Vous êtes sur la page 1sur 86

1: Introduction


Nuclear Magnetic Resonance Field-Cycling (NMR FC) relaxometry is a technique used

to study the spin-lattice relaxation rates, T 1 of samples as a function of magnetic field

strength or Larmor frequency. Studies of this nature provide information on molecular

dynamics of such samples.

Initially, relaxometry studies involved „shuttling‟ the sample between different fixed

magnetic field strengths to obtain the respective relaxation signals from the different field

strengths. However, commercial advances in instrumentation have allowed for a rapid

variation of the magnetic field experienced by the sample while it is fixed in position.

The technique of rapidly varying the field strengths experienced by the sample is called

Fast Field Cycling (FFC).

NMR FC relaxometry shows promise in the early diagnosis of cancer, Alzheimer‟s and

muscle-wasting diseases. [1, 2] This is because these conditions can be probed by

relaxation mechanisms involving mainly hydrated proteins whose relaxation rates are

measured noninvasively, in vivo. [1, 2, 3] Relaxometry finds applications in the study of

liquid crystal dynamics, hydration of paramagnetic metal ions and organometallic

complexes and dynamics of proteins. [4, 5, 6] The latter being the major application in this

study: to measure protein concentration from the dynamics of proteins.

The strength of NMR FC relaxometry lies in the fact that it is fast, non-destructive and

the equipment is able to access information (with sensitivity and resolution) that is

inaccessible to the NMR spectrometer. [4]


This chapter introduces the subject matter and also presents the plan of the project report.







underlying concepts of relaxometry.





clarify the

Chapter 3 describes the equipment used, sample preparation, pulse sequence selection

and the general methodology involved in obtaining R 1 . The results of experiments are

presented along with discussions in Chapter 4.

Chapter 5 concludes the report with a brief outline of future work. Spreadsheets of data

from the experiments and other relevant material inappropriate for insertion into the main

report are placed in Appendices.

1.1 Project Activity Schedule and Materials











preparation of samples, experimentation, data processing and writing up. The research















understanding of the technical aspects and operation of the instrument, and a general idea

of the study area.












experiment. All activities carried out during the experiments, progress of work, details of

meetings and resolutions with supervisors of the project were recorded in a logbook. This

record of activities were, indeed, very helpful during the writing up stage.

The resources for the project were bovine serum albumin (BSA), phosphorus buffer








(CuSO 4 ),






relaxometer, a MR scanner, a computer and student versions of Matlab, Microsoft Excel

and Word.

1.2 The Literature Review Process









relaxation processes, protein dynamics, nuclear quadrupole resonance, field-cycling and

NMR relaxometry were used in various search engines for related publications.

The search engines used were: Scopus, ISI Web of Knowledge, the Aberdeen University

library catalogue and the Internet. Some articles were also obtained in hardcopies from

my supervisor, Dr. Gareth Davies.

The relevant publications for this project have been duly referenced, where appropriate,

within this report. To the best of my knowledge, therefore, no such work has been

directly published elsewhere.

1.3 Significance

This study was aimed at determining the veracity of the proportional relationship between

the magnitude of quadrupole peaks and protein concentration in protein gel samples; and

also what were the effects on this relationship with the addition of non-protein gel

components or contrast agents. The sensitivity, to protein changes, of the NMR FFC

relaxometer used was also assessed; thus, providing first-hand information to enable a

further study of protein dynamics in future.



2: Background Theory and Literature Review

This chapter discusses the background of the project and how it fits into related cases of

study to solve problems in the study area. An emphasis is placed on the basic concepts of

relaxometry, sourced from related literature.

2.1 About the Project

This project exploits 1 H dipole- 14 N quadrupole cross relaxation to measure an effect that

should be proportional to the immobilised protein content of a gel. The goal was to use

the field-cycling technique to rapidly, electronically vary the magnetic field experienced

by the protein sample in the FFC NMR relaxometer at a switching time less than the T 1

relaxation constant of the sample so that the cross-relaxation effects can be explored at

specified field strengths of 16 mT, 49 mT and 65 mT, where quadrupole peaks are

observed in the dispersion profile. [1, 2, 5, 7, 8, 9, 10] However, for sensitivity and signal

homogeneity reasons, only field strengths of 49 mT and 65 mT were explored.

From the R 1 dispersion plots, a further processing technique was developed to extract and

compute the size of the quadrupole peaks, called change in relaxation rate, R 1 , due to

quadrupole processes.

R 1 has previously been shown to be proportional to the

concentration of protein in each sample. [1]

Finally, based on the effect of a contrast agent on the direct relationship between R 1 and

the concentration of protein in a given sample, a physiologically representative phantom

(with realistic T 1 and protein concentration characteristics) was produced, imaged and its

image analysed to assess the agreement between the MR imaging and the relaxometry


data sets. In the end, the sensitivity (to protein concentration changes) of the newly

acquired departmental “Stelar SMARtracer FFC relaxometer” used for the project was

assessed for the first time.

The next lines of discussion below present the background theory of field-cycling and its

applications in NMR relaxometry, with particular emphasis on protein concentration


2.2 Nuclear Magnetic Resonance

Nuclear Magnetic Resonance (NMR) is a process that involves the physical interaction

between radio waves at specified resonance frequencies and certain nuclei in a magnetic

field, the nature of the resonant interaction being dependent upon the magnetic quantum

mechanical properties of the nuclei. [11] After its discovery in 1938, it was further

developed by F. Bloch and E.M Purcell, independently, in 1946 for measurement of spins

and magnetic moments of nuclei. [12] Since this period, NMR has been applied to the study

of the chemical and physical properties of molecules (NMR Spectroscopy), biological

and medical applications (MR Imaging) and the structure and dynamics of solids, liquids

and gases (NMR Relaxometry). [13, 14]

The resonant interaction of the nucleus with the magnetic field can be explained by two

models: the Quantum Mechanical and Classical Models. [11]

The Quantum Mechanical Model explains that the ability of the nucleus to interact with a

magnetic field is due to its intrinsic spin that causes it to apparently precess about its axis.

The relative number of protons and neutrons present within a given nucleus determines

the total (intrinsic) spin, I, of the nucleus, where I is an integer or half integer value. This,


therefore, means that for a nucleus to interact with a magnetic field, I ≠ 0 necessarily;

which further means that such a nucleus possesses angular momentum, p, along the axis

of the spin related to the intrinsic spin, I, by: [11]

p = I


where = h/2, h is Planck‟s constant = 6.626 x10 -34 Js. [13]

The charge in the nucleus creates an effective current loop by rotating about its axis, thus

giving rise to a resultant magnetic dipole moment, µ, given by: [11]

µ = p = I


where is a constant called the “gyromagnetic ratio”, whose value is characteristic of a

given nucleus. For instance, the 1 H nucleus, which is the nucleus of interest in most NMR

experiments (including this project) due to its relative abundance of 99.98 % and greatest

NMR sensitivity, consists of a single proton and spin, I, value of ½. For protons,

= 26.75 X 10 7 rads -1 (/2= 42.58 MHzT -1 ). It is this dipole moment of the nucleus that

makes it behave like a little bar magnet (Figure 2.1), capable of physical interaction with

an external magnetic field. [11]

Direction of magnetic field
Direction of
magnetic field

Bar magnet

Spinning proton

Figure 2.1: A spinning nucleus has a net spin, I and a magnetic dipole moment, µ along its axis, analogous to a bar magnet. After reference 14


The quantum mechanical model further specifies that the magnetic dipole moment of the

nucleus can only have (2I + 1) orientations in the magnetic field, corresponding to

(2I +1) discrete energy levels. Thus, the 1 H nucleus with I = ½ has two quantised states:

“spin up” and “spin down”. In a magnetic field, therefore, it tends to align with or against

the field corresponding to low and high energy states respectively (Figure 2.2). The

energy difference between the two quantum states is proportional to the external

magnetic field, B 0 : [11]

E = µ

B 0 / I = B 0 Parallel B o Antiparall
B 0 / I = B 0
Antiparallel E

Parallel State

(2. 3)

Figure 2.2: Orientation of the nuclear spins in an external magnetic field, B 0 the antiparallel state is slightly higher in energy than the parallel state. After reference 14

Application of electromagnetic (em) radiation of suitable frequency, called the Larmor

frequency, L , causes transitions between the two energy states. This Larmor frequency

is related to the energy difference, E, between the two states by: [11]

E = h L = L

L = E / = B 0




where L is the frequency in hertz (Hz) and L is the angular frequency in radians per

second (rads -1 ) of the electromagnetic radiation to cause the transition from the lower to

higher energy states.

The populations of spins in the lower, N L , and higher, N U , energy states are described by

the Boltzmann distribution: [11]

N L / N U = exp (L / k T)


where k is Boltmann‟s constant and T is the absolute temperature.

Thus, only radiation of frequency, L , can cause this transition to occur and so the

resonance effect – the origin of the name of the phenomenon of “Nuclear Magnetic


The observable spins (after the application of the em radiation) in a NMR experiment are

those spins in the lower energy state, referred to as the “fractional excess”, given by: [11]

(N L N U ) / N U = L / k T = B 0 / k T


Thus, equation (2.7) suggests that to increase sensitivity to the NMR signal detection of

the equipment (imager or relaxometer), we must either decrease T (impractical in

patients, though possible in the samples used in this project) or increase B 0 .

