Thermal dosimetry from radio-frequency fields within a head phantom assessed during three tesla magnetic resonance imaging


Material Information

Thermal dosimetry from radio-frequency fields within a head phantom assessed during three tesla magnetic resonance imaging
Physical Description:
v, 167 leaves : ill. ; 29 cm.
Harder, George
Publication Date:


Subjects / Keywords:
Environmental Engineering Sciences thesis, Ph.D   ( lcsh )
Dissertations, Academic -- Environmental Engineering Sciences -- UF   ( lcsh )
bibliography   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph.D.)--University of Florida, 1998.
Includes bibliographical references (leaves 159-166).
Statement of Responsibility:
by George Harder.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002388706
oclc - 40471503
notis - ALZ3574
System ID:

Full Text








My thanks go out to my family and friends for their moral support and

encouragement, with special thanks to my sisters Virginia, Angie, Iris and Carol. My

appreciation and respect go to my committee for their confidence in my abilities. I want

to thank Dr. JeffFitzsimmons, Dave Peterson, Michelle Werner, Jim Scott and Dr. Randy

Dunsing for their patience in answering my questions. Special thanks to Dr. Emmett

Bolch for his moral and financial support during my extended period as a graduate

student. I wish to specifically thank Ben Warren for allowing me the time away from

work to pursue this degree.


ACKNOW LEDGMENTS ................................. ............ ii

ABSTRACT ............................................ ........... v


1 INTRODUCTION ............................................... 1
MRI Fields ............................... ................... 1
Static M agnetic Fields ................................. 1
Gradient M agnetic Fields .............................. .... 2
Radio-Frequency Fields ........... ......................... 4
Current Standards ............. ...... ............ ............. 6
INIRC-IRPA ..................... .......... ............ 6
AN SI ................ .......... ........... ............ 7
IE C . . . . . .. .. 8
FDA ...................... ......................... 10
Contrast ..................... ........... ............ 12
Specific Absorption Rate ................................... 13
Loading Factors .......................... .............. 14
SAR Development with Far-Field Assumptions .................. 16
SAR Development Without Far-Field Assumptions ................ 21

2 MATERIALS AND METHODS ............... ....................... 29
Head Phantom ................. ................. ........... .. 29
Thermocouple Measurements ............................... 30
Fiber Optic M easurements ................................... 32
Heating Protocol ........................................ 33
Thermal Imaging .................... ......................... 34
T1 Relaxation Time Imaging ............................ 34
Diffusion Imaging ................... .................... 37
Proton Chemical Shift Imaging ........ . . 40
Phantom Thermal Imaging .. .... ............ 42

3 R E SU LTS ........................................................ 48
Thermocouple Results ....................... .............. 48
Fiber Optic Results ................... ...... ............... 50
Heating Equation ......................................... 51
Cooling Equations ............. .......................... 54
Diffusion Imaging ................ ........................ 55
Procedures ............ ............. ..... ........... 55

4 D ISCU SSION ................... ................................. 78

5 CONCLUSION ........................ ........................ 82
Future W ork .......... ......................... ............. 83



B FOOD AND DRUG ADMINISTRATION (FDA) ....................... 116

C PHANTOM PROTOCOL THERMOCOUPLE ........................ 118

D PHANTOM PROTOCOL FIBER OPTIC ............................ 119

E HEAT STRESS INDEX ................................. .......... 120

F EXPERIMENTAL RESULTS ....................................... 131

G ARTIFACT TUNING ............................................. 140

H DIFFUSION IMAGING ................... ...... ................. 142

I DIFFUSION IMAGING PROTOCOL ................................. 143

J IM AGE TRANSFER ............................................... 144

K DIFFUSION IMAGE CONVERSION PROGRAM ....................... 146

LIST OF REFERENCES .......................................... 159

LIST OF SUPPLEMENTAL REFERENCES ............................. 162

BIOGRAPHICAL SKETCH ........................................ 167


Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



George Harder

August 1998

Chairperson: Dr. W. Emmett Bolch
Major Department: Environmental Engineering Sciences

Magnetic Resonance Imaging (MRI) provides radiologists with the tools to better

diagnose disease and view internal injuries. As field strengths increase, so does resolution

and so do potential health effects. This dissertation focused on tissue heating generated by

the radio-frequency (RF) energy. Thermal measurements were first made using

thermocouple probes. Experimental results, which were affected by the metallic nature of

the thermocouple, were discarded. The experiments were repeated using fiber optic

probes. Tests reveled uneven thermal energy deposition throughout the head phantom.

Temperatures nearer the outer edge increased sooner. Additionally, there was thermal

movement resulting in increased temperatures to the center of the head phantom up to an

hour after the RF fields were removed. Finally, a diffusion imaging technique was

performed prior to and after a scan. Analysis of the resultant images indicated diffusion

imaging may be able to monitor temperatures less than the 1 C as stated in literature.


The applied theory of magnetic resonance provided imaging and spectroscopy.

Increased field strengths enhanced resolution providing radiologists improved, non-

invasive diagnostic tools. Increased field strengths also brought increased potential for

health effects. The MRI system utilizes the nuclear spins of isotopes to define the

presence of the isotope. At the three Tesla (3T) MRI unit, jointly owned by the Veterans

Administration and Shands Teaching Hospital, the isotope of interested during the

research of this project was the proton (hydrogen). Since the human body contains a high

percentage of water, proton imaging would provide some type of image from every tissue

type and every location within the body. MRI produces three fields; the static magnetic

field, the gradient magnetic field and the RF field.

MRI Fields

Static Magnetic Fields

According the to Food and Drug Administration (FDA), clinical MRI systems in

the United States were permitted to function on a routine clinical basis at static magnetic

field strengths of up to 2T. This was recently increased to 4T. Above this level, evidence

of safety must be provided by the sponsor or device manufacturer prior to routine clinical

use. Static magnetic fields have no effect on skin or body temperature (Budinger, 1992).


Heat was only deposited in tissue when current flowed through the tissue and current was

only induced by a changing a magnetic field. In this respect tissue was invisible to static

magnetic fields. Patient movement within a static magnetic field was considered a

dynamic magnetic field from the tissue's viewpoint. However, this movement was slow

and was insufficient in creating a noticeable current.

Problems that have been reported by patients after entering the MRI chamber but

before the gradients fields were applied were dizziness and nausea. These effects mimic

motion sickness and have a similar root cause. Balance has always been perceived by the

circulating inner ear fluids. The calcium carbonate particles, called otoliths, moved against

the hair cells in the saccule and utricle sending false signals to the brain that the head was

moving. The brain received conflicting signals from the inner ear and the other senses.

Increased discomfort compounded the problem. For individuals easily susceptible to

motion sickness, very little head movement was required. Since this problem dealt with

the motion of the otoliths, this effect was also associated with gradient magnetic fields.

Gradient Magnetic Fields

Effects have been reported by patients subjected to changing magnetic fields. At

switching rates of two Tesla per second (T/s), patients have reported light flashes called

magnetophosphenes. Switched magnetic field excited the optic nerve inducing electrical

currents. The optic nerve responded as if it received a light pulse from the eye. Since

these were random and isolated, the result appeared as flashes of light. This effect was

also be achieved in a static magnetic field by rapid movement of the head (Kanal 1990).

At switching rates of 60 T/s patients have experienced neuromuscular excitation.

Faraday's Law of Induction stated that a conductor, in the presence of a time-varying

magnetic field, would induce a voltage in the conductor that was oriented perpendicular to

the rate of change of the magnetic field. When a voltage difference along a nerve

approached 6 V/m, neurological impulses trigger a peripheral nerve stimulation. This

peripheral nerve stimulation was an uncontrolled, involuntary skeletal muscle contraction

and/or twitching (Reilly 1989, Schaefer 1992). One FDA guideline to limit patient

exposure to time-varying magnetic fields specifically required demonstration, with "valid

scientific evidence," that the rate of change of magnetic field for the system was not

sufficient to cause peripheral nerve stimulation. At switching rates of 500 T/s cardiac

excitation became possible. A perturbation of an electrocardiogram was identified as a

superimposed signal of the gradient pulse (Budinger 1992). No long term effects were

demonstrated by the superimposed signal.

While biological tissue was predominantly transparent to dynamic magnetic fields,

there were still elements within the tissue that responded to magnetic fields. As magnetic

fields increased, these affected elements began causing noticeable effects. Currents,

induced by the changing magnetic field, flowed through the resistive body fluids to

produce heat. The amount of heat or energy produced was a function of the frequency,

intensity and duration of the exposure and the coupling between the RF coil and the


Radio-Frequency Fields

The predominant effect of RF fields was RF heating. The biological factors of

importance were conductivity, specific gravity of the tissue, size of the current loop,

circulation and the ability of the body to loose heat. An issue that often arose was not

whether localized heating existed but how much and what was its biological impact. The

majority of damage to cells and molecules was through ionization caused by heat. RF

fields were too high in frequency to stimulate excitable tissues electrically however these

fields deposited energy into tissues which raised their temperatures. The normal body

temperature remained approximately 37 C. The highest safe core temperature for

individuals was 39.4 C. The skin pain threshold was 43 C and proteins denatured at

about 45 C. This small range of temperatures became critical at higher field strengths.

The potential for localized heating and subsequent tissue damage increased as RF

frequencies increased to the point where the patient's size approached one-half

wavelength of the incident excitation frequency (Gordon 1992).

An area of major concern for MRI of the head was the eye. The vitreous body of

the eye contained a colorless, jelly mass between the lens and the retina. Heat removal

resulted from the formation and circulation of the aqueous humor as well as from the

ophthalmic artery. The eye had a reduced capability for heat dissipation as a result of a

lack of vascularization. Acute near-field exposures of RF radiation to the eye of

laboratory animals created cataracts as a result of the thermal disruption of ocular tissues.

The threshold for cataractogenesis had been estimated at 100 W/kg. An investigation

revealed no discernible effects on the eyes of rats resulting from MR procedures at

exposures that far exceeded typical levels used in the clinical setting (Shellock 1992).

Patients exposed to SAR levels of 0.5 to 1.3 W/kg had shown significant

elevations in temperatures in the cornea. The cornea was anterior to the lens and

contained no blood vessels. Heating of the cornea had the potential to lead to keratitis, an

inflammation of the cornea while heating of the lens accelerated the onset of cataracts.

The cornea was pain-sensitive and a conscious patient would be able to feel this heating

before it became severe.

Various underlying health conditions affected an individual's ability to tolerate RF

heating. The state of a patient's health, cardiovascular conditioning and obesity were

primary factors. Various medications (e.g. diuretics, beta-blockers, calcium blockers,

amphetamines, muscle relaxers and sedatives) also altered the thermoregulatory responses

to RF heating. The heat loss characteristics of the body were not applicable to head

phantom studies so tissue heating characteristics from MRI to the head phantom,

therefore, provided a worst-case scenario. Such conditions would exist for a stroke victim

with limited thermoregulatory responses in the head.

Static and dynamic magnetic fields produced temporary effects to patients with no

long term consequences. The effects from RF fields resulted in increased temperatures to

patients which could be dangerous. A head phantom provided a versatile platform for

conducting heating tests and defining methods to measure and monitor thermal changes.

In order to determine what method to use to measure temperature changes, the range of

acceptable temperature changes were derived from standards.

Current Standards

Radiation falls into one of two categories: ionizing and non-ionizing. Ionizing

radiation is defined as having sufficient energy to ionize an atom. MRI falls into the

category of non-ionizing radiation since its electromagnetic fields create insufficient

energy to ionize molecules, but sufficient energy for thermalization. To quantify this

thermal action the energy absorption over a small mass of material is defined as the

specific absorption rate (SAR) and is given in units of watts per kilogram. Although the

SAR defines the energy absorbed within tissue, it has always been impractical to measure

this energy change in live subjects. Instead the surface power density, given in W/cm2, has

been measured. Utilizing the tissue density and tissue attenuation characteristics, the SAR

can be inferred. The SAR has been the currently accepted means of indicating RF

exposure levels and several organizations have established guidelines for permissible levels

of exposure to static and varying magnetic fields and to RF fields.


