Professional Documents
Culture Documents
0
B
z
z
(4)
where
0
= 4 10
7
H/m is the permeability of free space.
The gravity force on the object is
F
g
=
m
gV (5)
where
m
is the mass density and g = 9.8 m/s
2
is the gravita-
tional acceleration. The ratio of magnetic force to gravitational
force is
Force ratio =
F
m
F
g
=
M
s
m
g
B
z
z
. (6)
For an object made from iron with M
s
2.2 T and
m
=
8900 kg/m
3
and a eld gradient of 2 T
2
/m, the magnetic force
is about 40 times the gravitational force. This will obviously be
Fig. 2. Diagram of the deection angle method for evaluation of translational
force for an implant.
Fig. 3. Measured force normalized to the gravitational attraction versus
distance along axis of bore of a 1.5-T MR system for Medtronic Itrel 3 IPG.
Positive distance is away from the center of the bore and a distance of 0
corresponds to the edge of the bore.
a massive force, and in general strongly magnetic objects are a
hazard in the MRI environment.
The magnetic force on an implant may be evaluated [4] using
the deection angle technique shown in Fig. 2. Along the axis
of a cylindrical bore with magnetization along the bore axis, the
force ratio is given on the tangent of the deection angle . As
an example, Fig. 3 shows the force versus position along the
axis of the bore of a 1.5-T MR system for an implantable pulse
generator (IPG) used for neurostimulation [5]. For this device,
the greatest force ratio is about 0.4, and this force occurs near
the edge of the bore.
B. Magnetic Torque
For a nonspherical magnetic object, the magnetization will
not be precisely along the magnetic eld and thus there will be
a magnetic torque. The torque will be in a direction such that
the long axis of a ferromagnetic object will tend to align along
the magnetic eld [6]. Fig. 4 shows the geometry for calculation
of the torque. The magnetic eld H
0
is assumed to be along the
z-axis and the object is assumed to be uniformly magnetized
with saturation magnetization M
s
in the xz plane.
The relevant magnetic energies are those due to external and
demagnetizing elds. Using SI units, the total magnet static
energy per unit volume is written as
W
T
=
M
2
s
2
0
(N
n
N
t
) sin
2
M
s
H
o
sin( + ) (7)
where N
n
and N
t
are the normal and transverse demagnetizing
factors, respectively.
NYENHUIS et al.: MRI AND IMPLANTED MEDICAL DEVICES: BASIC INTERACTIONS WITH AN EMPHASIS ON HEATING 469
Fig. 4. Geometry for evaluation of the torque on a soft ferromagnetic object.
is the device angle relative to the x-axis and is the direction of the
magnetization relative to the normal.
At equilibrium, it is required that W
T
/ = 0. Thus
M
2
s
2
0
(N
n
N
t
) sin 2 M
s
H
o
cos( + ) = 0. (8)
Make the denition
=
M
s
2
0
H
0
(N
n
N
t
). (9)
The energy minimization (8) can then be written as
cos( + ) + sin 2 = 0. (10)
The torque about the y-axis is
y
= M
s
H
0
cos( + ) volume (11)
where volume is the device volume.
The maximal amplitude of the torque is
max
=
M
2
s
2
0
(N
n
N
t
) volume. (12)
Note that the maximum torque is relatively insensitive to the
value of the static magnetic eld, though the angular depen-
dence will depend somewhat on the strength of H
0
.
The maximal difference in demagnetizing factors will occur
for a long needle-shaped object. In this case, N
n
N
t
0.5.
Note that the maximal torque is proportional to the square of
the magnetization for objects made from a single material. The
torque and magnetic force are correlated; objects exhibiting
minimal force will generally exhibit minimal torque. It is
noteworthy that the maximal torque does not increase with
eld strength.
Torque may be measured by placing the implant in a holder
suspended by a torsional spring [6]. The torque is proportional
to the deection angle of the spring from the equilibrium
position. The angular dependence of the torque is determined
by measurement of the deection angle as a function of the
device position. Fig. 5 shows the measured angular dependence
of the torque for an IPG used for a neurostimulation system in a
1.5-T static magnetic eld. The torque curves are well t by the
solid lines for = 0.25 and maximum torque = 0.018 N m.
There is qualitative agreement between the measured and the
tted torque, even though the magnetic structure of an IPG is
very different from the idealized geometry of Fig. 4.
Fig. 5. Measured torque versus angle for a Medtronic Itrel 3 IPG. The points
are measured and the solid line is calculated for = 0.25 and maximum
torque = 0.018 N m.
Magnetic forces on implants that are less than the force of
gravity (i.e., deection angle less than 45
max
= 16 T/s
_
1 +
0.36 ms
t
s
_
(13)
where t
s
is the duration of a rectangular dB/dt pulse. Ac-
cording to the IEC standard, MRI investigators may also use
experimental data obtained from human subjects to determine
the limiting gradient output.