The Classical Model rather permits spin orientation in all directions in the external

magnetic field, where the magnetic moment, µ, of each spin is being aligned with the

magnetic field, B 0 , by experiencing a turning force, L, given by: [11]

L= µ B 0



Due to the angular momentum, p, of the individual spins, the turning force, L, rather

causes them to precess about B 0 , similar to a gyroscope in a gravitational field, at a rate

of change of momentum given by: [11]

dp / dt = L =

B 0


d/ dt =  B 0 = L (since p = / )


where L = -B 0


Equations (2.10) and (2.11) infer that the nucleus spins at a rate of L rads -1 about the

direction of the applied field, where the minus (-) sign gives the direction of rotation.

This frequency of precession is equal to the frequency of resonance of equation (2.5) and

is called the Larmor frequency.

Ideally, in a NMR experiment, it is the overall effect of the individual magnetic moments

of all the nuclei of interest in a sample that is observed but not a single magnetic moment.

The combined observable magnetic moment (viewed as a large magnetic dipole moment)

is called “bulk magnetisation, M” and is given by:

M =Σ


At equilibrium, M has no transverse component to B 0 due to phase incoherence among

the many different making up M. However, due to the preference for these many to

align with B 0 , the result is that M has a longitudinal component along the B 0 direction, of

magnitude M 0 , called the equilibrium magnetisation, which is equal to the spin excess

detected as explained by the quantum model (Figure 2.3).


B o
z M o x y
M o

Figure 2.3: Equilibrium magnetisation, M 0 , aligned with B 0 . After reference 13

In order to detect this bulk magnetisation, it is caused to precess about B 0 by the

application of a second magnetic field, B 1 , to excite the NMR nucleus. B 1 is generated by

the passage of an oscillating current at the Larmor frequency, L , through a coil which

creates an oscillating magnetic field, B 1 , perpendicular to B 0 . The B 1 field displaces M

away from B 0 and causes it to execute a spiral path (Figure 2.4). When B 1 is switched off,

M continues its precession about B 0 , describing a cone at a flip angle, , to B 0 . depends

on the strength of the B 1 field and how long, t p , it acts: [11]

= B 1 t p


t p , it acts: [ 1 1 ]  = B 1 t p (2.13)

Figure 2.4: Spiral path executed by the bulk magnetisation, M 0 following a 90 0 RF pulse viewed from the laboratory frame of reference. From reference 12


At a given amplitude and time of the B 1 field, it is possible to tip M through 90 0 (where

B 1 is called a “90 0 pulse”) or completely invert M to the B 0 axis (where B 1 is called an

“inversion pulse” or “180 0 pulse”). The frequencies of B 1 used in NMR experiments

typically fall within the radiofrequency bandwidth (1 500 MHz) of the spectrum and so

B 1 is often referred to as a radiofrequency (RF) magnetic field and its pulses are RF

pulses. [11] RF pulses of 180 0 and 90 0 shall be explored in this project.

As long as the magnetisation, M, is flipped away from the B 0 axis, its transverse

component generates an oscillating magnetic field. This is then detected as a small

voltage (which also oscillates at L ) by electromagnetic induction into an RF coil, tuned

at L .

In the presence of a 90 0 RF pulse, this oscillating transverse component of M, and for that

matter, the induced signal in the RF coil, is maximum. At the end of the application of the

RF pulse, the signal decays gradually due to the relaxation of M, which causes it to return

to its equilibrium position, M 0 , parallel to the axis of B 0 (Figure 2.5). This detected

decaying signal in the absence of the RF pulse is called “Free Induction Decay” or FID.


0 90 RF Pulse Signal
RF Pulse


Figure 2.5: An FID after the application of a 90 0 RF pulse. After reference 13


It is usual to describe the behaviour of M in a “Rotating Frame of Reference” for ease of

visualisation, especially when more than two RF pulses are involved. [11] In this reference

frame, the x and y axes are replaced with x / and y / synchronised with the rotation of the

nuclear magnetic moments,

2.3 Relaxation Phenomena

Nuclear relaxation processes involve the thermal equilibration of the spin systems with

respect to longitudinal or transverse magnetisation components, multiple-quantum spin

coherences and longitudinal dipolar, quadrupolar or scalar order. [15] The effect of

relaxation is that it causes a gradual decay away of the FID detected in a NMR

experiment due to energy exchange “between spins” and “between spins and their

surroundings”. These interactions are called “spin-spin relaxation” and “spin-lattice

relaxation”, with time constants T 2 and T 1 respectively. [11] These two relaxation

processes together cause the magnetisation vector to return to its equilibrium position

parallel to the axis of B 0 . Thus, the T 1 and T 2 relaxation time constants are valuable

parameters for studies of molecular dynamics, [11] and so in this project, the T 1

components of the protein samples are used in the study of their molecular dynamics. A

parameter of interest, called the “relaxation rate” or R 1 , is evaluated as the reciprocal of

the T 1 values (1/T 1 ) in order to quantify the amount of polypeptides in the samples.

Conversely, the relaxation rate R 2 is also 1/T 2 .

Spin-spin relaxation effect is caused by dipole-dipole interactions between neighbouring

nuclear magnetic moments in an externally applied B 0 field. These interactions result in

variations in the precessional rates of the individual spins and hence loss of phase


coherence in the detected transverse magnetisation, M XY with time constant T 2 (Figure

2.6). Ordinarily, magnetic field inhomogeneities experienced by the sample cause

additional signal decay or “pseudo-relaxation” with time constant T 2inhom . Therefore, the

net loss of phase coherence is described by another time constant, T 2 * , defined as: [11]

1 / T 2 * = 1 / T 2 + 1 / T 2inhom


where 1/T 2inhom = B 0 ; with B 0 defined as the variation of the applied magnetic field

strength over the region of the sample.

magnetic field strength over the region of the sample. Figure 2.6: Loss of phase coherence between

Figure 2.6: Loss of phase coherence between the magnetic moments, µ i , due to spin-spin relaxation causes an exponential decay of the transverse magnetisation, M XY , with a time constant T 2 . From reference 11

To measure both the T 2 and T 1 relaxation times, a collection of RF pulses, with

predefined time intervals to control the reception and characteristics of the NMR signal is

used. This set of RF pulses is called a “pulse sequence”. There are several pulse

sequences available for the measurement of the T 2 and T 1 relaxation time constants but

usually, T 1 is measured by an “inversion recovery” pulse sequence whereas T 2 is

measured by a “spin echo” pulse sequence.



Figure 2.7: Measurement of T 2 in the spin-echo experiment (shown in the rotating frame of reference, x / , y / , z). After reference 11

The spin echo pulse sequence employed in the measurement of the T 2 relaxation time

comprises a 90 0 pulse, followed after an echo time TE/2, by a 180 0 pulse (Figure 2.7).

The 90 0 pulse flips the magnetisation vector, M, onto the y / axis (a), where the

component vectors making up M are caused to dephase with respect to one another due to

inhomogeneity of the applied magnetic field (b). Following the application of the 180 0

pulse, after the waiting time of TE/2, the “fan” of magnetisation vectors is flipped over to

the y / axis (c). The initial dephasing process after the 90 0 pulse is now exactly reversed

and the vector sum of the rephrasing magnetisation vectors reaches a maximum at time

TE, at the centre of an echo signal (d). In this case, instead of the initial FID after the 90 0

pulse, it is rather the spin echo signal after the 180 0 pulse that is acquired (e), the

amplitude, S 2 , of which is given by: [11]

S 2 = S 1 exp (-TE / T 2 )



where S 1 is the initial amplitude of the FID generated by the 90 0 pulse. The value of T 2

can, therefore, be calculated from equation (2.15) above.

Spin-lattice relaxation effect is caused by the interactions of the spin system with its

surroundings, resulting in energy loss from the spins to their surroundings. The energy






decay towards




M 0 ,



longitudinal component of the magnetisation vector, M z , with a time constant T 1 . Once

spin-lattice relaxation has returned the magnetisation vector to its equilibrium value, M 0 ,

there cannot be any transverse magnetisation component. [11] Hence, T 2 is always less

than or equal to T 1 . In fact, both T 1 and T 2 processes occur simultaneously with the only

restriction being that T 2 is less than or equal to T 1 . [13]

z x y
z M x y
than or equal to T 1 . [ 1 3 ] z x y z M
z M x y
z M x y

Figure 2.8: Return of the bulk magnetisation to its equilibrium magnetisation, M 0 , due to spin-lattice relaxation with time constant T 1 . After reference 14


The inversion recovery pulse sequence used for measuring the T 1 relaxation time of a

sample comprises a 180 0 pulse, followed after an inversion time TI, by a 90 0 pulse (Figure

2.9). The 180 0 pulse inverts the magnetisation onto the z axis such that its M z







M 0








magnetisation recovers towards its equilibrium value of +M 0 along the z axis (c).