The International Non-Ionizing Radiation Committee of the International

Radiation Protection Association (INIRC-IRPA) is an international association providing

universal guidance on a variety of non-ionizing radiation standards (INIRC-IRPA 1991).

For static magnetic fields, a patient's cardiovascular function should be monitored during

exposures above 2T to the head or trunk and above 5T for limbs. For time-varying

magnetic fields, a patient's cardiovascular function should be monitored during exposures

above 6 T/s. Time-varying magnetic fields should not exceed 20 T/s. For RF fields,

exposures up to one hour should not exceed a total energy deposition of 120 W-min/kg

which equates to a SAR of 2 W/kg. To avoid overheating, the following product of time

and SAR should not exceed:

60 W-min/kg (1 W/kg for I hour) averaged over the head,

120 W-min/kg (2 W/kg for 1 hour) averaged over the trunk or

180 W-min/kg (3 W/kg over 1 hour) averaged over the extremities.

Instantaneous SAR should not exceed:

4 W/kg in the head,

8 W/kg in the trunk,

12 W/kg in the extremities or

10 W/kg in 0.01 kg or more than 10 minutes for the eyes.


The National Council on Radiation Protection and Measurements (NCRP)

Scientific Committee 39, Report No. 67, is entitled Radiofrequency Electromagnetic

Fields: Properties, Quantities and Units, Biophysical Interaction and Measurements

(NCRP 67). Report No. 67 is a primary source upon which American National Standards

Institute (ANSI) C95.1-1982 is based and it concluded there is little possibility of directly

measuring the absorption of energy by biological bodies at the cellular level. This

becomes important when dealing with hot spot generation which is the concentration of

energy into a small volume. Cell death, especially in the brain area, can cause irreversible


The 3T MRI system, for proton imaging, operates at 128 MHz limiting the

equivalent power density to one mW/cm2 (Table 1-1). These guidelines make no

provisions for near-field radiation. In far-field radiation, generally one wavelength from

the antenna, the electric and magnetic fields become well behaved and remain orthogonal

to each other and the direction of propagation. This permits power density calculations

using either electric or magnetic field measurements. In the near-field, close to the

antenna, the electric field and magnetic fields have not aligned with each other.

Additionally, there is a reactive field. This field is the inductive or capacitive storage of

energy by the antenna itself. It does not propagate far from the antenna but still can

interact with tissue within its influence. ANSI C95.1-1982 provides results from either the

electric field or the magnetic field and uses far-field assumptions to calculate SAR. The

3T MRI system operates only in the near field. The equivalent power density must then

become a function of both the electric and magnetic fields.

The International Electrotechnical Commission (IEC) provides extensive standards

for RF exposures. A complete listing of the IEC standards are provided in Appendix A.

The following requirements are applicable for room temperatures below 24 C and where

the relative humidity does not exceed 60%. The requirements regarding SAR from RF

power further assumes that the SAR averaged over any 10 second period does not exceed

5 times the stated temporal average SAR limit.

a. Whole-body SAR. Three operating modes with regard to whole-body

SAR are defined, namely Normal Operating Mode, First Level Controlled

Operating Mode and Second Level Controlled Operating Mode

(Figure 1-1).

i. The Normal Operating Mode comprises values of whole-body SAR

not higher than 1.5 W/kg averaged over any period of 15 minutes.

ii. The First Level Controlled Operating Mode comprises values of

whole-body SAR not higher than 4 W/kg averaged over any period

of 15 minutes.

iii. The Second Level Controlled Operating Mode comprises values of

whole-body SAR that may exceed 4 W/kg averaged over any

period of 15 min.

b. Head SAR. Two operating modes with regard to head SAR are defined,

namely Normal Operating Mode and Second Level Controlled Operating

Mode (Figure 1-2).

i. The Normal Operating Mode comprises values of head SAR not

higher than 3 W/kg averaged over any period of 10 minutes.

ii. The Second Level Controlled Operating Mode comprises values of

head SAR that may exceed 3 W/kg averaged over any period of 10


c. Local Tissue SAR With the Use of Special Coils. Two operating modes

with regard to local tissue SAR are defined, namely Normal Operating

Mode and Second Level Controlled Operating Mode.


i. The Normal Operating Mode comprises values of local tissue SAR

in any one gram of tissue not exceeding 8 W/kg in the head or torso

or 12 W/kg in the extremities averaged over any period of 5 min.

ii. The Second Level Controlled Operating Mode comprises values of

local tissue SAR that may exceed the upper limit for the Normal

Operating Mode.


The Food and Drug Administration (FDA) provides the approval for MRI systems

in the U.S. Currently, static magnetic fields up to 2T are permitted for clinical magnetic

resonance systems. Above that level, the system must show evidence of safety before

routine clinical use. Therefore the 3T system has not been approved by the FDA for

patient use but has been permitted as a research unit. The FDA provided three

alternatives for patient exposure to time-varying magnetic fields.

1. Demonstrate that the maximum change in magnetic field strength (dB/dt)

of the system is 6 T/s or less. This is 1/10 the threshold of peripheral nerve

stimulation and almost 1/100 the possible threshold for cardiac excitation.

2. A. Demonstrate that for axial gradients (Figure 1-3):

i. dB/dt < 20 T/s for T > 120 ps;

ii. dB/dt < 2400/(t(gs)) T/s for 12 gs < T < 120 ps or

iii. dB/dt < 200 T/s for T< 12 us.

B. Demonstrate that for transverse gradients, dB/dt is considered to be

below the level of concern when less than three times the above

limits for axial gradients.

3. Demonstrate with "valid scientific evidence" that the rate of change of

magnetic fields for the system are not sufficient to cause peripheral nerve

stimulation by an adequate margin of safety (which the FDA establishes as

at least a factor of three).

Exposure to RF energy, which the FDA considers below the level of concern, is a

SAR of

1. 0.4 W/kg or less averaged over the body;

2. 8.0 W/kg or less spatial peak in any I gram of tissue; or

3. 3.2 W/kg or less averaged over the head.

The FDA has stated that the exposure to RF energy must not increase the body's

core temperature by I C and that the temperature not exceed 39 C in the trunk and 40 C

in the extremities. Normal body temperatures range from 35.5 C to 40 C with an

averageACI"* issued a document on September 29, 1997 (Appendix B) titled

"Guidance for Magnetic Resonance Diagnostic Devices Criteria for Significant Risk

Investigations." This document was designed to provide guidance only and did not

constitute a regulatory change. This guidance allowed exposures to static magnetic fields

up to four Tesla. Exposures to RF energy were also modified to SARs of

1. 4 W/kg averaged over the whole body for any 15 minute period;

2. 3 W/kg averaged over the head for any 10 minute period; or

3. 8 W/kg in any gram of tissue in the head or torso or 12 W/kg in any gram

of tissue in the extremities for any 5 minute period.

For this research, which dealt with the head, the SAR limits were lowered from 3.2 W/kg

to 3 W/kg. The time limits, which were not previously a part of the guidance, limits

applications of brief periods of higher energy depositions. Quantitative standards for

gradient magnetic fields was removed and replaced with the non-definitive provision that

gradient fields should not "produce severe discomfort or painful nerve stimulation."


The ANSI standards only address SAR in the far-field region where the electric

and magnetic fields have become aligned with each other. These standards provide no real

guidance for the MRI system. The INIRC-IRPA, FDA and IEC standards address time-

varying magnet fields with the predominant reaction of concern being peripheral nerve

stimulation which can be most damaging when affecting the heart muscle. Each standard

provides a safety factor of 10 with the new FDA guideline removing quantitative values.

Most peripheral nerve stimulation was not observed or anticipated until changing magnetic

fields exceeded 60 T/s. The 3T MRI can switch magnetic fields at this rate. The IEC

standards addressed a uniform method of measuring power deposition. Two methods of

SAR measurement are available. One method is designed specifically for whole body

SAR. The other method measures the average power absorbed by an object of known

mass to provide a representative indicator of SAR.

Until ratification of the new FDA guidance, the FDA standards must still be met

before patient use of the 3T system. There was no guidance on any uniform testing

procedure available. It becomes the responsibility of the particular facility to provide

proof of compliance. Therefore the results obtained during this research may be used to

provide proof of the safety of the 3T MRI system.

Specific Absorption Rate

Of the three parameters of concern during MRI; static magnetic fields, gradient

magnetic fields and RF fields, the RF fields provide effects resulting in possible permanent

harm to patients. By definition, the specific absorption rate is "the time derivative of the

incremental energy (dW) absorbed by, or dissipated in an incremental mass (dm) contained

in a volume element (dV) of a given density." (NCRP 67). Mathematically this is

represented by

SAR d dW( dW W1
dt dm) dt p(dV) kg

Realistically the amount of energy deposited within a unit of tissue can not be

measured. However, other characteristics can be measured. Power deposition can be

inferred if the amount of energy provided to the target tissue were known and certain

characteristics of the target tissue were also known. With the power deposition defined


and tissue mass measured, the SAR can be defined over any period of time. This section

describes three methods for the development of SAR equations: (1) using loading factors,

(2) using Poynting vectors with far-field assumptions and (3) using Poynting vectors

without far-field assumptions.

Loading Factors

Loading factors provided RF deposition measurements by defining the loading

factor of an empty coil (Q.) and the loading factor of a coil loaded with the patient (Q1)

(Werner 1993). The patient's loading factor, QT, was then calculated.

1 1 1
QT Q1 Qo

The RF power PT deposited in tissue in the presence of a linearly polarized B, field became

2r S Q,

where w = 27tf was the angular frequency, L was the coil inductance and S was a coil

constant giving the magnetic field produced per unit current in the coil. The required

value of B1 depended on the desired flip angle a and the RF pulse length t (Roschmann

1987). If the RF pulses had a rectangular envelope, the required linearly polarized B, was


B, = 2a/y,T

where y, was the gyromagnetic ratio of the considered nucleus i (for protons yi = 42.58


The SAR was the result of the scan time and mass of the target.


For the phantom this was a valid assumption. For the human head this was not

valid but could still be safely assumed at lower field strength MRI system. The amount of

tissue heating becomes unique to each part of the brain as a function of its electrical

characteristics. Some portions absorb more energy and would display increased

temperatures. Overall, temperature increases would remain reasonable for lower field

strength MRIs and would not have caused any permanent damage. However, at 3T,

power deposition per unit mass encroached on the SAR limits. At the higher energies, the

generation of hot spots became an issue. The tissue material with high conductive

properties would absorb more RF energy while those with lower conductive properties

would absorb less. The total power deposition per unit mass, averaged over the entire

head, would meet FDA guidelines but provide no method to determine hot spots. Using

loading factors was not an acceptable method for safeguarding patient injuries but may,

however, still provide a first step in determining the range of energy deposition.

SAR Development with Far-Field Assumptions

Another method described for calculating SAR was the Poynting vector. The

Poynting vector theorem was developed to find the power in a uniform plane wave made

up of an electromagnetic field. An electromagnetic field consists of two parts, the electric

field and the magnetic field. The ANSI C95.1-1982 protection guides provide an

equivalent power density equation based on one of the measurable fields.

SAR = a IE, I2

where a = tissue conductivity

p = tissue density

Ei = electric field

The Poynting vector was defined as the cross product between the electric and

magnetic fields and interpreted as an instantaneous power density with units of watts per

area (W/m2).