A metallic implant that has conducting components will tend
to concentrate the induced gradient currents [11]. The greatest
470 IEEE TRANSACTIONS ON DEVICE AND MATERIALS RELIABILITY, VOL. 5, NO. 3, SEPTEMBER 2005
Fig. 6. Qualitative distribution of currents in a human subject induced by
pulsed gradients for (a) the y-gradient coil and (b) the z-gradient coil.
focusing is expected to occur if the implant has the shape of
a long wire or forms a closed loop of sufcient size, such as
lead used for a neurostimulation system, cardiac pacemaker,
or guidewire. If there were indeed a signicant concentration
of the gradient currents, this could result in nerve stimulation.
However, electric eld enhancements less than a factor of three
should not result in nerve stimulation because inhomogeneities
in conductivity within the body also provide a similar level
of enhancement. The electrical impedance between metal and
surrounding tissues will tend to limit the concentration of the
currents. The higher intensity gradients available in echo-planar
or fast gradient echo MR imaging may enhance the likelihood
of nerve stimulation. The induced voltages may also interfere
with the operation of active medical devices.
The gradient elds induce a current in a metallic object.
This induces a magnetic moment in the object and thus there
is a torque induced by the MRI static eld. The implant will
then exhibit a high-frequency vibration, which is potentially
uncomfortable for large highly conductive implants. This effect
may have been observed in a cervical xation device [12].
IV. INTERACTIONS WITH THE RF FIELD
Fig. 7 shows a representative modulation shape of an RF
pulse associated with MRI. While the actual modulation is
complicated [13], it can be idealized by a sinc function. The
sinc pulse has a maximal amplitude of B
10
and the time
modulation dependence is given by
B
1
(t) = B
10
sinc
_
1
t
_
. (14)
The bandwidth f of the pulse is given by
f =
1
1
. (15)
Fig. 7. Modulation of an RF pulse used for MRI.
Fig. 8. Model for current induced in a lead wire by coupling with the electric
eld induced by the RF magnetic eld associated with MRI.
The pulse amplitude B
10
is adjusted to achieve the desired
ip angle of the nuclear magnetization. In the absence of loss,
the relationship between B
10
and ip angle is
B
10
=
f
(16)
where = 2.675 10
4
radians/(G s) is the nuclear gyro-
magnetic constant. As an example, consider a pulse with a
bandwidth of 1250 Hz and producing a ip angle = (180
(17)
where is the conductivity, E
10
is the electric eld amplitude,
and is the mass density. The IEC standard limits the whole-
body average SAR to 2 W/kg in normal mode operation [10].
Neglecting thermal transport, the time rate of temperature
rise dT/dt in response to the SAR is given by
dT
dt
=
SAR
C
(18)
where C
= 4186 J/(kg
=0
2
_
=0
I
t
(z
)
(2)
2
_
2
z
2
+ k
2
_
e
jkR
4R
d
dz
I
s
(z)ddz
+
L
_
z=0
2
_
=0
b
_
r=a
j
2r
m
dI
t
(z)
dz
(
m
i
)
2r
i
dI
s
(z)
dz
rdrddz
_
=
L
_
z=0
E
sz
(z)I
t
(z)dz (19)
476 IEEE TRANSACTIONS ON DEVICE AND MATERIALS RELIABILITY, VOL. 5, NO. 3, SEPTEMBER 2005
Fig. 13. Electric eld intensity versus position along the length of the
DBS lead for calculation of the current distribution shown in Fig. 14.
Z = 0 is the center of the Model 3389 DBS lead and the landmark is at
Z = 0.06 m.
where I
s
and I
t
are the unknown surface current on the
wire to be determined and the known testing current. E
sz
=
E
0
z, where E
0
is the incident electric eld. R is the
distance from the current source point to the evaluation point,
m
= j(/) is the complex permittivity of the medium,
is the conductivity of the medium, and
i
are the
permittivity of the medium and insulation, k =
m
is
the medium wave number, L is the length of the wire, a is
the radius of the conductor, and b is the outer radius of the
insulation.
Equation (19) can be solved by applying the Galerkin method
of moments, where the unknown current I
s
is assumed to be
the linear combination of the basis functions and the testing
functions I
t
are all of the basis functions. As a result, (19)
will form a set of linear equations with the number of variables
(the weights of the basis functions of the unknown current) and
equations equal to the number of basis functions. The set of
basis functions is the one described in [59].
A more detailed description of the method can be found
in [49] and [59]. Compared to the previous work, where a uni-
form electric eld was assumed, the calculations here make use
of the nonuniform electric eld in the phantom. The incident
electric eld distribution along the length of the lead position
was obtained from our FDTD program that includes the RF
coil and the phantom [55]. Fig. 13 shows the electric eld
distribution used in the calculations. The electric eld is scaled
for the calculations so that the maximum amplitude is 86 V/m,
which corresponds to an SAR of 1 W/kg.