Following the application of the 90 0 pulse, any M z component recovered (along the z

axis) is converted into transverse magnetisation, M xy (d). The magnitude and sign of the

resulting FID is indicative of how far the magnetisation had recovered along the z axis

during TI.

TI 180 0 90 0 RF (a) (b) (c) (d) Signal
180 0
90 0
(b) (c)
z (a) M 0 x / y /
M 0
x /
y /
0 RF (a) (b) (c) (d) Signal z (a) M 0 x / y / (c)
(b) (d)
B 1 -M 0
B 1
-M 0
B 1
B 1

Figure 2.9: The inversion recovery experiment, viewed from the rotating frame of reference x / , y / , z / . The 180 0 RF pulse is followed by a 90 0 RF pulse after an inversion time TI (longer than both pulse lengths). After reference 11


The value of T 1 in an inversion recovery experiment is calculated by curve fitting to a

multi-point TI data set (Figure 2.10) using the general equation describing the signal

strength, S proportional to the magnitude of M z , just before the 90 0 pulse: [11]

S M z (TI) = M 0 [1 2 exp (-TI / T 1 )]


where M 0 is the equilibrium magnetisation (equivalent to the proton density in the



Figure 2.10: Plot of M Z versus time in the inversion recovery pulse sequence. At t = 0, the inversion pulse ends; at t = TI at the end of the inversion pulse, the 90 0 pulse is applied, converting M Z into transverse magnetisation and generating an FID proportional to M Z (TI). Three curves are shown for samples with different T 1 values: the sample with the shortest T 1 value appears brightest in an MR image since its magnetisation recovers fastest to a large positive value. After reference 12

It is worth mentioning that the measured T 1 and T 2 values are influenced by a number of

extrinsic (operator-controlled) and intrinsic (sample property) factors. [11]

Molecular motions (tumbling) in solutions result in time varying magnetic fields and so

cause spin relaxation. This is because time varying fields at the Larmor frequency cause

transitions between the spin states and hence, a change in the longitudinal magnetisation,


M z . Only those rotational frequencies among the distribution of molecules at the Larmor

frequency affect T 1 . Since the Larmor frequency is proportional to B 0 , T 1 will therefore

vary as a function of magnetic field strength with the result that T 1 becomes inversely

proportional to the density of molecular motions at the Larmor frequency. [11, 13]

Moreover, the motional frequency distribution is dependent upon the temperature and

viscosity of the solution. Hence, T 1 will again vary as a function of temperature as shown

in Figure 2.11.

vary as a function of temperature as shown in Figure 2.11. Figure 2.11: Plot of number

Figure 2.11: Plot of number of molecular motions versus frequency at different temperatures: at the Larmor frequency indicated by 0 , T 1 (280 K) < T 1 (340 K). From reference 13

This makes keeping the temperature of samples constant essential in the experiments

within this project. Thus, the temperature of each sample did not vary by enough to cause

a significant influence on T 1 at a given field strength. The viscosity, however, does vary

significantly among samples (depending on the concentration) and

so a resulting

influence on T 1 . [13] A highly viscous solution would have a shorter T 1 than a less viscous

solution. This also required that samples be given long enough time to denature


considerably to gels to give high signal-to-noise ratio in the measurements of their

relaxation rates. Figure 2.12 shows spectral density functions of the variation of Larmor

frequency with sample viscosity.

of the variation of Larmor frequency with sample viscosity. Figure 2.12: Spectral density functions of Larmor

Figure 2.12: Spectral density functions of Larmor frequency variation with viscosity. From reference 13

Furthermore, variable fields which agitate the energy levels of the spin states dephase the

transverse magnetisation, M xy , with the result that the number of molecular motions less

than or equal to the Larmor frequency becomes inversely proportional to T 2 . Thus,

relaxation times get longer with increasing B 0 as there are fewer relaxation-causing

frequency components available in the random molecular motions. [13]

In addition to the above, the exchange of spin states between two spins affects T 2 but not

T 1 . T 1 is not affected because spin distribution between the lower and upper states is not

changed but phase coherence of the transverse magnetisation, M xy , is lost during the

process of spin exchange which consequently affects T 2 .


Lastly, the exchange of nuclear species between two or more molecules in a mixture by a

process called chemical exchange affects both T 1 and T 2 relaxation times. T 1 is affected in

this case because energy is transferred from one nucleus to another. [13] For instance, if

there are more nuclei in the upper state of specimen Aand a normal Boltzmann

distribution in B, the exchange process will force the excess energy from Ato B

with the effect that T 1 becomes smaller. This phenomenon has been observed in this

project from a mixture of the protein samples with some contrast agents. The chemical

exchange process reduces the T 1 relaxation time of the samples and so increased

relaxation rates are observed from them. T 2 is also affected because phase coherence of

the transverse magnetisation is not preserved during a chemical exchange process. [13]

2.4 Nuclear Quadrupole Resonance

Nuclear Quadrupole Resonance (NQR) is a physical phenomenon involving the nucleus

interacting with an external field similar to NMR discussed in section 2.2. In NMR, it has

been shown that the energies of nuclei with a magnetic dipole moment are split by an

external magnetic field in order to allow resonance absorption of energy that is related to

the difference between the ground state energy and the excited state. In NQR, however,

the energies of nuclei with an electric quadrupole moment are split by an electric field

gradient that has been created by the electronic bonds in the local environment. [6] The

nuclei are then set in a static inhomogeneous electric field, allowing the absorption of






[16] Since







environment without a static (or direct current) magnetic field, it is sometimes called

“zero-field NMR”. Many NQR transition frequencies rely greatly on temperature. [6]


Any nucleus with more than one unpaired nuclear particle (the nuclear particles being

protons or neutrons) will have a quadrupolar charge distribution.

Examples of such

nuclei are nitrogen-14 ( 14 N), chlorine-35 ( 35 Cl) and copper-63 ( 63 Cu). The resonance

interaction of this quadrupole with an electric field gradient supplied by the non-uniform

distribution of electron density (from bonding electrons) results in the NQR effect. This,

thus, makes the technique of NQR very sensitive to the kind of bonding surrounding the

quadrupolar nucleus. [6]

For an illustration of the NQR effect, in the simplest case of 35 Cl in solid Cl 2 , NQR is

involved in the precession of the angular momentum, I (and the nuclear magnetic dipole

moment, µ) of the nucleus, shown in Figure 2.13, as a flat ellipsoid of rotation around the

symmetry axis (taken to be the z axis) of the Cl 2 molecule fixed in the crystalline solid.


Z θ I I z eQ Cl 2
I z
Cl 2

Figure 2.13: Interaction of 35 CI nucleus with the electric field of a Cl 2 molecule. After reference 16


The constant angle, , (of the precession) between the nuclear axis and symmetry axis of

the molecule results from the torque exerted by the inhomogeneous molecular electric

field on the nucleus of electric quadrupole moment, eQ. Classically, the resonant

absorption occurs when the frequencies of the RF field and the precessing motion of the

angular momentum are equal.

The resonance interaction in NQR is considerably larger compared with the chemical

shift measured in NMR but can be approximated to zero (or will not virtually occur) in

the liquid phase. Therefore, NQR spectra can only be necessarily measured for samples

in a fairly solid phase. This condition is remarkably both a strong point and a weakness of

the NQR phenomenon. [6] Samples used in this project, therefore, had to be “cooked” over

prolonged periods to ensure complete denaturation to gels in order to measure the NQR

effect in them.


far, NQR spectra have been observed in the approximate Larmor frequency range of 1


1000 MHz [16] and in this project, a range of 1.5 to 3.5 MHz is used to probe the NQR

effect in the protein samples. These samples have 14 N- 1 H groups (in which the backbone

H atoms are strongly coupled to the spins of adjacent N atoms) [12] that serve as

“relaxation sinks” in order to enhance proton relaxation due to their interactions with the

14 N quadrupolar nucleus. [2, 5] This leads to reductions in the proton spin-lattice relaxation

time, T 1 , called “quadrupole dips” (where 1/T 1 is called a “quadrupole peak”), which

occur at three NMR frequencies corresponding to the 14 N nuclear quadrupole transitions.

Since 14 N has I = 1, this implies that (2I + 1) or 3 energy transition states are available in

the 14 N nucleus, thus, resulting in electrical asymmetry in the peptide nitrogen and hence,

the three (3) distinct transitions in the 14 N system to which the protons of the protein may


couple. The transition frequencies are observed at field strengths of 65 mT, 49 mT and 16















frequencies). [2, 4, 5, 7] Thus, at these stated magnetic field strengths, protein and water

proton spin-lattice relaxation rates become sensitive to the concentration of rotationally

immobilised peptide nitrogen due to field dependent heteronuclear cross relaxation

coupling between protein proton and nitrogen-14 spins that have been carried to the water

by proton homonuclear cross relaxation. [1]

Measurement of the water proton spin-lattice relaxation time or a signal amplitude that is

proportional to it (in this case, the size of the quadrupole dip or peak) may provide a

noninvasive measure of peptide bond concentration to reveal immobilised protein content

in most tissues. [1] The “quadrupole dip” effect was studied to a greater extent in the early

to mid 1980s and dips were measured in hydrated proteins and various biological

samples. Kimmich et al carried out the first in vivo demonstration of the effect in living

leeches. [8] This was followed by the work of Lurie in the 1990s in which quadrupole dips

were measured in vivo in human muscle and brain using a whole-body sized field-cycling

relaxometry and imaging system. [2, 10] Among other subsequent works, quadrupole dips

have been measured in collagen fibres, [8] calf lens [9] and multiple sclerosis plaques. [12]

In this study, however, the “quadrupole peak” effect shall be measured in vitro in protein

samples prepared by the author.