P = E(I) xH() ExH()

Under the conditions of the SAR equations defined by NCRP Report 67, the target tissue

was assumed in the far-field. At the antenna, both the electric field and the magnetic field

exist independently. Both begin propagating outward. Since there was no immediate

relationship between the two fields, their phase angles were random. In the far field, with


a predictable 90 relationship, an intrinsic impedance, %, was defined as the square root of

the ratio of permeability to permittivity.

permeability 1
\ permittivity e

Additionally, and simply, in the far-field, the intrinsic impedance, as a perfect dielectric,

was defined as the ratio of the electric field to the magnetic field


Using this identity, and averaging the Poynting vector, the magnetic field component of

the Poynting vector was removed.

1 E2
2p -
2 Yj

The human body can be considered a lossy dielectric in which case the intrinsic

impedance became a complex value. The complex value was written in polar form defined

by a vector length and a vector angle representing the phase difference of the electric and

magnetic fields. As will be demonstrated later, certain assumptions were made to reduce

the complex nature of the intrinsic impedance.

= __ __ j
\ +je )t


Additionally, as the electric field progressed into the tissue, it attenuated as a function of

the tissue's skin depth, e*"6, where d denoted distance into the tissue and 6 denoted skin

depth. The human body, and the head in particular, have different tissue types at different

depths and the skin depths of these tissues vary. For the development of the SAR

equation for the phantom an assumption of a homogeneous tissue mass was used.

The Poynting vector was modified to

1 (Ee -/)2 1 E2e -2d/ 1 E2e -2 _
2 T] 2 lm /0 2 -m

The average power from a continuously emitting source applied to a tissue was derived

from the total power and the total area of concern.

I E 2e-2d/6
average power = P X area /-6 x A
2 r .

The polar form for average power implied a real and imaginary portion of power. It was

only the real portion, that defined the reactive aspect of power, that lead to energy

deposition in tissue. The Poynting vector reduced to

1 E2
average power = e 2d/ cos x A
2 i.

To further simplify the intrinsic impedance, NCRP Report 67 assumed the target

tissue met the definition of a good conductor. The larger the ratio of conduction current,


o, to displacement current, we, the better the conductor. In a good conductor, the electric

and magnetic fields remain close to 900 out of phase which then supports the previous

assumption used to remove the magnetic field component from the SAR equation. This

was validated by calculating the phase relation when a > we for a given tissue.

tan-f 900

Tissue high in water content met this requirement. This included muscles and both white

and gray brain matter and the tissue equivalent material used in the head phantom. For

adipose laden tissue and bone, with low water content, the tissue conductivity and

displacement current are closer. However, for the development of the SAR equation a >>

we and the following modification was applied to the intrinsic impedance.

T( 2j /.! /0cos(tan'(1))
So + je8 a 0 o6 6


This was derived from the propagation constant equation

g e

For a good conductor, ao/e >> 1 and

Y = jW e -j-- = j -joiaO

= cVio cos450 + j (- sin450

12 1

The real portion of this equation was the attenuation constant which defined the amount of

attenuation per distance traveled in the medium. This distance was based on 1/e

attenuation which, by definition, also became the foundation of skin depth. The

attenuation constant, therefore, was the inverse of the skin depth, 6.

6= (2
tOp a

Using only the real portion of the intrinsic impedance

T1 V os (450)-
08 08

The simplified intrinsic impedance was applied to the average power equation

average power = Ee -2d6 x A
The SAR was defined as the average power per mass of the target tissue.

SAR = average power

Substituting the definition of the average power

o_ E2e -2d/8 x A
SAR = 2
p x volume

To simplify the equation, the volume of interest was defined as an area, A, down

to one skin depth, 8. Additionally, the power calculation defined the incident power at the

skin surface (d = 0).

SAR = o E2

SAR Development Without Far-Field Assumptions

Two assumptions were made in order to derive the SAR equation. First, that the


target tissue resided in the far-field and second, that the target tissue met the requirement

of a good conductor. These assumptions are reviewed in the next two sections.

Far-Field Assumption. The far-field region begins at some finite distance from the antenna

and stretches to infinity. In the far-field, the electric field strength, magnetic field strength

and power density are orthogonal to each other and one parameter can define the others.

P = Ex H E 120 tH2
120 tr

(where P was expressed in W/m2, E in V/m and H in A/m)

The area from the antenna to the beginning of the far-field defines the near-field

which consists of the radiating field in addition to a reactive (nonradiating) field. This

energy can be capacitive or inductive and can affect the electric and magnetic field vectors.

The reactive region remains adjacent to the antenna and extends to about one wavelength

from the antenna. Like the far-field, the electric and magnetic field intensity decreases

with distance but unlike the far-field the radiation still lacks the plane-wave characteristics

of the far-field electric and magnetic fields. No relative assumptions are available to

characterize one field by the other. Measurement of both fields are required to accurately

define the SAR. For a 3T system operating near 128 MHz, one wavelength extends to

2.34 m which placed the target tissue well in the near-field region. Far-field assumptions

for SAR were invalid but unfortunately "...recommendations of RF exposure limits are

based on the experimental data for the far-field exposures and are usually formulated in


terms of the far-field parameters. Only general guidance regarding the SAR was given for

the near-field exposures," (Stuchley 1985).

Good Conductor. The second assumption made in the SAR development was that the

material imaged with the MRI was a good conductor. The intrinsic impedance, ir, was

simplified under the assumption that a >> we, or that the ratio of conduction current to

displacement current (o/oe) was much greater than unity.

For the human head, o/Ae for bone and fat were 0.47 and 0.52 respectively, which

do not approach the definition for a good conductor. Since these were the incident tissues

of the head, the good conductor assumption for intrinsic impedance became invalid.

Likewise, for the head phantom, the polyethylene casing had a low water content and had

a conduction current that was smaller than the displacement current, o/me = 0.0002. The

electric and magnetic fields, even if they had been orthogonal would have lost cohesion

after transcending the bone and fat layers of the head or the polyethylene casing of the

head phantom. Brain matter, having a higher water content than fat proved to be a better

conductor and fell under the guidelines of a conduction current greater than the

displacement current, o/e = 2.71. The tissue equivalent material was constructed mainly

of water which had a conduction current that was four times the displacement current,

o/ae = 4.

With the good conductor assumption, the skin depth was defined as

1 ___o
6 2


To accurately define the skin depth, regardless of the tissue's conduction property will

require recalculation of the propagation constant.

The propagation constant combined an attenuation constant with a phase constant.

The attenuation constant defined the loss of signal strength as a wave propagated through

a medium. The phase constant defined the angular shift of the wave per distance traveled.

The inverse of the attenuation constant yielded the skin depth. Mathematically, the

equation for the propagation constant was y = a + jP3 and the equation for skin depth was

8 = 1/a.

Differences Between Near-Field and Far-Field Assumptions. Power deposition estimation

from a theoretical basis in the near-field, as required by the MRI, is complex. No

correlation between the electric and magnetic fields exists allowing only assumptions to be

made. The assumptions required to formulate a generalized estimate eliminates the minor

imperfections that give rise to individual power deposition values for each individual

imaged (Mahony 1995, Bottomley 1985, Bottomley 1981, Boesiger 1992).

The skin depth was used for comparison between near-field and far-field

assumptions. Using far-field assumptions,

(a eRe (o = 0.00804

6 = = 0.5 m
1/2n x 128 x 106 x 47 x 10-7 x 8.04 x 10-3

Next the propagation constant, which contained no far-field assumptions, was used to

calculate skin depth.

Y = j(0 1 jj-

y = 127.5 /89.97'

6 = 127.5cos(89.970) = 0.248m

The far-field assumption tended to overestimate the skin depth of the polyethylene

outer husk by a factor of two. Therefore, using far-field assumptions would cause an

under-estimation of power deposition in "poor" conductors. For the head phantom, this

would include the out husk. For the human head this would include adipose tissue and

bone. Since both of these are near the outer surfaces, the under-estimation of power

deposition would be similar, still allowing use of the head phantom in human head power

deposition studies.

Whoe-Body SAR
Controlled Operating Mode




Time (min) 15
Figure 1-1. IEC Whole-Body SAR Guidelines.

Head SAR
Controlled Operating Mode


Time (min) 10
Figure 1-2. IEC Head SAR Guidelines.


dB/dt vs Time
200- \

T/s -


12 time (us) 120

Figure 1-3. FDA Axial
Gradient Magnetic Field

Table 1-1. ANSI C95.1-1982 Protection Guide for Radio-Frequency Electromagnetic

Frequency Equivalent power (Electrical field)2 (Magnetic field)2
range densityab
MHZ mW/cm2 V2/m2 A2/m2
0.3 3 100 4 X 105 2.5
3.0 30 900/f2 4 X 103 (900/f) 0.025 (900/f)
30-300 1.0 4 X 103 0.025
300-1500 f/300 4 X 103(f300) 0.025 (f/300)
1500- 100,000 5.0 2 X 104 0.125
a measured 5 cm or greater from any object in the field and averaged for any 0.1 h (6 min.).
b (electric field)2/1200i or 127c(magnetic field)2, whichever is greater.


Experiments were performed using invasive and non-invasive temperature

measurements at the VA/Shands 3T MRI Unit. Invasive measurements were made using

thermocouples and fiber optic cables. A thermocouple system was inexpensive but greatly

affected the results. A fiber optic system was expensive but immune to RF fields which

provided better results. Temperatures were measured non-invasively using a diffusion

MRI technique. All measurements were performed on a head phantom installed in the

General Electric (GE) Head Coil.

Head Phantom

The head phantom used through these experiments was created by Charles

Webster and Jim Scott using a plaster cast of a volunteer. A polyethylene covering was

form-fitted over the plaster cast which was then broken and removed. The outer husk was

constructed of 6.35 mm (1/4") polyethylene which had known electrical properties of

permittivity, intrinsic impedance and skin depth (Table 2-1). The polyethylene material

contained a low water content and was suspected of being a poor conductor.

The polyethylene mold was filled with tissue equivalent material (Table 2-2). The

tissue equivalent material was formulated by Jim Scott to provide similar loading

characteristics in the GE Head Coil as the human head (Olsen 1992, Chou 1984).

This material had a density of 1.01 g/cm3 and a total volume of 5250 ml.

Equipment was not available to determine the permittivity and conductivity of the tissue

equivalent material but certain assumptions allowed useable test results. Tissue, such as

brain matter, has a specific heat of 3.5 J/cm3'C. The tissue equivalent material has a

0.00461 to 1 ratio of NaCl to water. Using this value with known values of specific heat,

the tissue equivalent material specific heat was extrapolated (Table 2-3)(Leonard 1984).

Therefore, the tissue equivalent material should have thermal characteristics close to that

of the brain (Table 2-4).

Thermocouple Measurements

Temperature measurements in the tissue equivalent material were made using a

thermocouple connected to an Atkins Model #39641-T. The T-Type indicated the

thermocouple was made of copper-constantan. The junction of the copper and constantan

was exposed allowing for fast response times. The Atkins unit provided temperatures to

within 0.1C or 0. IF. Since 0.17F was a smaller unit, a slight increase in accuracy was

obtained from measurements made in Fahrenheit. Temperature measurements were later

converted to Centigrade.

A 6.35 mm (1/4") hole had been drilled into the location of the right eye of the

head phantom and used as the insertion point for temperature reading devices. Originally,

the thermocouple was inserted into the phantom and left in the phantom during the scan.

Interference from the RF signals distorted the readings on the Atkins unit so a low-pass

filter was installed. The calibration of the Atkins unit was invalidated with the addition of

the low-pass filter. The unit was then disconnected from the thermocouple during the

scan. A pre-existing phantom protocol had been programmed into the Advanced NMR

(ANMR) console. This protocol was modified to provide a higher range of output power.

The modified protocol is listed in Appendix C.

