
r--
~--.
~.-.

,
!.!

38.09 -
m“”’” 1.5-GHz
at about
temperature
due to 750-MHE
case, the peak temperature
the center
due to 1.5-GHz
heating,
of the eyeball,
heating
even though
of 40.4°C
We see that
is higher
the magnitude
is located

than
the lens
that
of
I the heating potential in the lens area follows an exactly
21 -
-.4 , opposite behavior. Evidently, this is because the surface
20 ;
1
cooling mechanism of the model eye is more effective for
19 ; r
3e.5” / the shallow 750-MHz heating pattern than for the rela-
18 \ r

17 ; L tively deep 1.5-GHz heating pattern. This is a strong in-

16[ /’ ‘l\ll[ 44W” dication that tissue thermodynamics must be taken into
39.0” L
account when analyzing microwave heating of organs.

I‘[ (“---J
Is ‘

‘w
L. .-, L-. .1
14 Fig, 9 is a graph of the peak eye temperature versus
13 40.09 time, for the 1.5-GHz case. The two curves plotted repre-
L--l ~.... L.., ~---
12 39,3”
L- --- J L.-, sent the behavior of this temperature for the extreme
r;
lit
k\
.x._-l ; i--- : 39.5” values of the range of h of ( 13b). Irradiation is assumed to
lot
———L———
..-— 39.0- begin at t = O and continue from then on with a con-

;1,,, !-T:CIP,I
II
,,,>-733:
R 13 14 15 16 17 18 19 II 12 13 14 15 16 i? IS 19
stant
h = 50 W/m2
10 min
power level of 100 mW/cm2.

after
We see that for

the
● ‘C thermal
start of
equilibrium
heating, with
is reached
a nmximum
about
tem-

(a) (b) perature of about 40.4”C. However, for h = 5 W/m’. ‘C,

Fig. 7. Maps of the computed steady-state temperatures at lattice eqtiilibrium is not attained until much latlw, estimated
symmetry plane z = 193 for Pin. = 100 mW/cmz and h = 50 to be about 45 min. For this case, the maximum achieved
W/m3. “C. (a) Incident frequency = 750 MHz. (b) Incident
frequency = 1.5 GHz. temperature is about 56”C. This implies tll~at the peak
896 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, NOVEMBER 1975

44 I I I I I I I
REFERENCES
[1] F. G. Hirsch and J. T. Parker, “Bilateral lenticular opacities
43 occurring in a technician operating a microwave generatm,”
Am. Med. Assoc. Arch. Ind. Hyg., vol. 6, pp. 512-517, Dec. 1952.
3 [2] M. M. Zaret, “An experimental study of the cataractogenic
: effects of microwave radiation, ” Zaret Foundation Rep. RADC-
~ 42
. TDR-64-273 (Contract AF 30(602)-3087), Oct. 1964.
[3] L. Birenbaum, G. M. Grosof, S. W. Rosenthal, and M. M.
Zaret. “Effect of microwaves on the eve.” IEEE Trans. Bio-
41 med. kng., vol. BME-16, pp. 7–14, Jan.” 1969.
[4] R. L. Carpenter, D. K. Biddle, and C. A. Van Ummersen,
“Opacities in the lens of the eye experimentally induced by
40
exposure to microwave radiation,” IRE Trans. Med. Electron.,
vol. PGME-7. mJ. 152–157. Julv 1960.
[5] S. M. Michael@, “Biological ;ffects of micr?wave exposure,”
in Proc. Btologwal Eflecta and Health Implications of Microwave
39 Radiation Symp., Med. College Virginia (Richmond, Sept.
1969), Rep. BRH/DBE 70-2, Sept. 1969, pp. 35-58.
[6] U. S. A. Standard, “Safety level of electromagnetic radiation
t
38
with respect to personnel,” U.S.A. Standards Inst., New York,
N. Y.. IJSAS CXM.1. 1966.
[7] A. W. Guy, “l?utu;e research directions and needs in biologic
v
electromagnetic radiation research,” in Biologic E.fects of Non-
1 1 1 I I 1 I
37! ionizing Radiation, Ann. N. Y. A cad. Ski., vol. 247, pp. 539-
0 100 200 300 400 500 600 700
545, Feb. 1975.
TIME FROM THE BEGINNING OF IRRADIATION (seconds)
. . A. W. Guv. J. C. Lin. P. Kramar. and A. F. Emerv. “Meesure-
[81
Fig.9. Range of thepeak eyetemperature versus time for incident ment of absorbed power patterns in the head and ejes of rabbits
frequency = 1.5GHz andPinc = 100mW/cmz. exposed to typical microwave sources,” in Proc. 1974 Conf.
Precision Electromagnetic Measurements (London, England,