2.5 Field-Cycling NMR Relaxometry

In normal NMR and MRI, the magnetic field strength, B 0 is fixed and very stable

throughout the signal measurement and imaging processes respectively. In Field-Cycling


NMR, however, the B 0 field is switched rapidly to different values during the pulse

sequence. If the B 0 field switching time is less than the T 1 relaxation time of the sample,

the technique is called “Fast Field Cycling” or FFC.

The process of varying the field strength, B 0 experienced by the sample can be achieved

by either moving the sample through a magnetic field gradient (i.e. “shuttling” the sample

rapidly between different magnetic fields) or by keeping the sample fixed in position

while varying the magnetic field it experiences. [5, 17, 18, 19] This latter option of field-

cycling application is what is being employed in this project by the use of an instrument

called the FFC NMR relaxometer. The technique is called field-cycling (FC) NMR

relaxometry. This is a NMR technique used to determine the longitudinal relaxation time

(T 1 ) over a range of magnetic field strengths at a fixed temperature. However, the

boundaries of this range are not specific (Figure 2.14): the lower limit depends on the

local fields while the upper limit is mainly determined by technical choices. [4]

limit is mainly determined by technical choices. [ 4 ] Figure 2.14: Time (  )

Figure 2.14: Time () and angular frequency () scales covered by various NMR techniques. The ranges indicated refer to proton resonance. From reference 5

T 1 is particularly suitable in relaxometry studies because T 1 in tissues is strongly

dependent on field strength (gets longer with increasing field strength) while T 2 is hardly

influenced by changes in field strength. [7, 11]


More generally, FC NMR relaxometry is used to study the behaviour of nuclear

relaxation phenomena dependent on internal and/or external parameters, among which

are molecular and supramolecular structures, temperature, viscosity, pH, the influence of

paramagnetic and ferromagnetic agents and the magnetic field strength. [7] The theory of

the technique is based on the physical aspects of nuclei relaxation to the ground state

after being excited by an RF pulse. [3]

While the conventional NMR approach studies molecular motions at constant frequency

by varying the temperature in biological systems, NMR relaxometry does exactly the

opposite. This is advantageous because when the temperature is varied, the sample

suffers from severe limitations arising from freezing or denaturation of the sample

material. Therefore, the only possible way to study the protein motions by relaxation

spectroscopy is to vary the Larmor frequency this is called Nuclear Magnetic

Relaxation Dispersion (NMRD). [9] Variation of the Larmor frequency means variation of

the external flux density, B 0 (according to equation 2.5). The range of variation, however,

is limited in conventional NMR spectrometers due to the fixed flux density provided by

the magnet in use. [15]

The dispersion depends on the type of motion and on the observed nucleus. The most

informative 1 H NMR spectroscopic frequency range for rotational motions is centered

around 10 MHz for small motions (molecular weight, M w 10 4 Da) and smaller

frequencies for larger proteins and protein aggregations [20] (e.g. BSA with M w 66 ×10 3

Da). However, at these low fields, resolution is low and so, practically, only one signal

can be detected. Exploring the low frequency range, on the other hand, is beneficial in

isolating the typical relaxation features associated with molecular processes characterised


by very long correlation times like molecular surface dynamics and collective effects.

The range, thus, significantly enhances the spatial domain usually explored by the

relaxation techniques. This advantage of low field relaxometry has been exploited in this

project, where only the abundant water protons (with high S/N) can be conveniently

studied under such low sensitivity. [20]

For the purpose of analysis and good understanding of NMR relaxation experiments with

biological materials, it is important to discuss the relevant interactions and fluctuations at

the molecular level in such materials. This is necessary because spin-lattice relaxation is

generally due to spin couplings modulated by the fluctuations within the system. [8]

The interaction of water with biopolymers in general is considered to be the main source

of relaxation in biological tissue, and is therefore of paramount importance for the

interpretation of both relaxation data and MR images. [5] This interaction is paramount as

water molecules have sufficiently rapid motion such that on the time scale of a relaxation

time, they have thousands of encounters with the cellular constituents; the collective,

average influence of the constituents on the water protons is manifested in 1/T 1 . [9]

Fundamentally, three (3) competitive mechanisms contribute to the spin-lattice relaxation

of this hydration water: [5]

I. Restricted rotational diffusion of water molecules about axes perpendicular to the

local surface. This process can be described by an exponential correlation

function with a correlation time not much longer than that in bulk water.


II. Reorientation








somewhat, rough and coiled surface of the protein. This depends on both the

surface topology and the effective diffusivity along the surface.

III. Tumbling of the protein molecule as well as its hydration shell. An exponential

function can be used as a correlation function to describe this process.

However, restricted rotational diffusion is feasible only at high frequencies (10 MHz).

The low frequency dispersion is influenced by the RMTD mechanism of water and

tumbling of the hydrated protein molecule. The later mechanism is only possible if the

protein molecules are not prevented by mutual sterical hindrance at high protein

concentrations. Aside tumbling of the protein molecule itself, if not immobilised, spins

are relaxed by backbone fluctuations and side-group motions.

Only proteins in aqueous solutions undergo tumbling and side-group motions; the latter is

only possible at frequencies greater than or equal to 10 8 Hz (or 100 MHz). [21] Since the

protein samples used in this project were denatured to gels, and the maximum frequency

range explored was from 0.5 to 8.0 MHz, the only two possible relaxation mechanisms

were RMTD and backbone fluctuations.

Surprisingly, the R 1 dispersion profile is qualitatively the same regardless of the presence

of a bulk-like water phase. The same dispersion features occur even at water contents as

low as 25% by weight, where the hydration cells are just saturated and where protein

molecules are immobilised. In the absence or relatively low content of water (such that

water proton signals are negligible) protein internal motions become relevant in the spin-

lattice relaxation dispersion of protons. [5] This is when backbone fluctuations dominate


in the low frequency field-cycling window. A direct indication of backbone fluctuations

are 14 N- 1 H and (if deuterated water is used) 2 H- 1 H quadrupole dips originating from

relaxation sinks formed by quadrupole nuclei in the amide groups 14 N and (after

deuteron exchange) 2 H. The additional quadrupole interactions experienced by these

nuclei cause a much tighter coupling to the lattice and consequently produce enhanced

(faster) spin-lattice relaxation. This phenomenon reveals itself as “quadrupole dips” in the

T 1 dispersion (Figure 2.15). [5, 9]

in the T 1 dispersion (Figure 2.15). [ 5 , 9 ] Figure 2.15: Origin of

Figure 2.15: Origin of quadrupole dips in the case of 14 N- 1 H amide groups (a) Magnetic field dependence of the 1 H (H ) and the three 14 N ( (1) , (2) , (3) ) resonance frequencies in amide groups. (b) Proton spinlattice relaxation dispersion; the quadrupole dips arise at the resonance crossings of 1 H and 14 N. From reference 5


Thus, the condition for the occurrence of quadrupole dips or peaks is that molecular

motions are restricted such that motional averaging is incomplete on the time scale of the

experiment. This is true for dry or hydrated (but rotationally immobilised) proteins as

rotational reorientation of the peptide environment on the time scale of hundreds of

nanoseconds will average the nuclear electric quadrupole interaction to zero. [1]

Specifically for protein backbone fluctuations, spin-lattice relaxation is universally

subject to a power law after elimination of potential contributions from side-group

motions: [5, 21]

T 1  L b

which is expressed as a rate:

R 1  L -b ; (where R 1 = 1/T 1 )



where the exponent „b‟ is usually in the range 0.65 to 0.85 at room temperature and

decays to lower values upon temperature reduction. [5] Kimmich and Nusser have

reported that „b‟ changes at about 200 K from a constant value above this temperature to




decreasing temperatures.



they reported,



manifestation of a transition from “ergodic” (at high temperature) to “nonergodic”

(associated with frozen states) behaviours in the time scale of the experiments. The

conclusion, therefore, is that interpretation of water relaxation data on the basis of

intramolecular protein dynamics alone is misleading. [21]

Hence, for a complete description of the relaxation data referring to protons bound to

protein backbones, the two rate contributions below are essential: [8]


1/T 1 = R 1 = R p + R q


where R p refers to homonuclear dipolar coupling between protons (i.e. proton-proton

interaction) and R q refers to heteronuclear dipolar interaction between protons and non-

resonant nuclei (i.e. the quadrupole peaks).