A 48 minute scan at 18 watts average power (as indicated on the Omega process

indicator) exhibited a 3.9 C rise in temperature using the Atkins thermocouple. The

temperature immediately dropped an average of 1.1 *C within two minutes of the end of

the scan. Further testing revealed that the temperature of the tissue equivalent material

was raised by the presence of the thermocouple cable during the scan. This invalidated


The experimental method was adjusted. Prior to the scan the thermocouple was

inserted to one of three depths: the top surface, middle of the tissue equivalent material or

the back of the head phantom (Figure 2-1). The temperature was read for approximately

two minutes or until the reading on the Atkins unit stabilized. After the initial temperature

was determined the thermocouple was removed and the head phantom scanned for a

predetermined time using the modified Phantom Protocol. At the conclusion of the scan,

the phantom was removed from the MRI unit and the thermocouple reinserted.

Temperature measurements were generally obtained within 30 seconds of termination of

the scan and were subsequently recorded at 0.1 F intervals at one minute intervals. To

improve results, fiber optic cables with ten meter extensions were purchased.

Fiber Optic Measurements

Three fiber optic probes were purchased to replace the thermocouple cables. The

Fluoroptic Thermometer used a patented fiberoptic probe manufactured by Luxtron.

The tips of the fiber optic cables use the fluorescence of phosphors to sense temperatures.

The probes are immune to electromagnetic interference and can operate with the -200 C

to 450 C range. Since the fiber optic probes were immune to RF effects, the probes

remained in the phantom during the scan. The probes were linearly positioned

approximately four centimeters from each other. The exact location of the probes was

determined for each experiment through imaging the head phantom. Using the landmark

as a common reference. The hole at the right eye was plotted sagitally and axially. The tip

of each probe was then plotted. To account for any rotation of the head phantom between

testing dates, the outer periphery was plotted.

Using the drill hole location and the outer bounds of the head phantom, the

distance from the surface to the probe head was calculated. This distance was defined in

millimeters and represented the "concentric ring" (Figure 2-2).

Once the locations of the probes were determined, the temperature measurements

were recorded. Some experiments were performed on weekends allowing the head

phantom twenty-four hours to come into thermal equilibrium with the room which ranged

form 17.2 to 17.8 C. Even allowing forty-eight hours for the head phantom to come into

equilibrium, the head phantom would still have differing internal temperatures. On the

average, the temperature varied 0.2 *C between points within the head phantom. To

compensate for variations between probes and starting temperatures, each probe was

evaluated independently. Equations were derived using absolute temperature variations


Heating Protocol

A head phantom imaging protocol had been previously programmed into the

ANMR console. Appendix D lists the program's parameters. The output power was

adjusted by increasing or decreasing the Transmission Repetition (TR) value. The TR was

the length of time between subsequent starts of a pulse sequence. The phantom protocol

used a spin echo which had a 90 pulse followed by a 1800 pulse (Figure 2-3). The TR

was the length of time between subsequent 900 pulses.

By decreasing the TR value, the pulses cycled in faster succession which increased

the average power being transmitted. The majority of experiments were conducted at an

average power of 18 watts as indicated on the Omega. The head phantom was placed into

the GE Head Coil in the same orientation as a patient. The "landmark" of the imager was

set on the bridge of the nose of the head phantom. This represented the iso-center of the

head phantom and represented the spatial location of x=0, y=0 and z=0 of the imaging

chamber. The patient carriage containing the GE Head Coil and head phantom was then

sent into the imaging chamber. The Phantom Protocol was selected on the ANMR

console and the TR adjusted to obtain a desired average power deposition. A 8 mm

camcorder was positioned to record the fiber optic probe results, the average power as

indicated on the Omega and the current time. The camcorder images were used in post

imaging analysis to enter time, average power and temperature into a database program.

Thermal Imaging

MRI affects tissue by depositing thermal energy. This thermal energy can alter

some parameters measurable by MRI. Temperature resolution using MRI was first

applied to imaging hypothermia patients. For hypothermia patients, MRI had the dual

benefit of providing heat directly to the body core and imaging the temperature changes

within the body. Unwanted heat deposition was also the problem for general imaging.

The MRI system can be self-regulating by monitoring heat deposition during a scan.

Imaging relies on the detection and quantification of signals from the tissue. Thermal

imaging described in literature required a reference thermal image (Delannoy 1991, Hall

1990, Schwarzbauer 1995). The temperature was then defined for this reference image.

This method required extensive calibration within each tissue type. Establishing a

reference temperature had an inherent accuracy problem which propagated into a larger

error during subsequent imaging. To reduce this error, the first diffusion image only

defined the baseline. Regardless of the starting temperature, any differences in subsequent

images would provide a difference in temperature from the reference image. Quantifying

temperature changes in this method should reduce errors and increase accuracies. There

were at least three characteristics of protons that lent themselves to measuring changes in

temperature: TI relaxation time, diffusion and proton chemical shift.

Tl Relaxation Time Imaging

The TI relaxation time, also known as the Spin Lattice or Longitudinal Relaxation

Time, was defined as the characteristic time constant for the return of the proton's

longitudinal axis to alignment with Bo, the main magnetic field. A Tl map of the target

tissue can be generated by an RF pulse of sufficient frequency and strength, acting in a

direction 900 from B0 (Figure 2-4). Shutting off the transmitter and using the transmit

antenna as a receiver, the signal strength was a function of the quantity of the nuclear

spins realigning with Bo and provided the T1 map. After absorption of the RF energy, the

absorbed energy dissipated into the surrounding molecules at an exponential rate. After

one TI period, 63% of the protons are realigned with B0, also known as their rest value.

After two TI periods, 86% have realigned and after three T1 periods, 95% have realigned

(Figure 2-5). As temperature increased, the relaxation time decreased (Nelson 1987).

Normally the TI for protons in water was two seconds.

For thermal imaging, a reference image would be taken at a known temperature.

After tissue heating, another Tl image was taken and the images compared. The

difference in TI times defined the temperature change. One aspect of thermal imaging

was no change in temperature of the cerebral spinal fluid and blood from short, localized

heating tests. The fluids present during a slice selective RF pulse would have moved to

another physical location by the time a second image was taken. The replacing fluids, not

having absorbed any RF energy, would already be at their rest value. The temperature

would appear to remain constant. Early in the development of TI mapping a sensitivity of

about 2 C was obtained (Parker 1983). Since the FDA standards that the MRI system

must abide by restricted temperature increases to only 1 C, this method would prove

unuseable. The sensitivity was directly related to the scan time which lasted five minutes.

During this scan period, the entire tissue continued to absorb RF energy which in turn

continued to increase the temperature. The solution to making TI imaging a viable

thermal indicator was to shorten the scan time.

Using an inversion-recovery Snapshot FLASH MRI technique reduced acquisition

times to approximately two seconds (Deichmann 1992, Nekolla 1992). An inversion

recovery technique began with a 180 pulse which inverted the magnetization and flipped

the proton axial spin opposite to Bo. After a time period, t, a 90 pulse would be applied

(Figure 2-6). Transverse magnetization would be detected (Figure 2-7). The inversion-

recovery Snapshot FLASH MRI technique took images within milliseconds of the RF

pulse by using smaller flip angles. Since smaller flip angles require less RF energy,

temperature increases were minimal. One sacrifice for this rapid measurement was a

reduction of signal-to-noise. This can be compensated for by averaging several images in

quick succession. The precision of the Tl measurement had improved to better than three

percent with this method (Schwarzbauer 1995). A three percent precision at a normal

body temperature of 37 *C was 1.1 C. This method would still not satisfy FDA

requirements since a reading of 0 C change could in reality exceed the 1 C limit.

In addition, temperature changes were not the only factors affecting the Tl

relaxation times. The presence of free radicals and water concentration also affected TI

(Knuttel 1986). These factors became a concern when their concentrations changed from

the time the reference set of images were taken until the time the final TI images were

taken. A larger quantity of protons in the second Tl image would increase the apparent

percentage of realignment of nuclear spins with Bo. This increased quantity, and

subsequent apparent increased realignment would promote a misconceived perception of


increased tissue temperature. The MRI system only generates non-ionizing radiation, so it

was unlikely the generator of it created additional free radicals during the imaging process.

Since the free radicals do possess a polarity which was subject to magnetic fields, only if

the free radicals were uniformly distributed, whereby the movement of radicals out of a

particular location matched the movement of radicals into the particular location, could

the effects of free radicals be ignored.

Two additional factors that affected the T1 relaxation times were perfusion and

diffusion. Perfusion was the pseudo-random flow at low velocities of blood moving along

finely divided structures of the capillaries. Diffusion was the Brownian motion of

individual molecules moving with large random thermal velocities. Brownian motion was

the constant erratic movement of particles being struck by other particles. Temperature

changes affected the rate of Brownian motion by affecting the viscosity and diffusion of

water in tissue.

Diffusion Imaging

In biological tissue, a fraction of the water movement in capillaries was affected by

diffusion (LeBihan 1988). Water flowing down a linear path has an overall direction

down the path, but upon closer attention, the water molecules collide and deflect off each

other creating a random motion. This interaction was calculated by differentiating the

Stokes-Einstein relation between viscosity and diffusion.

dD dT-
D kT) T

where k = Boltzmann constant,

E, = the activation energy for translational molecular diffusion,

D = the translational molecular diffusion and

T = temperature.

A temperature change map can be derived from two diffusion images (LeBihan


kT (D -Do)
T To = -

An initial reference diffusion image, Do, was required as well as the initial temperature.

Using Ea = 0.21 eV, and applying a 1 C increase in temperature, the diffusion coefficient

increased by 2.4%. This value was dependent on an accurate E,. For water the activation

energy remained relatively constant at 0.21 eV. For tissue, this value may be different and

should be determined for each tissue type present in the image. If this activation energy

varied as a function of either TI and/or T2, an image weighted in favor of one or the other

would be required. The intensity of the resultant image would suffice to assign E, values

to each tissue type. Then a pre- and post-exam diffusion image sequence would map the

temperature changes.

Both perfusion and diffusion images can be obtained using the Intravoxel

Incoherent Motion (IVIM) method. The velocities of molecules in a particular tissue type

were directly related to the kinetic energy of the molecules. This kinetic energy, in the

form of thermal energy can be imaged using a spin-echo sequence followed by a second

sequence with increased gradient pulses. The increased gradient pulses increase the

effects of the perfusion and diffusion. Since the pulse sequences are identical, except for

the gradient pulses, Tl and T2 effects were negated. This leaves only an apparent

diffusion coefficient (ADC) which was a combination of both perfusion and diffusion. A

third sequence with an even stronger gradient pulse greatly reduced perfusion while only

moderately reducing diffusion effects. A comparison between the second and third

sequences defined a diffusion image. A comparison of the diffusion image with the ADC

image from the first two sequences produced a perfusion image. Temperature changes do

not greatly affect perfusion. Perfusion, the blood flow through capillaries, may increase

from increased cardiac action caused by the deposition of heat elsewhere in the body. The

body's response to thermal changes through a long imaging session were possible but

would only reflect an overall temperature increase throughout the body. Appendix E,

"Heat Stress Index", described this effect in greater detail. Localized heating, which was a

consideration for imaging at 3T, would not be accurately represented by perfusion


The IVIM method was sensitive to motion artifacts. For head imaging, cardiac

gating would reduce artifacts from blood flow, and to a lesser degree, cerebral spinal fluid

(CSF) flow. Eye movement artifacts would be reduced by preplanning the direction of the

phase-encoding gradient pulses. Increased IVIM images using faster techniques would

also reduce motion artifacts. The brain was a non-moving organ and thermal imaging

resolution would be possible as long as patient movement was reduced. The eye, as a


result of its limited vascularization, was the tissue most susceptible to thermal damage and

the most difficult to image using the IVIM method. An alternative would be to use echo-

planar imaging.