temperature within the model eye is in the range of 40.4°C .[91. ‘UIY’
A. Anne.1974)’M. ‘i!”25&257”
alto. O. M. Salati. and H. P. Schwan. “Penetra-
tion and’ thermal dissipation of ‘microwaves in tiss~es,” Univ.
to 56°C for the assumed range of values of h. Pennsylvania, Philadelphia, Tech. Rep. RADC-TDR-62-244,
For comparison with the results presented here, we note Contract AF30(602)-2344, ASTIA Dec. 284981, 1962.
[10] H. F. Cook, “A physical investigation of heat production in
the temperature distribution within the rabbit eye ir- human tissues when exposed to microwaves,” Brit. J. Ad.. .
radiated at 2.45 GHz, computed by Guy et al. [25]. For phyS., vol. ~, pp. 1–6, 1952.
[111 A. R. Shamro. R. F. Lutomirski. and H. T. Yura. “Induced
P. Inc = 100 mW/cm’, a maximum temperature of 40.O”C fields and heating within a cranial structure irradiated by an
was calculated on the center line of the eyeball, approxi- electromasmet,ic rdane wave.” IEEE Trans. Microwave Theoru
Tech. (Sp%al I&ue on B~;logical E.fects of Microwaves), vo~
mately ~ of the eyeball diameter from the front surface. MTT-19, pp. 187–196, Feb. 1971.
Thermal equilibrium was reached within about 20 min of [121 J. A. Stratton. EZectromaanetic Theorv. New York: McGraw-
Hill, 1941, pp.’563-573. “
the start of irradiation. The magnitude and position of the [13] D. E. Livesay and K.-M. Chen, “Electromagnetic fields in-
peak rabbit-eye temperature and the time required to duced inside arbitrarily shaped biological bodies,” IEEE Trans.
Microwave Theory Tech. (Part II of TWO Parts—1971 Symp.
reach the steady state agree well with the 1.5-GHz re- Issue), vol. MTT-22, pp. 1273-1280, Dec. 1974.
sults (with h = 50 W/m2 ● “C) for the human-eye model [14] T. K. Wu and L. L. Tsa$ “Numerical analysis of electromag-
netic fields in biological tissues, ” Proc. IEEE (Lett. ), vol. 62,
reported in thk paper. This agreement offers promise that pp. 1167–1 168, Aug. 1974.
[15] B. H. McDonald and A. Wexler, “Finite-element solution of
a relation between the microwave heating of the rabbit
unbounded field problems,” IEEE Trans. Microwave Theory
eye and of the human eye can be established. Tech. (1972 S’ymp. Issue), vol. MTT-20, pp. 841–847, Dec. 1972.
[16] D. R. Wilton and R. Mittra, “A new numerical approach to
VII. CONCLUSIONS the calculation of electromagnetic scattering properties of two-
dimensional bodies of arbitrary cross section,” IEEE Trans.
The absorbed power distribution pattern within a de- Antennas Propagat., vol. AP-20, pp. 31 W317, May 1972,
[17] A. Taflove and M. E. Brodwinl “Numerical solution of steady-
tailed model of the microwave-irradiated human eye has state electromagnetic scattering problems using the time-
been computed for two different frequencies of plane-wave dependent Maxwell’s equations,” IEEE Trans. Mi~owave
Theory Tech., vol. MTT-23, pp. .623-630, Aug. 1975.
irradiation, 750 MHz and 1.5 GHz. The temperature dis- [18] B. Carnahan, H. A. Luther, and J. O. Wilkes, Applied Numerical
tribution induced by the microwave heating within the Methods. New York: Wiley, 1969, pp. 452461.
[19] P. L. Brian, “A finite-difference method of high-order accuracy
model eye has been computed. The results indicate that for the solution of three-dimensional transient heat conduction
a distinct hot spot exceeding 40.4°C probably occurs deep problems,” Amer. Inst. Chem. Eng. 3., vol. 7, pp. 367–370, 1961.
[20] S. Duke-Elder and K. C. Wybar, System of Ophthalmology.
within the eye at a frequency of 1.5 GHz. The results Vol. II: The Anatomy of the Visual System. London, Englancl:
Kimpton, 1961, ch. 5.
imply that hot spots within the eye are to be expected at
[21] H. Davson, Ed., The Eye. Vol. 1: Vegetative Physiology and
frequencies higher than 1.5 GHz, due to wave-concentra- Biochemistry. New York: Academic, 1962.
[22] C. C. Johnson and A. W. Guy? “Nonionizing electromagnetic
tion effects of the eyeball-orbit juxtaposition. Comparison
wave effects in biological materials and svstems, ” Proc. IEEE.
of the temperatures within the human-eye model with vol. 60, pp. 692–718,-June 1972. “
[23] A. Roulier, ‘[Calculation of temperature increase in the eye pro-
those calculated by Guy et al. [25] for the rabbit eye
duced by intense light,” Bull. Math. Biophys., vol. 32, pp. 403-
indicates the possibility y of relating the microwave heating 427.1970.
[24] H. ‘P:- Schwan, “Microwave radiation: biophysical considera-
patterns within the respective organs. A more detailed
tions and standards criteria,” IEEE Trans. Biomed. Eng., vol.
model of the human eye and computer runs for “frequencies BME-19, pp. 304-312, July 1972.
[25] A. W. Guy, J. C. Lin, P. O. Kramar, and A. F. Emery, “Effect
higher than 1.5 GHz are required to explore this possible
of 2450-MHz radiation on the rabbit eye,” IEEE Trans.
relation. Mia-owave Theory Tech., vol. MTT-23, pp. 492-498, June 1975.