2.6 Technical Aspects of FC NMR Relaxometry

In a typical FFC multi-block experiment, a high magnetic field, called the “polarization

field, B 0 P ” is applied to pre-polarize the sample in order to increase signal intensity;

thereafter, the sample is allowed to relax in a second field, called the “relaxation field,

B 0 E ” which can be set to any value, including zero. The field switching time, Swt,

determines the shortest measurable T 1 and so a faster and more flexible electronic method

for switching the B 0 E field is required. Therefore, a low-inductance, air-coil magnet is

used in the relaxometer (Figure 2.16) with power supplies capable of switching the field

electronically to any desired value in a matter of milliseconds while maintaining the high

field stability and homogeneity, at the same time, required by NMR. The magnetic flux

density relevant for relaxation is different from that during signal detection. While the

relaxation field may be varied over several orders of magnitude, the “detection field, B 0 D

is kept fixed at the highest possible value by tuning the RF probe to this particular

detection field at which the signal, an FID generated by the application of a 90 0 pulse, is

detected. [4, 5]


Cooling Solenoid enclosure Magnet Magnet Power Supply and Cooling System Control Circuits Probe for Switching
Magnet Power
Supply and
Cooling System
Control Circuits
for Switching
Temperature Control
RF Unit
Data System

Figure 2.16: A block diagram of the FFC NMR Relaxometer. The „Cooling System‟ and „Magnet Power Supply and Switching Control Circuitry‟ are the units specifically designed for field-cycling purposes. 5 After reference 4

The Fourier transform of this detected FID gives the magnetisation, M (), within the

sample after relaxation for a period, , in B 0 E . [13] The magnetisation decays exponentially

as a function of , starting from the equilibrium magnetisation at B 0 P towards B 0 E . For a

sample containing a single type of spin, the relaxation towards the equilibrium value is

monoexponential with a rate constant R 1 at B 0 E described by the equation: [15]

M () = M E + [M P M E ] exp [-R 1 (B 0 E )]


where M E and M P are the amounts of magnetisation present in the sample during the

relaxation and polarisation periods, respectively. The relaxometer then reduces the data in

each -block to a single value, S () proportional to M () in the curve for a number of B 0

values (Figure 2.17). [15, 17]



Figure 2.17: A monoexponential plot from a relaxation experiment using the “Stelar SMARtracer FFC relaxometer”

Figure 2.17: A monoexponential plot from a relaxation experiment using the “Stelar SMARtracer FFC relaxometer”

In practice, the FFC experiment can be carried out by a number of available pulse

sequences. Four of these were explored in this project: Non-polarised (NP), Pre-polarised

(PP), Non-polarised/Pre-polarised (NP/PP) hybrid and Inversion Recovery (IR) pulse


Pre-polarisation Relaxation Detection D B 0 E B 0 B T x Acq RD Swt
Swt 1
Swt 2
Swt 0

Figure 2.18: The Non-Polarised (NP) Pulse Sequence. After reference 17


The NP sequence (Figure 2.18) is appropriate for T 1 relaxation measurements at

comparatively high relaxation fields (usually above a few MHz). [17] It involves the

preparation of a rather null initial longitudinal magnetisation by allowing the sample to

relax in a null field for a period, RD (recycle delay) which triggers off the sequence. After

a switching time, Swt 1 , the magnet is switched to B 0 E , and is kept constant for the variable

period during which the sample‟s magnetisation grows towards the equilibrium

magnetisation. The magnet is then switched to B 0 D followed by a 90 0 RF pulse (after a

switching time, Swt 2 ) to generate an FID. The field is then switched off and the entire

sequence is repeated.

Pre-polarisation Relaxation Detection
D B 0 P B 0 E B 0 B T x Acq Swt 0
T x
Swt 0
Swt 1
Swt 2
Swt 0

Figure 2.19: The Pre-Polarised (PP) Pulse Sequence. After reference 17

The PP sequence (Figure 2.19) is appropriate for T 1 relaxation measurements at low

fields, possibly, down to zero. [17] In this case, the initial magnetisation is prepared by a

high polarisation field, B 0 P over a considerable long polarisation time, t P this results in a

high initial magnetisation value. After a switching time, Swt 1 the magnet is switched to

B 0 E and is kept constant over a variable period during which the magnetisation decays



towards the equilibrium value. From this stage onwards, the processes involved are

exactly the same as in the NP sequence.

The NP/PP hybrid sequence is a combination of the NP and PP sequences in such a way

that a switching field value is specified (before acquisition) at which the relaxometer

switches from the NP (at high fields) to PP (at low fields) pulse sequences.


to PP (at low fields) pulse sequences. Pre-polarisation Relaxation Detection P B 0 90 B 0
Relaxation Detection
P B 0
sequences. Pre-polarisation Relaxation Detection P B 0 90 B 0 D E B 0 B T





Relaxation Detection P B 0 90 B 0 D E B 0 B T x 0
Relaxation Detection P B 0 90 B 0 D E B 0 B T x 0
E B 0




0 180



P B 0 90 B 0 D E B 0 B T x 0 180 0

t P


Figure 2.20: The Inversion Recovery (IR) Pulse Sequence. After reference 17

The IR sequence (Figure 2.20) involves pre-polarising the sample in the B 0 P field, then

switching to B 0 D where the first 180 0 RF pulse is applied to invert the pre-polarised

magnetisation. The magnet is then switched to B 0 E where the sample is allowed to relax

for a variable period . The field is then switched to a final B 0 D field where a 90 0 RF

pulse is applied to generate the FID. [17]


Optimum performance of a typical FFC NMR relaxometer (especially in keeping the

initial magnetisation constant for all the -blocks) is limited by the following crucial

technical factors irrespective of the pulse sequence in use. [15]

Signal-to-noise ratio (S/N): According to Curie‟s law, M P B 0 P , where M P is the

magnetisation reached in the polarisation field, B 0 P . The induction signal increases

proportionally to B 0 D , whereas noise is proportional to the square-root of the carrier

frequency (i.e. B 0 D ). Therefore:

S / N (B 0 D ) 1/2 M ()


The magnetisation after the evolution period follows:

M () B 0 P exp [-/ T 1 (B 0 E )]

So that S/N is limited by:

S / N (B 0 D ) 1/2 B 0 P

exp [-/T 1 (B 0 E )]



Therefore, field levels B 0 D and B 0 P must be as high as technically possible for maximum


Thermal Stability: The B 0 P field is applied as long as required for attaining thermal

equilibrium, typically, a period t P 5T 1 (B 0 P ). The B 0 D field is applied within a very short

interval, t D required for obtaining an FID. Thus:

t P >> t D



Since the polarisation period is long, and the magnet coils are mostly resistive, joule

heating, and thus thermal drifts of the field, becomes unavoidable during the field cycle.

It is therefore necessary to restrict the B 0 P field to a level still compatible with thermal

stability while the B 0 D field is kept as high as the magnet power supply permits.

Field Homogeneity: Homogeneity within the sample should correspond to the stability of

the B 0 D field. B 0 D must therefore be reproducible within an accuracy of 10 -5 , as the field

homogeneity in the sample does not necessarily need to be better than 10 -5 provided that

a suitable magnet design is used. [15] The commonest cause of inhomogeneous field in a

sample is the sample extending beyond the homogeneous bounds of the RF coil. This can

be avoided by confining the sample to the volume of the RF coil. [13] This is achieved in

practice by filling an NMR tube with a sample to a height not greater than the length of

the RF coil, and positioning the tube with the sample in the centre of the coil.



3: Methodology In the previous chapters, the problem has been defined and background concepts in the

problem domain discussed in relation to other publications. This chapter is a first-step

presentation of how the main aim of the project will be accomplished.

3.1 The FFC NMR Relaxometer and Accessories Measurements of the relaxation rates of the protein samples were done in the relaxometer

shown in Figure 3.1 the principle of operation has been discussed in section 2.6. It

comprises a multifunctional NMR console for carrying out NMR measurements. [23]

NMR console for carrying out NMR measurements. [ 2 3 ] Figure 3.1: “Stelar SMARtracer FFC

Figure 3.1: “Stelar SMARtracer FFC relaxometer” and Accessories













temperature/flow-rate control unit and the main magnet (Figure 3.2).

control unit and the main magnet (Figure 3.2). Figure 3.2: The PC-NMR Console The pulse sequence

Figure 3.2: The PC-NMR Console

The pulse sequence timing and parameter settings of all measurements (and hardware)

are controlled by a powerful NMR software package (Section 3.2) installed on a Pentium

M, 1.6 GHz, 512 MB ram, HD 60 GB single board computer with 4 USB and Ethernet

ports. A high-resolution LCD monitor is connected to the computer for a visual output of

all measurements.

RF pulses are delivered from three (3) independent RF transmitter channels, each being

programmable from DC to 90 MHz. However, only one of these is chosen as the default


transmitter channel; while the other two can be configured, if desired, as supplementary

channels or even as DC outputs for driving external gradients.

NMR signals from samples, after the RF excitation of their spins, are directly acquired by

a digital receiver. Signals detected are usually of low intensity and if not amplified, could

be heavily corrupted by noise. Hence, a preamplifier amplifies these detected signals by

the digital receiver before they are recorded and further displayed on the monitor.

The receiver (or probe) can be tuned, electronically, to signals within a range of 500 kHz

to 90 MHz with a maximum spectral width of 10 MHz. However, it has been tuned for

signal acquisition at a fixed spectral width of 7.2 MHz, ensuring a constant S/N in all

measurements. This fixed detection field (or receiver frequency) enables phase sensitive

detection by the probe, making signal accumulation possible in four separate buffers:

“real and imaginary” and “phase and absolute”.