Echo-planar imaging used a gradient echo technique consisting of a rapid sequence

of phase steps between each repeated pulse sequence. The echo-planar imaging provides

rapid images, from 15 to 30 images a second. This produced virtually real-time images for

functional studies of the brain. During brain activity there is a rapid, momentary increase

in the blood flow to the active area of the brain. Therefore, echo-planar imaging allows

mapping of brain activity. Echo-planar imaging also becomes a benefit to limit motion

artifacts by imaging in a fraction of a second. Since the echo-planar imaging gradients

follow in rapid succession, the signal-to-noise decreased. However, the large quantity of

data points from the multiple sequences improved the estimation of the perfusion and

diffusion (Turner 1990). With decreased imaging times, diffusion imaging without regard

for perfusion artifacts may also become possible.

Proton Chemical Shift Imaging

A third method of non-invasive temperature measurements used chemical shift

imaging. Fat protons and water protons have differing chemical shifts when imaged in a

MRI. Generally, chemical shift imaging was used as fat suppression imaging. The pure

water proton chemical shift was proportional to a change in temperature by approximately

-0.01 ppm/C (Ishihara 1995). This temperature dependence resulted from the stretching,

rupturing and/or bending of the hydrogen bonds. Temperature changes were difficult to


detect using the chemical shift because of the small shift. An accuracy of I C was usually

expected which was at the FDA allowable threshold of 1 C change in tissue temperature.

This meant that a reading of 1 C from the chemical shift could actually be from 0 C to 2

C. The chemical shift was detected by measuring the time between the 90* pulse and

1800 pulse and comparing this to the difference in time between the 180 pulse and the

spin echo (Figure 2-3). The time difference between sets of images was the result of


Chemical shift imaging used the echo planar imaging technique to help reduce

motion artifacts from diffusion-weighted techniques. However, the echo planar imaging

had a strong sensitivity to magnetic susceptibility. Even though human tissue was mostly

not affected by the magnetic field, at 3T even small differences in the electrical properties

of tissue caused a dephasing of spins and frequency shifts. This made the chemical shift of

a particular tissue weighted by the concentration of KCI and NaCI in the tissue. To use an

absolute temperature difference between the same tissue over a heating period, the

concentration of electrolytes would have to be assumed constant over a temperature

range. The chemical shift of protons was also affected by the pH of the tissue. Since the

MRI was non-ionizing, hydrogen ions were not created, but since they are polarized they

may be moved from a region of high density to a region of low density. Again, for

absolute temperature measurements between two time frames, the pH should not change

rapidly and the reference set of images, if taken just prior to the exam, would still provide

a good baseline for comparison of post-exam chemical shift images.

Phantom Thermal Imaging

Of the three thermal imaging techniques described above, diffusion imaging using

echo planar imaging, appeared the best alternative. Neither Tl nor chemical shift imaging

was solely affected by temperature and both were affected by diffusion. This would

require techniques to eliminate diffusion effects. The best method to do that was from

two sets of diffusion weighted images to remove the diffusion affects. This, in effect,

relied on running diffusion techniques in addition to other techniques to image T1 or the

chemical shift as a function of temperature. Since diffusion changes from one image to

another was solely affected by temperature, using just diffusion techniques reduced

imaging time thereby decreasing the effect from RF deposition in tissue.

The benefit of using diffusion imaging was three-fold. First, with less techniques,

less RF energy deposition took place during temperature measurements. Second, with

shorter imaging techniques, results would be easier to post process. Third, while not

applicable to phantom studies, with shorter imaging techniques the effects of patient

motion would be reduced.

Figure 2-1. Thermocouple
Probe Locations.

Figure 2-2. Concentric Rings Used to
Define Location in Head Phantom.



90 90

Fiu TE--i

Figure 2-3. Spin Echo Sequence.TR

Figure 2-3. Spin Echo Sequence.

Figure 2-4. 90 RF Pulse.



0 T1 T2 T3T4

Figure 2-5. Ti Recovery Time.


-r >

Figure 2-6. Inversion Recovery Pulse.


Percent Brain
wkh Bo CSF


Figure 2-7. Water and Fat
Discrimination Using Tl.

Table 2-1. Electrical Properties For Polyethylene Husk of Head Phantom.

Loss Tangent Permittivity Intrinsic Impedance Skin Depth

a (U-s_ F 1(() 6(m)

0.0005 2.26 27.9 0.2 0.248

Table 2-2. Ingredients to Create Tissue Equivalent Material.
7.5% TX-51 290g/gal. Water
0.45% NaCl 17.44g/gal. Water
0.07 mole Gadopentetate Dimiglumine 0.544 ml/gal. Water

Table 2-3. Thermal Properties of Tissue Equivalent Material in Head Phantom.
NaCI Ratio Thermal Conductivity Specific Heat
(W/mOK) (J/cm3K)
0.000765 0.674 0.025 3.52
0.00306 0.646 0.097 3.53
0.00461 0.629 3.54

Table 2-4. Electrical Properties of Tissue Equivalent Material in Head Phantom.
Loss Tangent Permittivity Intrinsic Impedance Skin Depth

a (-s\) R() F() 6(m)

2.71 61 34.3 326 0.06


Three sets of results were obtained from this project. The first set was

thermocouple results. These results were recorded just prior to heating with the MRI

system and then immediately after the scan. The thermocouple wire generated

interference and increased temperatures when left in the head phantom during the scan.

The second set of results were from the fiber optic cables which were immune to RF fields

and provided information during the heating of the head phantom. The third set of data

was from a diffusion imaging technique. Previous thermal data from invasive temperature

measurements determined the diffusion technique could provide accurate temperature


Thermocouple Results

All plots from experiments with thermocouple probes displayed a combination of two

curves (Figure 3-1). The predominant curve during the first minutes resulted from the

creation of a "depletion zone." After insertion the thermocouple wire became a heat sink

for the surrounding tissue equivalent material and accelerated the heat loss for the first

two to four minutes. Using the plots from the first four minutes, an equation was derived

of this initial cooling. Since the first recorded temperature was approximately thirty

seconds after the scan ended and a few seconds after the thermocouple was inserted, the

initial temperature was calculated from the assumption that the cooling followed an

exponentially decaying function over time with the following general format.

F, F.e

Fo is the unknown initial temperature, FI is the known temperature approximately thirty

seconds after the scan, t, is the thirty seconds of thermal decay time and is the unknown

thermal decay constant.

From data gathered after the initial four minutes, a second exponential curve was

derived which defined the expected cooling curve of the phantom. This cooling curve was

then adjusted to compensate for the initial heat loss provided by the thermocouple wire.

This method depended on the validity of the thermocouple cable acting as a heat sink.

Temperature measurements were made just below the top surface of the

phantom's husk and through the tissue equivalent material to the back of the head

phantom. There was more fluctuation in temperature measurements at the top surface of

the phantom which was exposed to the air. The back of the head phantom was placed on

a cushion which provided insulation from heat loss. Temperature measurements from this

region provided stable results (Table 3-1). The heating of the phantom from RF

transmissions were compiled and a heating equation, as a function of the average power

read off the Omega process indicator and the scan duration, was derived.

Temp. Inc. (C) = ((4E-4 X Power2 (1/9)E-4 X Power3) X Scan TimeX5/9)


This equation provided close approximations to actual thermocouple results up to

thirty watts average power. Complications were found when measuring temperatures in

the middle of the head phantom. After the scan was complete and the thermocouple

installed, temperatures continued to rise. This indicated that thermal movement would

become an issue of concern when mapping temperatures. By suggestion this also

indicated the RF energy did not penetrate the entire phantom during a scan. If it had, then

core temperatures would have matched the outer surfaces. This demonstrated the

phenomenon of "skin effect." At high frequencies, a current carrying conductor maintains

a flux on the outer edges only. This theory is used by the power industry which allows

them to use high tensile strength, but low current carrying potential, material to string long

power lines. Only the outer sheath requires current conductor properties. This same

effect was demonstrated during thermal testing of the head phantom. With this

complication and relying on the subjective interpretation of the heat sink properties of the

thermocouples, accurate results with thermocouples were suspect.

Fiber Optic Results

Three fiber optic probes were simultaneously inserted into the head phantom. The

fiber optic probes provided temperature measurements during the scan to better

characterize the heating of the head phantom. Appendix F contains the raw data and

graphs derived from phantom testing using the fiber optic probes. The average power

measurements were obtained from the Omega at one minute intervals. The shaded

portions of the graphs represent the period the imaging was performed.

Heating Equation

An exponential function was assumed for the heating of the tissue.

T = T 1 e 'It

Two assumptions were used in determining this function. First, the tissue equivalent

material, similar to human tissue, would not have an unlimited capability to absorb thermal

energy. At some point the tissue would stop increasing in temperature as energy was

applied. This implied some upper threshold, T.. Since both the tissue equivalent

material and human brain contain mostly water, this upper threshold would be similar and

would also be far above the pain threshold of an individual. Therefore, the determination

of this threshold was not significant to this research. Only the assumption that this

threshold was above the current operating temperatures was required. Second,

temperature increases as a function of power would not be linear as they approached the

threshold. As the temperature increases there would be a constant decrease in the amount

and rate that additional energy could increase temperatures. This implies an exponential

function. Since the function changes exponentially, it was assumed that there would be a

thermal half-life, t.. Experimental data would be used to define these parameters.

The thermal data results were measured using three fiber optic probes and the

location of each probe tip was determined using the scanned image. Table 3-2, 3-3 and 3-

4 list the locations of the probe tips in relation to the outer edge of the head phantom and

to the eye hole drilled into the head phantom. There was no probe #1 cable used during

the testing. Appendix F lists the results of the temperature readings for the first thirty

minutes of the experiment. For the majority of the experimental scans, the thirty minute

period included the entire scanning time of twenty-two minutes and several minutes


There were four effects apparent in the data. Three of the effects dealt with a

delay time which was the time, in minutes, since the beginning of the scan until there was a

noticeable increase in temperature.

1. The delay time was a function of distance from the outer surface. This was

intuitive once the other effects were defined and validates some of the

thermocouple results. The electric field attenuated as it passed through the tissue

equivalent material. This attenuation deposited energy in the outer layer thereby

heating it first. The remaining energy, deposited farther into the tissue, heated the

tissue equivalent material at a lower rate.

2. The temperature gradient within the head phantom caused a delay by "passing"

thermal energy to the lower temperature tissue equivalent material. Temperatures

varied 0.5 *C within one centimeter of the probes at the start of Test #3. The plots

showed a slower increase in temperature where the tissue was warmer. This

demonstrated an increased thermal conductivity of the warmer tissue equivalent

material. The results from Test #9 also demonstrated this pattern. There the

temperature gradient between probes was 0.8 *C over two centimeters. The higher

temperature zone increased in temperature at a slower rate than the inner core

even though it was closer to the skin and should have heated first.

3. The lower the starting temperature of the head phantom the sooner a temperature

increase was registered. If the head phantom had sufficient time to remove thermal

energy then the delay time in seeing a response from a scan of 18 watts average

power was one minute per centimeter depth. The higher the starting temperature,

the delay time increased by twenty-five percent for every one degree centigrade. A

comparison of the experimental results from Test #2, which had a higher starting

temperature, to that of Test #1, where the temperatures were stable, demonstrated

this point. With just under a one degree centigrade change in starting

temperatures, delay times in registering a temperature rise increased from ten

minutes to twelve minutes at a depth of 74 mm.

4. After the completion of the scan, temperatures at the inner core continued to rise.

Test #1 showed a continued increase in temperatures after completion of the scan

at 62 mm and 73 mm depths. The temperature rise at the 41 mm depth continued

to rise for approximately 5 minutes before stabilizing. This reinforced the concept

that the electric field was not uniformly distributed throughout the head phantom

and that the majority of the thermal energy was deposited along the outer edges.