Fig. 14 shows currents induced on the lead wire for the
electric eld distribution in Fig. 13. The peak current amplitude
on the wire is about 0.14 A. There is a discontinuity in current
at the end of the wire. There is thus a charge accumulation at
the end, which results in a concentration of electric eld.
The SAR distribution in watt per kilogram surrounding the
lead model was obtained using (17). From the calculated SAR
distribution, the temperature distribution near the end of the
Fig. 14. Calculated real, imaginary, and magnitude components of the current
distribution along a 28-cm-long insulated wire according to the model in
Fig. 13 with 6 mm of insulation removed from the end. The electric eld
distribution is shown in Fig. 13.
lead was evaluated by using the standard bioheat equation
given by
C
p
T
t
= K
2
T + SAR b(T T
b
) (20)
where T = T(r, z, t) is the temperature (
C) at time t, C
p
is
the specic heat (J/(kg
2
_
_
1
2
. (21)
For /(2) = 64 MHz and relative dielectric constant
/
0
= 77, the wavelength
0
in the absence of loss is 0.52
m. For the phantom material with conductivity = 0.27 S/m,
the wavelength is calculated to be 0.48 cm. With no insulation,
a maximum in temperature rise is expected for lead length
approximately equal to half a wavelength. However, there is no
evidence of resonance in the measurements and calculations in
NYENHUIS et al.: MRI AND IMPLANTED MEDICAL DEVICES: BASIC INTERACTIONS WITH AN EMPHASIS ON HEATING 477
Fig. 15. Solid line is the measured temperature rise at the surface of electrode
0 (probe 1 in Fig. 9) versus length of section of the Medtronic Model 3389 lead.
Circles are the calculated temperature rise according to the model in Fig. 12.
The electric eld is that shown in Fig. 13.
Fig. 15. This is because the insulation increases the length for
maximal heating since its dielectric constant is less than that of
the phantom material.
XI. EFFECT OF LOOPS ON THE RF-INDUCED
TEMPERATURE RISE DURING MRI
Baker et al. [35] reported that loops in the lead could reduce
the RF-induced temperature rise in a neurostimulation system
used for DBS. In measurements with bare and insulated wires,
Nyenhuis et al. [61] found that the temperature rise for the
bare wires decreased as the number of loops increased from
0 to 2, whereas the temperature rise for the insulated wires
was relatively independent of the number of loops. To further
investigate the potential effects of loops on heating of a DBS
system, we consider a lead wire model with the geometry
shown in Fig. 12. The induced current is calculated for loops of
2-cm diameter centered 11 cm from the end of the lead. The end
to end length of the lead wire is constrained to be 28 cm. The
background electric eld is assumed to be 86 V/m, yielding an
SARof 1 W/kg, and the frequency is 64 MHz. Other parameters
are the same as used to calculate the current distribution shown
in Fig. 14.
The calculation is made using the method of moments. As
for the current calculations on straight wires, it is assumed
that the lead wire is a good conductor and thus the boundary
condition is that the electric eld is zero at the surface of the
wire. By including the polarization charges and currents due to
the insulator and assuming that the wire is thin, the electric eld
integro-differential equation to calculate the induced current on
an insulated wire is given by [62], [63]
E
i
t
=
j
m
_
L
_
k
2
G
a
(s, s
)I(s
) s s
+
m
i
dI(s
)
ds
G
a
(s, s
)
s
+
i
i
dI(s
)
ds
G
b
(s, s
)
s
_
ds
(22)
Fig. 16. Calculated current distribution versus normalized distance along the
length of the lead wire model of Fig. 12 with zero, one, two, three, and
four loops of 1-cm diameter and 3.l4-mm pitch. The 2-cm-diameter loops are
centered 11 cm from the right side. The end to end length is 28 cm and there
are 6 mm of insulation removed from the end. The uniform background electric
eld is 86 V/m, yielding a background SAR of 1 W/kg. Other parameters are
the same as used to calculate the current distribution shown in Fig. 14.
where E
i
t
is the tangential incident eld; I(s
) is the current
at s
; s and s
m
= j(/) is the complex permittivity of the medium;
k =
m
is the medium wave number; and are the
permittivity and conductivity of the medium;
i
is the permittiv-
ity of the insulator; G
a
(s, s
) = e
jkR
a
/4R
a
and G
b
(s, s
) =
e
jkR
b
/4R
b
are Greens functions dened for inner radius
a and outer radius b, where R
a
=
_
|r(s) r(s
)| + a
2
and
R
b
=
_
|r(s) r(s
)| + b
2
; and r(s) and r(s