Temperature of the sample whose relaxation rate is being measured is very important as

far as the accuracy of the measurement is concerned. Therefore, the console is further

equipped with a variable temperature control(VTC) and flow-rate controlunit. These

two controls respectively set fixed temperature and flow-rate of air (which ideally should

be nitrogen gas) driven through an air flux heater to a temperature sensor just below the

placement depth of the sample tube (in the magnet bore). This conditioned air is drawn-in

(through an inlet valve) from the surroundings, sieved and pumped out (through a tube

connecting its outlet valve to the air flux heater) by an electrically operated air pump.

Finally, the main magnet whose B 0 field is being cycled is an air-cored resistive solenoid

through which an electrically controlled AC is passed to generate the required fields in a


typical field-cycling experiment. Joule‟s heat in the magnet is transferred through a high

specific heat capacity liquid (called galden) in heat exchangers to continuously running

tap water. The galden is circulated in the heat exchangers by an electric pump. This

ensures that the magnet is operated near room temperature; in order to avoid considerable

drifts in the field homogeneity. Thermistors are connected to the pipes conveying the

water between the tap and the magnet heat exchangers to monitor the amount of heat

produced from the magnet during operation.

3.2 Software The NMR software package installed on the computer of the relaxometer is called

“AcqNMR” (Figure 3.3). It comprises a huge library of pulse sequences and parameters

for most NMR and NQR experiments. It has an option to be “Run online”, which is best

to select at start up before it is initialised. This option allows saving the results (for

reasons of backups) to a storage space on the network server of Aberdeen University.

There is a multi-page control tab where the acquisition parameters are entered. Most

editable entries in this window essential for the measurements are in the “Main

Parameter” sheet. These editable items include: sample type, experiment (i.e. pulse

sequence type), data acquisition file, maximum T 1 , polarisation time, polarisation field

and relaxation filed. However, since the relaxation field is usually a range, and each pulse

sequence has a unique, specified evaluation format, another window, the “NMRD Profile

Wizard”, under the “Actions” command (on the Windows main menu bar) is used for

these specifications; other parameters are also specified here. The “profile Wizard”

enables the required acquisition parameters in the program window to be “synchronised”

to the NMRD profile parameters before the sequence is run.


Figure 3.3: The main program window of the “ AcqNMR ” software Within the profile

Figure 3.3: The main program window of the AcqNMRsoftware

Within the profile window, the user is also expected to provide a user-specified file name

and location to which the acquisition data will be saved. Files are saved in “*.sef” (stelar

export file) and “*.sdf” (stelar data file) formats, usually accessible in „Notepadbut can

be exported to MS Excel for further editing. The “.sdf” document contains all the

acquisition parameters used by the system for a particular experiment while the “.sef”

document contains the acquired data (arranged in columns) from the experiment.

Once all parameters are set and the “Execute” button on the profile wizard is hit, the

software triggers all other hardware of the relaxometer to start „work‟; except the galden

and air pumps which run continuously, independently of the software. However, the

temperature of the air pumped into the system and the number of scans at any given time


are displayed by the software on the immediate mode bar, just above the multi-page

control tabs.

For complete information on the relaxometer and software, the reader is referred to

references 23 and 24.

3.3 The Field-Cycling MRI System

The phantom was imaged in a home-built, whole-body sized MR scanner (Figure 3.4)

which operates on field compensation method of field-cycling: B 0 D field generated by the

subtraction of variable field strengths from a fixed field by the use of two coaxial

magnets. [25]

a fixed field by the use of two coaxial magnets. [ 2 5 ] Figure 3.4:

Figure 3.4: The Field-Cycling MRI System

The outer, primary magnet (constructed from ferrite) generates a fixed, vertically oriented

B 0 D field of 59 mT. This is a whole-body sized permanent magnet with an approximate


mass of 2600 kg and a clear bore of 65 cm. Gradient coils are constructed into this


The inner, secondary magnet (made from sheets of copper conductor), when energised,

generates an antiparallel variable magnetic field to the primary magnet‟s field; thus,

reducing the net field at the sample. It is a resistive saddle-shaped coil with an outer

diameter of 64 cm and is well-fitted into the primary magnet, creating a free bore

diameter of 52 cm.

The composite support structure of the magnet and the ferrite are non-conducting and so












laboratory is equipped with an air-conditioning unit to maintain an ambient temperature

of 22 ± 0.5 0 C as the ferrite magnet has a relatively large temperature coefficient

(-0.2% 0 C -1 ). The temperature within the ferrite magnet structure itself is stabilised by a

continuously operating fan that circulates air around it. The inner coil is also water cooled

on its outer surface, and a network of platinum resistance thermometers is coupled to it

for monitoring its temperature during operation.

The transmit/receive coil used was a split-solenoid RF coil assembly made from a 6 mm

diameter copper pipe, length 12 cm and diameter 32 cm, tuned to 2.5 MHz. A series of

trimmer capacitors are attached to the coil to allow for resonance matching to be done

with the sample in place. This coil assembly is housed within a cylindrical RF shield.

All operations of the scanner are controlled by a commercial MRI console through

software on an IBM-compatible PC. Gradient and field-cycling waveforms are generated

from the console through four programmable 12-bit digital-to-analogue converters.


Programmable TTL logic waveforms are also generated to enable/disable external

hardware. The console is also capable of RF pulse synthesis (for NMR excitation), pre-

amplification of the NMR signals generated and phase-sensitive detection of the signals.

The console is also capable of image transformation and display.

3.4 Sample Preparation and Handling

The protein aggregate used for the entire project was bovine serum albumen (BSA). It

was prepared with either one or two of the following agents: agarose, phosphate buffer

saline (PBS) and copper sulphate (CuSO 4 ). While the BSA and agarose were in dried

forms, the PBS was diluted according to suppliers assay 1:3 parts of deionised water.

Separate CuSO 4 solutions of the required molarities were prepared by dissolving a

weighed mass of CuSO 4 crystals in 1 litre of deionised water.

mass of CuSO 4 crystals in 1 litre of deionised water. Figure 3.5: From left to

Figure 3.5: From left to right: a volumetric flask of CuSO 4 solution, BSA (in white container), PBS, the composite phantom and in front, NMR tube containing BSA gel


Now that the agents used in preparing the samples for the relaxometry measurements

have been presented, the sample preparation procedure shall now be discussed.

Three different types of BSA gels were prepared: „BSA in PBS‟, „BSA and agarose in

PBS‟ and „BSA in CuSO 4 solution‟. The use of PBS was because it easily dissolves BSA

and also provides similar conditions to proteins in human tissues. All samples were

prepared in glassware thoroughly washed with a powerful decontaminant, rinsed with

deionised water and air-dried.

Preparation of BSA in PBS

The required percentage by weight of BSA was weighed (in grams) on an electronic

balance and carefully transferred into a small beaker of 5 cm 3 of dilute PBS. This BSA-

PBS co-mixture was gently warmed in a water bath at a temperature of 40 0 C while

stirring with a spatula. The temperature was slowly raised to 80 0 C, while stirring

continued. This continued for approximately 10 minutes to achieve complete dissolution

of the BSA in the PBS. 1 ml of the resulting solution was syringed into a NMR tube,

which was further heated (with a smaller beaker covering its open end to prevent

significant vapour loss from the contents) in the water bath. The temperature was raised

to 95 0 C and heating continued for 30 minutes. During this period, the BSA-PBS co-

mixture had denatured to gel.

Preparation of BSA and Agarose in PBS

The required percentages by weight of BSA and agarose were weighed (in grams) on an

electronic balance, separately, and transferred gently into the same beaker of 5 cm 3 of

PBS. The BSA-agarose-PBS mixture was then gently warmed, while stirring with a


spatula, in a water bath at 40 0 C. The temperature was slowly increased to 95 0 C while

stirring continued. A total heating time of 10 minutes, approximately, was allowed for a

complete mixture to be attained. 1 ml of the resulting solution was syringed into a NMR

tube which was heated (with the open end covered with a smaller beaker) at 95 0 C

continuously for 35 minutes, during which period the contents have transformed into gel.

Preparation of BSA in CuSO 4 Solution

The required percentages by weight of BSA were added to 5 cm 3 of CuSO 4 solution in a

small beaker and gently warmed, while stirring with a spatula, in a water bath at 40 0 C.

The temperature was slowly raised to 80 0 C. An approximate period of 10 minutes was

allowed to attain a uniform mixture. 1 ml of the resulting solution was syringed into a

NMR tube which was then further heated (with a smaller beaker covering the open end)

in the water bath at 90 0 C for 30 minutes. During this period, the contents of the tube had

changed to a gel.

Sample Handling

In each of the above cases, after the gel had been formed, the sample tube was removed

from the water bath, wiped dry with a tissue paper, covered with a tight-fitting cork at the

open end and labelled by contents name and preparation date. This labelled tube was

then refrigerated overnight.

Before an experiment with a sample (from the fridge), it was warmed for a period of 10

minutes at room temperature (25 0 C) before placement in the relaxometer; then left for

another waiting period of 10 minutes to ensure complete equilibration of its temperature

just before data acquisition.