Several experiments were run at lower average power. These scans generated less

thermal energy and, as expected, the rate of temperature increase was less. Since the

temperature increase appeared as an exponential function, a "thermal half-life" of

temperature increase was derived. From test results, this value appeared as a function of

the average power (Table 3-5). These values are subjective and derived only from the

appearance of the curve. A plot of the natural logs of the half-life against the average

power allowed a slight modification of the numbers to create an exponential function

relating average power to half-life (Figure 3-2).

/ = e 6.375 0.134 (avg power)

Using this relationship between thermal half-life and average power reading, as displayed

on the Omega, a heating equation was defined.

1 -0.693 x t
T = To ( 1 e (6370. ))

Cooling Equations

Cooling equations would provide additional information on the thermal

characteristics of the head phantom tissue equivalent material. Ideally the tissue

equivalent material would begin cooling at the end of the scan. This, however, did not

take into account the thermal movement within the tissue. Thermal measurements taken

with fiber optic cables (Figure 3-3 through 3-8) show temperatures taken after the scan

was complete and no further energy was deposited into the head phantom. The tissue

equivalent material closer to the outer edge of the head phantom began reducing

temperatures. The tissue equivalent material closer to the core continued to increase in


The increased temperatures made creating a cooling curve impossible. Some

useful information was still obtained from these results. Thermal energy traveled in the

direction of cooler tissue equivalent material. This direction was towards the center of the

head phantom. This implied that the polyethylene outer husk of the phantom provided

insulation to the tissue equivalent material and minimal heat transfer to the atmosphere.

This was also demonstrated when temperatures had not stabilized within the head

phantom after twenty-four hours of sitting at room temperature.

Diffusion Imaging

Diffusion imaging was preferred over other imaging techniques since the only

parameter that affected it was temperature. Other imaging techniques, such as T2-

weighted and perfusion, also required a diffusion image technique to remove the diffusion

effects from the image. The diffusion technique was run on the head phantom prior to

heating. After the diffusion technique was run and images captured, an imaging technique

was performed on the head phantom that generated approximately 18 watts, as indicated

on the Omega, for approximately twenty-two minutes. Immediately after the head

phantom was heated, another diffusion technique was performed. A comparison of the

two spatially different diffusion techniques would indicate changes as a result of

temperature alone.


To capture the diffusion images required the InstaScan computer system

connected to the ANMR unit. Figure 3-9 displays the connections between the InstaScan

computer, the ANMR Console, the uf3T Sun Workstation and the Windows-based

computer. Also listed are the program names used to transfer and manipulate images.

Prior to diffusion imaging, "ghosts" needed to be tuned from the system. Ghosts were

unwanted artifacts of the original image. Appendix G lists the steps performed on the

InstaScan computer and the ANMR console for ghost tuning.

Diffusion Imaging. Once the ghost images were attenuated, the diffusion technique was

run. Appendix H lists the steps performed on the InstaScan computer and the ANMR

console for diffusion imaging. Appendix I lists the parameters for diffusion imaging on the

ANMR console. The diffusion technique provided seven images. The first image was a

T2-weighted image. The next three images were diffusion images, at the slice location,

for the three vectors X, Y and Z. The final three images were another set of diffusion

images with increased gradients. For phantom studies the second set of images were not

required. The diffusion technique actually provides both perfusion and diffusion which

was described earlier as the ADC. Perfusion is blood flow through small capillaries and a

second diffusion image with increased gradients reduces the perfusion effects more than

the diffusion effects. Since the phantom did not have capillaries, there were no effects

from perfusion. However, for this diffusion technique to be effective on humans, a second

diffusion technique will be required. If desired, the second set of images can be subtracted

from the first set of images to provide images that predominately contain perfusion effects

and a small portion of diffusion effects.

Image Conversion. The diffusion images were physically located on the InstaScan

computer. In order to manipulate them, they were first moved back to the ANMR console

by selecting the images from the database on the InstaScan computer and selecting

TRANSFER. From this location they were retrieved from the uf3T Sun Workstation.

Appendix J lists the steps to transfer the images from the ANMR console to the Sun

Workstation. Once on the Sun Workstation the files were converted to another format

that was transferred to a Microsoft Windows-based computer for further processing.

Image Manipulation. Appendix K lists the program written by the author for this research.

The program was written for a Windows 95 platform and was designed to take a bitmap

image and digitize each pixel into a long integer indicating its color. The location of the

pixel, given in X and Y coordinates, and the number associated with its color, was saved

in a Microsoft Access database. Once the images were saved into the database, the

program would take the images from the first diffusion run and subtract the images from

the second diffusion run. The result represented the change in the two images. Since a

change in diffusion images was only affected by temperature, the values from the program

characterized temperature changes.
Figure 3-10 is a display of the entire set of images. The first row shows the seven

images from the diffusion imaging performed prior to heating the head phantom. The

second row shows the seven images from the diffusion imaging performed after a twenty-

two minute spin echo sequence generating approximately 18 watts as indicated on the

Omega. The third row shows the comparison of the two images. Since the image transfer

process involved several steps, some noise was expected from signal degradation in the

final images. As a check for undesired degradation of the images, the second row of

images was saved twice from the InstaScan computer. This was accomplished by

transferring the same set of images twice back over to the ANMR system. Each set was

saved as the next sequential series number. At that point the second, duplicate set of

diffusion images were processed identically as the original. The images were transferred

to bitmaps and digitized. The two sets of identical images were then subtracted from each

other and displayed in the fourth row.

Image Analysis. The images in the first row of Figure 3-10 were provided by the diffusion

technique taken prior to heating the head phantom (Figure 3-11). The same heating

technique that had been used for fiber optic tests was applied to the head phantom to

increase its temperature. This technique applied eighteen watts average for twenty-two

minutes. The diffusion technique was run for a second time and the results displayed in

row two of Figure 3-10. The first diffusion image (the second image in the series) is

displayed in Figure 3-12. The two images were digitized and a number was assigned to

the color of each pixel. The color values derived for each pixel of each image were

subtracted to provide a method to quantify the change in diffusion. This was validated by

test results beginning with the Atkins thermocouple unit and later with the fiber optic

system. The fiber optic probes indicated that the temperature increased along the outer

rim of the head phantom before temperatures would increase deeper in the head phantom.

This would translate into an initially greater change in diffusion along the outer portions of

the head phantom but that by the end of the scan temperatures would be similar and the

overall diffusion image would be more uniform.

The numerical value which defines each pixel in the image is a long integer

representation of its color. The numerical format consisted of six hex digits representing

the colors red, green and blue with each color having a magnitude of 0 to FF (hex). For

instance, a pure red color would have the hex number 000OFF, a pure green color would

have the hex number OOFFOO and a pure blue color would have the hex number FF0000.

Since the images were displayed on a gray scale, the values for red, green and blue were

identical (i.e. white equals 000000, light gray equals 111111, dark gray equals AAAAAA

and black equals FFFFFF). This simplified numerical interpretation by allowing

subtraction of only one color which had a range of 0 to 255. Changes in these numbers

would indicate the magnitude of change in diffusion, and thereby, change in temperature.

Figures 3-13 through 3-15 represent T2-weighted images that accompanied the

diffusion image technique. The images had noise attached to the images as they were

processed. This noise transferred as darker pixels when converted to mathematical values.

The noise was later reduced by programming a filter to remove numerical values below 50

(on a scale of 0 to 255). Relevant portions of the scan had image values above 50. The

filter value was derived experimentally through trial and error. Also present on the image

was a horizontal line. The line represented the location of the displayed histogram.

The top histogram (Image 002 window) was taken from the pre-heating diffusion

image and the middle histogram (Image 004 window) was taken from the post-heating

diffusion image. The scale of the window was from 0 to 255. These two images appeared


similar so a composite image was created. The histogram from the post-heating diffusion

image was subtracted from the pre-heating diffusion image.

The third histogram is the Composite window which was scaled vertically from

100 to -100. Image composites where the post-heating images were less intense than the

pre-heating images would have the histogram generated upwards. The downward

generated histogram indicated that the intensity of the pixel increased in intensity, which

translated into an increased numerical value when digitized by the computer, as

temperatures increased. Three slices, as indicated by the horizontal line on the right side

image of Figures 3-13 through 3-24, were taken from each axis of the diffusion image.

Slices were taken from the top, which represented the eye and nose area of the head

phantom, the middle and the lower or back portion of the head phantom.

The Composite images from all subtracted diffusion results displayed histograms in

the negative direction. This indicated the intensity of the images increased with increased

temperature. Had the Composite histogram been random, where equal numbers of points

would have been positive and negative, then this diffusion imaging technique to measure

temperature changes could be considered non-effective. The histograms demonstrated

diffusion imaging does provide a means of measure temperature changes.

The diffusion images represent diffusion changes in three orthogonal vectors. To

quantify one numerical results for each pixel, a vector sum of the three diffusion images

was obtained. For this research, the average intensity of all the pixels was calculated for

each vector. The average was obtained from adding the intensity differences of the three

slices taken from each axis. The sum of the Composite images from Figure 3-16 through

3-18 provided the average intensity change for the X-axis. The sum of the Composite

images from Figure 3-19 through 3-21 provided the average intensity change for the Y-

axis. The sum of the Composite images from Figures 3-22 through 3-24 provided the

average intensity change for the Z-axis. These averages were summed using the following


Pixel Intensity = X2 + y2 + Z2

The average change in intensity of the composite images displayed in Figures 3-16 through

3-24 was 40.7 on a scale of 0 to 255. The imaging technique used for heating the head

phantom generated sufficient thermal energy to increase temperatures by approximately

0.5 C. These results indicated temperature changes less than 1 *C can be measured.

Phantom Cooling Curve


Figure 3-1. Cooling Curve Using Thermocouple Probes.

Thermal Half-life


Figure 3-2. Natural Log of Thermal Half-Life.

4 6 8 10 12 14 16 18 20
Average Power (W)

Cooling Curve
(Test #4 Ring = 64)


20.1 -.----------

0 10 20 30 40 50 60 70
Time After End of Scan (min)
Temperature = -0.00524 t + 20 72

Figure 3-3. Test #4 Post Scan Temperature Measurements.

Cooling Curve
(Test #4 Ring = 30)

20.9 -.. .. ..- --- .- .
20.8 -
20.7- i

S20.5 -j --

20. --- --+----4--\--------I ----

0 10 20 30 40 50 60 70
Time After End of Scan (min)
Temperature = 0.00041 t + 20.58

Figure 3-4. Test #4 Post Scan Temperature Measurements.

Cooling Curve
(Test #4 Ring = 48)

21 -_

20.9 -- -------------,

0 10 20 30 40 50 60 70
Time After End of Scan (min)
Temperature = 0.00202 t + 20.60

Figure 3-5. Test #4 Post Scan Temperature Measurements.

Cooling Curve
(Test #6- Ring = 43)



0 20 40 60 80
Time After End of Scan (min)
Te 2mp0.erature = -00026 t + 20.15

Figure 3-6. Test #6 Post Scan Temperature Measurements.
Figure 3-6. Test #6 Post Scan Temperature Measurements.

Cooling Curve
(Test #6- Ring =25)


19.6 _____
CL19.5 -

19.4 -----------------

0 20 40 60 80
Time After End of Scan (min)
Temperature = 0.00414 *t+ 19.46

Figure 3-7. Test #6 Post Scan Temperature Measurements.

Cooling Curve
(Test #6 Ring = 42)

20.2 -

4D 20

0 20 40 60 80
Time After End of Scan (min)
Temperature = 0.00287 t + 19.78

Figure 3-8. Test #6 Post Scan Temperature Measurements.

--- Ml InstaScan
- xidbmmaint





Portable Gray Map

Figure 3-9. Diffusion Imaging Equipment and Software



Paint Shop Pro )

Portable Gray Map Bitmap

Diffusion Image Transformation
digitized comparisonscomposites


.l ow- ..