The sample volume of 1 ml was crucial so that the sample did not extend beyond the

homogeneous bounds of the RF coil in order for it to experience a homogeneous

magnetic field. The necessity for a homogeneous filed also required that the sample

remained fixed, without movement of any sort, within the relaxometer (Figure 3.6)

throughout the time scale of the experiment.

(Figure 3.6) throughout the time scale of the experiment. Figure 3.6: Sample positioning during a relaxation

Figure 3.6: Sample positioning during a relaxation measurement in the relaxometer. The clamp and load provide a firm grip and stable position of the sample tube against expulsion by the air pressure around it.

3.5 Pulse Sequence Selection and Implementation

Ideally, the same T 1 relaxation rate measures should be obtained regardless of the pulse

sequence used. However, it was necessary to find the most appropriate pulse sequence

that could show least errors in the relaxation measurements, consistency and robustness

to RF pulse fluctuations (due to temperature drifts for example). For this reason, the four

pulse sequences discussed in section 2.6 were each assessed on these bases to arrive at

the best choice of pulse sequence.


Experiments were performed with an initially prepared BSA sample of 20% by weight.

The same parameters were set for each measurement except for the pulse sequence type









deliberately varied



percentage offsets) when experiments to test pulse sequence robustness to RF pulse

fluctuations were to be carried out. Each pulse sequence was applied in four different

experiments; their order of selection for application was done by randomised trials. The

parameters that were kept constant for each pulse sequence were:

Sample temperature ( 0 C):

25.0 ± 0.2

T 1 Maximum value (s):


Repetition delay (s):


Polarisation field (MHz):


Polarisation time (s):


Detection field (MHz):


Relaxation time (s):


Number of -blocks:


Number of averages:


The relaxation field values were chosen over two ranges: 0.5 to 8.0 MHz and 1.5 to 3.7

MHz. All other parameters on the software of the relaxometer are either default or

hardware-specific control/configuration parameters and so were not changed. It was also

possible to type-in the names of the experimenter and sample.

The following were the results obtained with measurements on the 20 % BSA sample

with the four pulse sequences. Each pulse sequence experiment was repeated four times,


with three repetitions per experiment. Therefore, each point plotted in each graph is an

average of twelve measurements. This was done to ensure a high signal-to-noise ratio.

Comparison of the Pulse Sequences in the 1.5-3.7 MHz Range The average standard error in the relaxation rate measurements was evaluated over the

entire Larmor frequency range (in the dispersion profiles shown in Figure 3.7) of all the

sequences. In order of good accuracy, the IR sequence recorded a standard error of

± 0.02 s -1 , the NP sequence recorded ± 0.03 s -1 , the PP sequence recorded ± 0.05 s -1 and

the NP/PP hybrid sequence recorded ± 0.07 s -1 .

5 4.5 4 3.5 3 2.5 1.5 2 2.5 3 3.5 4 R 1 (s
R 1 (s -1 )

Proton Larmor Frequency (MHz)


NP/PP5 4.5 4 3.5 3 2.5 1.5 2 2.5 3 3.5 4 R 1 (s -1

PP/S5 4.5 4 3.5 3 2.5 1.5 2 2.5 3 3.5 4 R 1 (s -1

Figure 3.7: R 1 Dispersion profiles of the four pulse sequences compared in the 1.5 to 3.7 MHz Larmor frequency range

These results showed that the IR sequence was most reliable in the range of Larmor


where the

sequence was the worst.

quadrupole peaks


were expected whiles


NP/PP hybrid

Comparison of the Pulse Sequences in the 0.5-8.0 MHz Range The average standard errors were again evaluated (in the dispersion profiles shown in

Figure 3.8) for the sequences over a wider range of Larmor frequencies, to include

background spin-lattice relaxation processes. In these measurements, the NP/PP hybrid

sequence recorded a standard error of ± 0.26 s -1 , the PP sequence recorded ± 0.28 s -1 , the

IR sequence recorded ± 0.44 s -1 and the NP sequence recorded ± 1.17 s -1 .

8 7 6 5 4 3 2 1 0 0.5 5 R 1 (s -1
R 1 (s -1 )

Proton Larmor Frequency (MHz)

IR8 7 6 5 4 3 2 1 0 0.5 5 R 1 (s -1 )

NP8 7 6 5 4 3 2 1 0 0.5 5 R 1 (s -1 )

NP/PP8 7 6 5 4 3 2 1 0 0.5 5 R 1 (s -1 )

PP8 7 6 5 4 3 2 1 0 0.5 5 R 1 (s -1 )

Figure 3.8: R 1 Dispersion profiles of the four pulse sequences compared in the 0.5 to 8.0 MHz Larmor frequency range

The error values increased among the sequences due to the wider Larmor frequency

range used but in comparison, the NP/PP sequence was the best in this case because the

signal-to-noise ratio in the measurements were actually averaged from a NP sequence

between 8.0 MHz and a sequence switch over frequency of 2.5 MHz; then PP sequence

from 2.5 MHz down to 0.5 MHz, thus showing an overall improved S/N ratio.

However, by separate application of the two sequences, the PP sequence was better than

the NP sequence in this range though both showed higher errors than when combined.


It is the T 1 relaxation of an inverted magnetisation that is recorded in the IR experiment

and so increasing the range of Larmor frequencies for the measurement (and for that

matter, a longer time of measurement) would mean an availability of a smaller transverse

magnetisation towards the lower frequencies. Therefore, the signal recorded was much

corrupted by noise towards the lower frequencies and the relaxometer even had to abort

the experiment at a Larmor frequency of 0.89 MHz.

When the error difference between the two ranges of Larmor frequency was evaluated to

assess the effect on the observed errors in switching from one range to a relatively larger

one, the following results were obtained: the NP/PP sequence recorded a standard error

change of ± 0.19 s -1 , the PP sequence recorded ± 0.23 s -1 , the IR sequence recorded

± 0.42 s -1 and the NP sequence recorded ± 1.14 s -1 .

Based on these error analyses and the understanding of the operations of these sequences,

a decision was arrived at to choose between the IR and PP pulse sequences. This was

because the PP sequence as a stand-alone sequence showed a very close accuracy to the

combined NP/PP sequence which rather was more likely to show larger errors around the

sequence switching over frequency (Appendices C and G) and also put stress on the

switching circuitry. The IR sequence, on the other hand, would be ideal if frequencies

below 1 MHz were not to be explored. Therefore, a further test was carried out to confirm

between which of these two sequences would be robust in situations of RF power offsets.


Comparison of the Pulse Sequences in Respect of Tolerance to RF Pulse Offsets

The 90 0 RF pulse height is fixed in the relaxometer and so the duration (4.5 s) was

deliberately mal-tuned to ± 10%, ± 20% and ± 40% in order to compare the resulting

dispersion profile with that obtained from using the standard 90 0 RF pulse duration.

Dispersion profiles over a range of 0.5 to 3.7 MHz of the IR and PP pulse sequences were

then compared.

With a fixed standard error of ± 1 s -1 (error bars fitted by MS Excel), the profile for the

IR sequence (Figure 3.9) showed consistent correspondence in the mal-tuned data set

with the „standard‟ data set. However, with the – 40% RF offset, significant deviations

were recorded over a Larmor frequency range of 0.89 to 2.09 MHz. The operation was

aborted below this range due to T 1 Max overflow” error; i.e. a considerable decay away

of the magnetisation to be recorded. Apart from this deviation over this range, the IR was

insensitive to RF offsets.

7 6.5 6 5.5 5 4.5 4 3.5 3 2.5 0.5 1 1.5 2 2.5
R 1 (s -1 )

Proton Larmor Frequency (MHz)

-40%4 3.5 3 2.5 0.5 1 1.5 2 2.5 3 3.5 4 R 1 (s -1

-20%4 3.5 3 2.5 0.5 1 1.5 2 2.5 3 3.5 4 R 1 (s -1

-10% 90 deg
90 deg

10%4 3.5 3 2.5 0.5 1 1.5 2 2.5 3 3.5 4 R 1 (s -1

20%4 3.5 3 2.5 0.5 1 1.5 2 2.5 3 3.5 4 R 1 (s -1

40%4 3.5 3 2.5 0.5 1 1.5 2 2.5 3 3.5 4 R 1 (s -1

Figure 3.9: R 1 dispersion profiles with the RF pulse offset compared to the profile with the 90 0 RF pulse in the inversion recovery sequence experiment.


With the same standard error of ± 1 s -1 , the PP sequence (Figure 3.10) showed significant

deviations in response to RF offsets of -20% and -40% over a Larmor frequency range of

1.18 to 3.70 MHz, and also slight deviations in response to + 20% offset over a range of

2.0 to 3.70 MHz.

7 6.5 6 5.5 5 4.5 4 3.5 3 2.5 2 1.5 0.5 1 1.5
R 1 (s -1 )

Proton Larmor Frequency (MHz)

-40%3 2.5 2 1.5 0.5 1 1.5 2 2.5 3 3.5 4 R 1 (s -1

-20%3 2.5 2 1.5 0.5 1 1.5 2 2.5 3 3.5 4 R 1 (s -1

-10% 90 deg
90 deg

10%3 2.5 2 1.5 0.5 1 1.5 2 2.5 3 3.5 4 R 1 (s -1

20%3 2.5 2 1.5 0.5 1 1.5 2 2.5 3 3.5 4 R 1 (s -1

40%3 2.5 2 1.5 0.5 1 1.5 2 2.5 3 3.5 4 R 1 (s -1

Figure 3.10: R 1 dispersion profiles with the RF pulse offset compared to the profile with the 90 0 RF pulse in the pre-polarised sequence experiment.

Therefore, the IR sequence showed enough evidence to be reliable and so was the pulse

sequence of choice in the subsequent experiments to measure the relaxation rates.

3.6 Extraction of R 1 from Background Relaxation Processes

It has been discussed in section 2.5 that the background T 1 relaxation processes in

proteins can be described by a power law: [5, 21]

R p = k L -b



Relaxation rates over a range of 0.5 to 6.5 MHz were taken and the resulting dispersion

profiles were curve-fitted to generate an exponential equation of the form of equation 3.1

using MS Excel (Figure 3.11). Relaxivities at 0.5 MHz, 2.1 MHz and 2.8 MHz were not

included in these plots. The reasons were to avoid the significant error to be imposed on

the measurement at 0.5 MHz by the sequence in use and to also exclude the quadrupole

peaks (which occur at 2.1 MHz and 2.8 MHz) from the background plots.

10 9 8 7 R p = 7.5541 L -0.383 R² = 0.9977 6 5
R p = 7.5541 L -0.383
R² = 0.9977
R p (s -1 )

Proton Larmor Frequency (MHz)

Figure 3.11: Illustration with a background relaxation curve of a 20% BSA sample. The R 2 value

expresses the „goodness of the fit‟

Relaxation rates were measured again over a narrower range (1.5 to 3.5 MHz), with

particular consideration for inclusion of the quadrupole peaks (Figure 3.11).


7.5 7 6.5 6 5.5 5 4.5 4 1.5 2 2.5 3 3.5 R 1
R 1 (s -1 )

Proton Larmor Frequency (MHz)

Figure 3.12: R 1 dispersion profile of the 20% BSA sample showing the quadrupole peaks at 2.1 MHz and 2.8 MHz Larmor frequencies, resulting from the molecular interaction of water protons with protein protons of the immobilised peptide ( 1 H- 14 N) bonds in the protein molecule; R 1 = R p / + R q

The Larmor frequencies (L ) of Figure 3.12 were substituted into the equation of the fit

(of Figure 3.11) to generate another set of relaxation rates, R p / these were the

relaxations due to the background processes that added to the quadrupole processes (R q )

to constitute the R 1 dispersion of Figure 3.12. Subtraction of R p / from R 1 (i.e. the

relaxation rates for the quadrupole peak data set) gave the values of R 1 (or R q ), which

were then plotted versus the L values of the quadrupole peak data set (Figure 3.13).


1.8 A n = 0.5h (a + b) 1.6 A QP = Σ A n
A n = 0.5h (a + b)
A QP = Σ A n
h 2.5
R 1 (s -1 ) / 10 -6

Proton Larmor Frequency (MHz)

Figure 3.13: The quadrupole peaks generated by plotting the relaxation difference (R 1 ) versus the

Larmor frequencies of the quadrupole peak data set. The R 1 axis has been normalised to 10 -6 to cancel out the „mega‟ (10 6 ) unit of the Larmor frequencies so that the areas are in units of per squared second

(s -2 )

The sizes of the peaks were then obtained by „stripping‟ them into smaller trapezia (on

the frequency axis), calculating the area of each trapezium and summing the areas up. In

order to avoid the introduction of significant noise into the computation, the lower tails of

the peaks were ignored; thus, L ranges of 1.8 to 2.5 MHz and 2.5 to 3.0 MHz were

chosen for the peaks at 2.1 MHz and 2.8 MHz respectively.

Therefore, for this 20% sample, the areas under the peaks at 2.1 MHz and 2.8 MHz were

evaluated (Appendix L) to be 0.55 ± 0.03 s -2 and 0.66 ± 0.05 s -2 respectively, giving a

total area of 1.21 ± 0.04 s -2 .


3.7 Imaging Relaxometry Measurements The IR sequence was also used with gradients at the detection field to image the

phantom. In this case, an initial 10 ms adiabatic fast passage was used to provide a

uniform inverted magnetisation. The field was then switched to an evolution value of

59 mT for 30 ms and then returned to a detection field of 59 mT where the signal was

read out, after a 70 ms delay, by a 90 0 pulse. K-space data was collected at five evolution

fields: 59 mT, 62 mT, 65 mT, 68 mT and 71 mT. The data were used to generate T 1

colour images in a Matlab program. Dispersion plots were obtained from regions of

interest to compute R 1 at 65 mT by interpolation of the R 1 values at 59 mT and 71 mT.



4: Experimental Results and Discussions This chapter presents the results of the relaxometry measurements in different protein

concentrations and also image analysis of a phantom based on these measurements. The

same constant parameters outlined in section 3.4 were used in the experiments (except

where specified) and the R 1 values reported therein were evaluated as described in

section 3.5. Ten and thirty points were recorded for the background and quadrupole

curves respectively.

4.1 Correlation between R 1 and Concentration of BSA This section presents results of the measured background effects (Figure 4.1) and

quadrupole effects (Figure 4.2) for five different BSA concentrations. The reason was to

determine the relationship between the peak size and the gel concentrations (Table 4.1).

14 12 10 8 6 4 2 0 0.5 1.5 2.5 3.5 4.5 R p
R p (s -1 )

Proton Larmor Frequency (MHz)

5% BSA12 10 8 6 4 2 0 0.5 1.5 2.5 3.5 4.5 R p (s -1

10% BSA12 10 8 6 4 2 0 0.5 1.5 2.5 3.5 4.5 R p (s -1

15% BSA12 10 8 6 4 2 0 0.5 1.5 2.5 3.5 4.5 R p (s -1

20% BSA12 10 8 6 4 2 0 0.5 1.5 2.5 3.5 4.5 R p (s -1

25% BSA12 10 8 6 4 2 0 0.5 1.5 2.5 3.5 4.5 R p (s -1

Figure 4.1: Background relaxation rate curves of five BSA concentrations








R 1 (s -1 )


5% BSAR 1 (s - 1 ) 6


10% BSA5



15% BSA  4


20% BSA3


25% BSA2










Proton Larmor Frequency (MHz)


Figure 4.2: R 1 dispersion profiles of the five BSA concentrations of Figure 4.1






R 1 (s -1 )/10 -6

5% BSA R 1 (s - 1 )/10 - 6


10% BSA1.5

15% BSAR 1 (s - 1 )/10 - 6 5% BSA 1.5 10% BSA 1   20%



20% BSA 



25% BSA  0.5









Proton Larmor Frequency (MHz)


Figure 4.3: Quadrupole peaks of the five BSA concentrations


The areas under the peaks are summarised in Table 4.1.

BSA Concentration (% by weight)






Total area under quadrupole peaks (s -2 )

0.22 ± 0.01

0.55 ± 0.02

0.84 ± 0.04

1.21± 0.04

1.77 ± 0.07

Table 4.1: Correlation between the area under the peaks and protein concentration of the BSA samples

From these results, it was obvious that the area under the quadrupole peaks is directly

proportional to the concentration of polypeptides in the BSA samples. The linear

correlation is shown in Figure 4.4.

2 1.8 R² = 0.9824 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 5
R² = 0.9824
Area under Peaks (s -2 )

Concentration of BSA (% by Weight)

Figure 4.4: Linear correlation between the total areas under both peaks and the protein concentrations of their respective BSA samples. The same correlation is true between each separate peak and the concentration of the sample (Appendices I and J). R 2 is the „goodness of the fit‟

The gradient of Figure 4.4 was evaluated to be (0.08 ± 0.003) s -2 (%) -1 , representing the

sensitivity of the relaxometer to changes in BSA concentration. In other words, the

relaxometer measures relaxation rates to an order of accuracy of (0.08 ± 0.003) s -2 for

every 1% by weight BSA.


4.2 Effect of Agarose on the Relaxation Processes in BSA

At relatively low concentrations, particularly below 5% by weight, it becomes very

difficult, if not impossible, to de-nature BSA to gel. Therefore, the denaturing process

was enhanced by preparing some low BSA concentrations in a matrix of agarose to help

immobilise the protein molecules. BSA concentrations of 1%, 2% and 5% were prepared,

and separate samples of these concentrations were again prepared with 0.5% agarose in

each. The temperature on the VTC was set to 30 0 C in these and the next block of

experiments because the ambient temperature varied around 28 ± 0.2 0 C during that

period. So it was necessary to set the VTC to a point that could remain stable throughout

the experiments.

Background relaxation curves (Figure 4.5) were subtracted from R 1 dispersion profiles

(Figure 4.6) to generate the quadrupole peaks (Figure 4.7) for these six samples.

2.5 2 1% BSA+ AGR 1.5 % BSA 1 % BSA + AGR 2 1