Figure 3-10. Diffusion Technique Images and Changes in Images.

Figure 3-11. Diffusion Image Prior to Heating Head

Figure 3-12. Diffusion Image After Heating Head

ImagsUJ2- -

teue04-z 2

CoIlee "

Figure 3-13. Histogram From Slice of T2-Weighted Image.

nag OV4 2

ce*w 91M 2

Figure 3-14. Histogram From Slice of T2-Weighted Image.

Linek 0 -
agSe ? 2

Im aIM 2 ..: ?. .. *' -.- '



Figure 3-15. Histogram From Slice of T2-Weighted Image.

Imap eW 2 .

h b_ .e 2

Figure 3-16. Histogram From Slice of Diffusion Image X-Axis.



Image 0-2
lnMx 09M.-2

Figure 3-17. Histogram From Slice of Diffusion Image X-Axis.

Image SO2 2

hwwON-i 2"

- ai.iliL lU *

Cmtgo^^ m^e f *

Figure 3-18. Histogram From Slice of Diffusion Image X-Axis.

...... . .

us.-. IF

I"- .. -

Figure 3-19. Histogram From Slice of Diffusion Image Y-Axis.





Figure 3-20. Histogram From Slice of Diffusion Image Y-Axis.

J- ...... -

Uk !"

mage 802 .2

*I _. __.


Figure 3-21. Histogram From Slice of Diffusion Image Y-Axis.

knegeOt- 2

C'm .,

Figure 3-22. Histogram From Slice of Diffusion Image Z-Axis.

Lin -f

. .. ... .o _.




L.ia& ~Ls~x

.-. .. .. .. .. "

Figure 3-23. Histogram From Slice of Diffusion Image Z-Axis.



he, m0-2

Compat, -. ... o. .
I r. .... ... ,

Figure 3-24. Histogram From Slice of Diffusion Image Z-Axis.

- -


Table 3-1. Temperature Measurements Using Thermocouples.

Date Power (W) Location Start Temp Stop Temp Scan Time
(F) (F) (mmn)
11/13/96 30 back 70.6 73.5 48.22
11/13/96 30 back 73.1 76.2 48.22
11/20/96 31 front 69.5 70.1 10.02
11/20/96 31 front 68.7 70.7 10.02
11/20/96 31 back 69.8 71.5 10.02
11/20/96 31 back 71.1 71.8 10.02
11/25/96 30 front 70.5 73.8 48.42
12/16/96 17 back 68.8 70.4 32.07
12/16/96 17 back 70.2 72.3 32.07
12/28/96 14 back 67.7 70.0 42.73
12/28/96 14 back 69.4 72.3 42.73
12/28/96 14 back 70.9 73.0 42.73
12/29/96 14 front 68.1 71.5 42.73
12/29/96 14 front 70.9 73.4 42.73
12/29/96 14 front 72.5 74.7 42.73
1/4/97 9 back 67.2 68.7 42.73
1/4/97 9 back 68.3 69.4 42.73
1/5/97 37 back 69.7 74.5 42.45
1/5/97 37 back 74.0 78.6 42.45
1/5/97 37 back 77.9 81.9 42.45

Table 3-2. Head Phantom Probe #2 Location.
locationn on Drawing Concentric
Test Dx Dy x y Ring No.
#1 12.2 26.3 70 52 41
#2 14.1 25.4 72 51 40
#3 0 22.5 68 49 38
#4 0.9 15 59 41 30
#5 1.9 13.1 60 39 28
#6 3.7 10.3 62 36 25
#7 -8.1 29.2 50 55 30
#8 6.5 10.3 685 36 25
#9 -2.8 19.6 55 46 35

Table 3-3. Head Phantom Probe #3 Location.
locationn on Drawing Concentric
Test Dx Dy x y Ring No.
#1 24.4 84.4 82 110 62
#2 35.7 65.7 94 92 74
#3 9.4 57.2 67 83 47
#4 36.5 30.99 94 57 46
#5 13.1 62.8 71 89 51
#6 4.7 88.1 63 114 43
#7 -14.1 59.1 44 85 24
#8 28.1 50.6 86 77 86
#9 1.9 62.8 60 89 40

Table 3-4. Head Phantom Probe #4 Location.
Location on Drawing Concentric
Test Dx Dy x y Ring No.
#1 34.7 121.9 93 148 73
#2 52.8 95.3 111 121 91
#3 18.8 86.3 77 112 57
#4 26.2 67.5 84 94 64
#5 15.0 101.2 73 127 53
#6 3.7 95.6 62 122 42
#7 -30.0 94.7 28 121 8
#8 42.2 80.6 100 107 80
#9 16.0 99.3 74 125 54

Table 3-5. Thermal Half-Life Using Fiber Optic Cables.
Average Half-Life
5 300
10 150
20 40


In the first series of thermal test, thermocouples were considered a cost effective

method of testing. Analysis of data revealed three major problems. First, the

thermocouple wires, being metallic, interfered with the imaging of the head phantom by

creating artifacts. For use with diffusion imaging the exact locations of the thermocouples

are required. With artifacts on the images it was difficult to plot the exact location of the

thermocouple tip. Additionally, the artifacts would also affect the diffusion image. This

would not provide an accurate indication of the change in diffusion as a function of

temperature, especially near the thermocouple probes which were the temperature


Second, the wires conducted the RF fields and increased the temperatures in the

tissue equivalent materials surrounding the wires. This created a false indication that more

energy was being deposited into the tissue resulting in greater temperature increases. The

equation of average power as indicated on the Omega and length of the scan indicated a

greater increase in temperatures over a shorter time interval. A comparison of this

equation with the one derived from fiber optic system results, revealed an increase in

temperature increases in the tissue equivalent material as a function of average power and

scan time. From the standpoint of safety, the thermocouple results had an inherent safety

margin by over estimating temperature increases. This margin, however, limited the

capabilities of 3T MRI.

Third, after the thermocouples were removed from the head phantom during a

scan and later reinstalled, the wire acted as a heat sink. The wires, being at room

temperature, were slightly cooler than the tissue equivalent material of the head phantom.

Once installed into the head phantom, the cooler wire acted as an excellent thermal

conductor and rapidly lower temperatures. Readings of 1 C drop in temperature within

two to four minutes were common. One method to reduce this would be to shield the

thermocouple end in plastic. This would insulate the wire from the tissue equivalent

material but would also slow down the response time of the thermocouple. Since the

cooling of the head phantom is slow this may be possible, however, the thermocouple wire

would still conduct heat faster until it reached equilibrium with the tissue equivalent


Fiber optic cables were immune to magnetic and RF fields. Three fiber optic

probes were simultaneously installed in the head phantom. Measurements of temperatures

were possible and revealed characteristics that would never have been seen with

thermocouple wires. Since the RF output power was constant during a long scan, it was

assumed that temperatures would increase linearly. The majority of temperature increases

were exponential. This exponential curve started earlier in the scan the closer the probe

was to the outer edge of the head phantom. This demonstrated that RF deposition, and

subsequently temperature increases, were greater on the outer edge. Thermal energy did

move inward and at the end of a scan the entire head phantom had increased in


temperature. Additionally, test results from Test #9 showed a slower rate of temperature

increase along the outer edge in comparison to an inner core probe. This validates the

assumption that the capability of the tissue equivalent material to increase temperatures

from a given power deposition decreases as the tissue equivalent material increases in

temperature. The amount of energy deposited in the tissue equivalent material nearer the

core was less than the energy deposited near the outer edge, but since the inner tissue

equivalent material was lower in temperature the deposited energy increased the

temperature more. These mechanisms were not detectable with the thermocouple system.

Fiber optic cables, remaining installed during a scan, will be instrumental in testing

diffusion images at different stages of a scan.

Deriving cooling curves would better quantify thermal energy deposition. The

temperature of any volume of tissue can be mathematically defined as simultaneously

receiving energy and dissipating energy during a scan.

Vol... = Jf(E EOur)

After a scan, with no source of energy, the temperature of a volume of tissue is a function

of energy loss only.

TVolume,,, = fE

Defining the energy deposition properties of the 3T system would be a matter of

subtracting the cooling equation from the heating equation. However, during these tests,

there was constant thermal movement up to an hour after the scan.

These results would prove useful in better characterizing the tissue equivalent

material. Unless the same process is performed on human tissue, there would be no

correlation and the information would not be useful. Additionally, the human head has a

thermal regulatory mechanism which is not available to the head phantom. A better

characterization of the tissue equivalent material would not improve test results.

Diffusion imaging, being only affected by temperature, can provide non-invasive

measurement of temperature increases. The FDA's guideline limiting temperature

increases to any 1 gram of tissue to I C can only be met through imaging. Any other

method requires invasive techniques that would not be comfortable to patients. Diffusion

is the only technique developed so far that is only affected by temperature. All other

imaging techniques are affected by some other parameter besides temperature. Until

another technique is developed or the other parameters can be mitigated, further testing

using diffusion imaging is warranted.


For the results obtained during this research to be valid for the human head, the

head phantom must be a good representative of the human head. Physically the head

phantom is an identical shape of a human head. Its electrical loading characteristics were

defined through an average loading of volunteers from 3T MRI. Using an established

recipe and retesting in the 3T imager, a tissue equivalent material to simulate the human

head was determined (Olsen 1992, Chou 1984). The primary component in the recipe was

water which is also the primary component in brain tissue. This suggests the thermal

characteristics of the head phantom and human head would also be similar. The outer

husk of the head phantom was an insulator similar to the skull and skin of the human head.

Both contain low water content and should have similar thermal characteristics.

Fortunately, with the use of diffusion imaging, exact duplication of test results between the

head phantom and human head may not be required.

Diffusion imaging had proved effective in determining the change in temperatures

in-vivo (LeBihan 1989, Stehling 1991, DePoorter 1994, McCain 1995, Mischler 1995).

Results mentioned in literature indicated an accuracy of 1.0 oC. Temperatures during the

diffusion imaging for this research used a heating technique that raised temperatures by

0.5 *C which was easily noticeable in test results.

This research has demonstrated that a diffusion imaging technique, already

available on MRIs, can be used to measure and monitor temperature increases in the

human head with resolution better than 1 C. An on-line program, similar to the one

written for this research would provide real time analysis of thermal changes by utilizing a

quick, initial diffusion image and then subsequent images throughout the session to

compare the change in diffusion. Further testing is required to better characterize the

temperature increase relationship to RF fields.

Future Work

Accuracy in determining temperature changes through diffusion imaging would

entail using the fiber optic probes in conjunction with the diffusion imaging technique.

The result would be a description of the change in pixel intensity with change in

temperature. Temperature gradients existed in the head phantom and would be a cause

for increased errors in temperature comparisons between the diffusion image results and

the fiber optic results. To minimize this effect, the diffusion image should be run on the

slice containing the fiber optic probe tips.

A variety of heating techniques resulting in a variety of different temperature

increases from 0.1 C to 2 C should be performed. The representative diffusion images

would then be transferred, digitized and analyzed using software similar to that used

during this research. Since noise was added each time an image was processed, an

analysis program should be written for the InstaScan unit or the Sun workstation. This

would eliminate the need to transpose the images from a Unix bitmap to a portable gray

map to a Windows@ bitmap and finally digitized. A program written for the InstaScan

would also reduce the time required to analyze images.

The next step in applying this technique to humans would be to run the diffusion

imaging technique on a cadaver head. The cadaver head, like the head phantom, should

start at room temperature. Artificially warming the cadaver to normal body temperature

would present a problem of maintaining that temperature. Once removed from the

warming mechanism, the head would begin to cool. The change in temperature would

skew the temperature changes as a function of MRI and also skew diffusion results.

Temperature increases by the RF fields would be negated by the cooling of the head. The

amount of temperature increase as a function of the length and strength of the RF fields

would have to be assumed as similar for tissue at normal body temperatures (37 *C) and at

room temperature (18 C).

Fiber optic probes would be inserted into the various brain and optic tissues. A

diffusion image would be acquired in the location of the fiber optic probe tips. A heating

technique would then be applied to the cadaver head. Post-heating diffusion images

would then be compared in the same manner as the head phantom diffusion images. Only

a handful of images, providing different temperature changes, would generate a

relationship between tissue type and pixel intensity as a function of temperature. This

pixel intensity verses temperature change would provide a conversion factor for each

tissue type. Most brain tissue (white matter and gray matter) have a high water content.

Temperature changes as a function of the RF fields would most likely be similar. As long

as the thermal characteristics of the brain tissues remain similar between room temperature


and normal body temperature, the conversion factor would be valid for living tissue at

normal body temperature.


Operating Mode Ranges for dB/dt

Normal Operating Mode

dB/dt < 20 T/s for 120 ps < T
dB/dt < 2400/T T/s for 2.5 ps < T 120 gts
dB/dt < 960 T/s for T < 2.5 ps

Where the pulse width, T, is defined as the duration of the change of magnetic flux
density expressed in microseconds (ps). For sinusoidally or other continuously
varying periodic magnetic fields the duration of change shall be considered to be
the half period the magnetic flux waveform.

For values of T not exceeding 2.5 ps, the limit on peak dB/dt is based on heating,
assuming a sinusoidal waveform. Therefore, the limit on peak dB/dt of 960 T/s
may be exceeded for non-sinusoidal waveforms, provided that the r.m.s. value of
dB/dt is less than 960 T/s and the peak dB/dt is less than 2400/T T/s.

First Level Controlled Operating Mode

dB/dt < 20 T/s for 3000 ps < T
dB/dt < 60000/T T/s for 45 ps < T < 3000 ps
dB/dt < 1330 T/s for T < 45 ps

For values of T not exceeding 45 ps, the limit on peak dB/dt is based on heating,
assuming a sinusoidal waveform. Therefore, the limit on peak dB/dt of 1330 T/s
may be exceeded for non-sinusoidal waveforms, provided that the r.m.s. values of
dB/dt is less than 1330 T/s and the peak dB/dt is less than 60,000/T T/s.

Second Level Controlled Operating Mode

Comprises values of dB/dt which exceed the upper limit for range of the First
Level Controlled Operating Mode.

Operating Modes

First Level Controlled Operating Mode

Magnetic Resonance equipment that is capable of values of dB/dt and/or SAR
above the upper limits of the Normal Operating Mode, shall comply with the

a) Before the start of each scan, an indication of the operating mode defined
by the maximum dB/dt value for the scan and a prediction of the value of SAR that
will actually be used during the scan shall be displayed at the Operators console.

b) If the value of dB/dt or SAR is such as to enter the First Level Controlled
Operating Mode, the attention of the Operator shall be drawn to this condition by
a clear indication on the Operators console and provision shall be made so that a
deliberate action of the Operator shall be necessary to start the scan. A record of
these values shall be inseparable from the image data.

c) A means of control shall be provided to ensure that values of dB/dt and
SAR do not exceed the upper limits for the First Level Controlled Operating
Mode. This control shall be independent of Operator input as the patient size,
weight or position.

Second Level Controlled Operating Mode

Magnetic Resonance equipment which allows values of dB/dt or SAR within the
Second Level Controlled Operating Mode shall comply with the requirements
above and in addition:

a) Shall include specific security measures that allow entry in the Second
Level Controlled Operating Mode. These measure shall include an indication to
the Operator that the operating conditions are potentially hazardous and that these
conditions should be not applied for normal clinical use.

b) The specific security measures shall be so constructed that it can be made
accessible only under the authorization of the medically responsible person.

c) The specific security measures shall be effective for each scan for which the
predicted values of dB/dt or SAR are above the limits for the First Level
Controlled Operating Mode.

d) The specific security measures shall involve a key-lock, a combination
local, a software password or other protective devices.

Checking Compliance

The accuracy of the predicted SAR with the specification shall be checked by
measurement at the maximum SAR of which the Magnetic Resonance equipment is
capable or at a SAR of 3W/kg, whichever is lower.

The upper limit of the First Level Controlled Operating Mode shall be measured to check
that it complies with the maximum SAR of which the Magnetic Resonance equipment is
capable or at a SAR of 4 W/kg, whichever is lower.

The requirement to prevent exceeding the upper limit of the First Level Controlled
Operating Mode for dB/dt shall be demonstrated by measurement using the maximum
gradient rate.

Limits for dBidt

dB/dt (Ts)

Range of Second Level Controlled Operating Mode
IL 4d

First Level Controlled Operating Mode

10" 10' 10' 103 10' 10'
T (psi

Figure 103 Limits for dB/dt.

Compliance is checked by measurement of the maximum trapezoidal dB/dt or the
maximum sinusoidal dB/dt component, that the system is capable of generating under any
conditions allowed for clinical use in the First Level Controlled Operating Mode.

Method of measurement of dB/dt to demonstrate compliance with the given requirements:

Method of measurement of maximum trapezoidal dB/dt:

a) Auxiliary definitions:

search coil

magnet isocentre

Patient space radius

dB/dt pulse width, T:

Search coil
position 2

a small diameter coil used to measure gradient field rates of
normally defined by magnet geometry, in this document it is
defined as the zero crossing for all three gradients.
the distance from magnet isocentre to the patient support (without
pads) along the lines y = x (see figure 103).
the Full Width at Half Maximum of a single dB/dt pulse or the half
period of a sinusoidal dB/dt waveform (see figure 104)


Search coil
/ \5 \ position 1

\ / /

S/ isocentre

Figure 103 Search coil placement for measurement of dB/dt.

LA- I/ -


a: magnetic field change waveforms
b: dB/dt waveforms


Figure 104 Magnetic field change waveforms and dB/dt waveforms.

b) Test hardware:

1) Search coil

The search coil shall be circular and shall be small with respect to the gradient coils
under test to ensure accuracy. The search coil consists of n turns of wire with a
radius of r meters. The axial length of the coil shall be less than 20% of its

The search coil shall be no more than 50 mm in diameter. The response of the
search coil shall be determined by calculation. The instantaneous magnitude of the
component of dB/dt coaxial with the search coil shall be determined from the peak
voltage, V, induced in the coil by the time varying magnetic flux:

dt nrV

For example a typical coil would consist of 15 turns of copper wire of 0.8 mm
diameter on a form of 50 mm diameter (r = 25 mm) resulting in a circular coil
approximately 12 mm long. An induced voltage of 200 mV would then result in a
dB/dt = 6.79 T/s coaxial with the search coil.

2) Voltage measurement device

The device used to measure voltage induced in the search coil shall have a high
input impedance and sufficient bandwidth to prevent signal attenuation, e.g. a
storage oscilloscope.

The voltage measuring device (storage oscilloscope) shall be placed at a location
where it is accurate and not affected by magnetic fields.

The search coil shall be connected to the voltage measurement device by means of
a low inductance cable, e.g. a twisted pair to avoid ringing on the waveform that
may be experienced with coaxial cables.

3) Positing device

A means shall be provided for positioning and aligning the search coil in the
magnet in a stable and reproducible manner. The device shall permit positioning of
the search coil throughout the region normally occupied by the patient while
maintaining the search coil coaxial with the static magnetic field.

c) Maximum gradient rate measurement for trapezoidal gradients:

Turn off or maximally attenuate the radio frequency (RF) transmitter to prevent

Position the search coil int he Magnetic Resonance equipment using the positioning device
so that the search coil axis is aligned with that of the static magnetic field, regardless of
patient orientation.

The amplitudes, A,. and A, are defined as the maximum gradient strength that can be
generated by the systems under normal scan conditions. The transition time t,, t2, t3 and t4
shall be the minimum that the system can generate using the standard pulse sequence
generation software and hardware used by the manufacturer. Begin simultaneous pulsing
of all three gradients (x, y and z) using the maximum amplitude bipolar gradient

Move the search coil to locate the position of the maximum induced voltage that can be
found within the region of the Magnetic Resonance equipment normally occupied by a

patient. Once this position is established, the coil shall be immobilized to prevent
movement-induced voltage from interfering with the measurement.

The following procedure locates the approximate position of maximum dBjdt in the
patient space of the Magnetic Resonance equipment with an axial field magnet and whole-
body gradient coils. Consider a cartesian coordinate system with the Z-axis pointing in the
direction of the static field and the Y-axis pointing vertically (e.g. as in figure 103). For
the maximum gradient rate measurement for trapezoidal gradients procedure, the points of
maximum dB/dt will occur where each of the gradients is near maximum and all the
gradient fields add. Typically, this is near the maximum winding density for the z-coil and
near the wall of the bore for X- and Y-gradients.

Assume that the distance, p, from the magnet isocentre to the patient support along the
lines y = x is the same as the maximum distance the patient may extend above the
magnet isocentre along the lines y = x, then the maximum dB/dt value is located at the

Where Z4 is the Z coordinate of maximum dB/dt.

Assuming all three gradients are applied simultaneously and with equal amplitudes, the
highest induced voltages will be found near the walls of the magnet shroud near the ends
of the gradient coils along the lines y = x (see figure 103).

Search coil position I in figure 103 shall be located at:

Y = X -P

Similarly, search coil position 2 shall be located at:

Y = -X P

Place the search coil coaxial with the static magnetic field at position 1 and constrain the
search coil so it may move only in the Z direction. Move the coil in the Z direction to the
position where the maximum voltage is induced in the search coil. Repeat this procedure


for position 2. The results from the position which yields the greatest search coil voltage
shall be recorded.

Measure the peak voltages which correspond to the maximum dB/dt for transitions t,, t2,
t3 and t4 (see figure 101).

Record the pulse width, T, of the dB/dt waveform corresponding to the t,, t2, t3 and t4 (see
figure 101).



2 I
--' "AA-max

Maximum amplitude bipolar gradient amplifier input waveform
(without pre-emphasis)

Figure 101 Waveform for performing measurements of acoustic noise and dB/dt.

d) Maximum sinusoidal dB/dt:

Step 1: turn off or maximally attenuate the radio frequency (RF) transmitter to
prevent interference.

Step 2: position the search coil in the Magnetic Resonance scanner using the
positioning device so that the search coil axis is aligned with that of the static
magnetic field.

Step 3: turn off all gradients other than the sinusoidal gradient of interest. Apply
the maximum amplitude and frequency waveform that the system can generate using
the standard pulse sequence generation software and hardware used by the

Step 4: move the search coil to locate the position of the maximum induced
voltage that can be found within the region of the machine normally occupied by the
patient. Once the position is established, the coil shall be immobilized to prevent
movement-induced voltage from interfering with the measurement.

Step 5: for most coils a suitable search procedure is as follows:

for x gradients, fix the coil along X = p, Y = 0
for y gradients, fix the coil along X = 0, Y = p
for z gradients, fix the coil along X = 0, Y = p

Step 6: then search along Z to locate the position of the maximum induced voltage
that can be fond within the region of the Magnetic Resonance equipment normally
occupied by the patient.

Step 7: measure the peak voltage which corresponds to the maximum dB/dt.

Record the pulse width, T, of the dB/dt waveform.

Repeat the above procedure for any other sinusoidally driven gradient.

a) Report of results:

The following data shall be recorded for the maximum gradient rate measurement for
trapezoidal gradients:

Parameter Dimension

maximum gradient pulse amplitude mT/m
patient space radius, p M
ti us
t2 ps
t3 us
t4 us
measured voltage voltage indicating T/s

The following data shall be recorded for the maximum sinusoidal dB/dt method:



- maximum gradient pulse amplitude
- patient space radius, p
- pulse width, T
- measured voltage

voltage indicating T/s