Daniel Alamidi: Lung imaging using oxygen

Lung imaging using oxygenenhanced MRI in small animals
Master of Science Thesis in Medical Radiation Physics
Daniel Alamidi
Supervisor: Prof., Lars E. Olsson
August 2010
Imaging centre
AstraZeneca R&D
Mölndal, Sweden
Department of Radiation Physics
Sahlgrenska Academy
[email protected]
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
Abstract
BACKGROUND Oxygen enhanced magnetic resonance imaging (OE-MRI) is a novel method
of assessing regional ventilation and oxygen diffusion from the alveoli into the capillaries of
the lung. The paramagnetic nature of oxygen and deoxyhemoglobin in blood shortens the T1
of the oxygenated tissues. MRI can visualize the effect by a signal increase on T1 weighted
images acquired with subjects during breathing of increased oxygen concentration compared
to room air. Data on the utility of this technique in rodents are limited.
OBJECTIVES To develop a protocol for OE-MRI on freely breathing live rodents.
METHODS A T1-measurement protocol for the lung was established. Measurements on both
Nickel-doped agarose gel phantoms and living rat lungs for optimization with respect to
repetition time/inversion time/echo train length at a 4.7 T MRI scanner were performed. An
optimized cardiac triggered inversion recovery RARE imaging sequence was developed. In
order to achieve adequate signal from the lung parenchyma, maintaining practical acquisition
times, and compensating for rapid physiological motion.
RESULTS The optimized parameters from the phantom studies were: repetition time of 6000
ms, 6 inversion time values and an echo train length of 6. The mean T1 was measured with
cardiac triggering to 1682 ± 203 ms (12 %) and 1769 ± 188 ms (11 %) in the right and left
lung, respectively.
CONCLUSIONS An optimized T1-measurement protocol was established for OE-MRI in
rodents. Due to hardware problems that affected the images with ghosting artifacts no
conclusions can be drawn about the oxygen-induced changes.
ii
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
Summary in Swedish (Sammanfattning på svenska)
MR-bildtagning av lungor i smådjur med förhöjd syrenivå
Kroniskt obstruktiv lungsjukdom (KOL) är en folksjukdom som idag enklast diagnostiseras
med hjälp av spirometri. Med en spirometer kan man mäta lungfunktionen men dessvärre inte
erhålla någon detaljerad information om sjukdomens utbredning i lungorna. För att vidare
orientera sig i sjukdomens distribution gör man en lungröntgen eller datortomografi, vilket
utsätter kroppen för joniserande strålning.
Under senare år har Magnet Resonans (MR) kameror gjort stora framgångar vid bildtagning
av lungor. Genom att använda sig av syre som kontrastmedel har en ny lungfunktionsmetod
växt fram, syreförstärkt MR bildtagning. Det fria syret som objektet andas in har en
paramagnetisk egenskap vilket kommer att åstadkomma en signalskillnad i lungvävnaden.
Signaländringen kommer att reflektera lungfunktionen och dess förmåga att överföra syre
från alveolerna i lungan till blodet. Denna bildtagningsmetod har under det senaste decenniet
börjat tillämpas i den kliniska verksamheten med goda resultat.
Syftet med detta arbete var att utarbeta ett MR protokoll för mätning av den signaländring
som uppkommer vid inandning av syre på smådjur. Till en början optimerades protokollet
med simuleringar och mätningar på MR fantom (geler). Därefter gjordes mätningar på fritt
respirerande råttor.
iii
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
Abbreviations, acronyms and symbols
129
Xe
He
81m
Kr
99m
Tc
3D
B0
BOLD
BPM
CO2
COPD
CT
CV
Diffusion
3
DTPA
ETL
FID
FOV
GRE
IDL
IR
M0
MR
MRI
Mxy
Mz
NEX
O2
OE-MRI
Perfusion
PFT
RARE
RF
ROI
SD
SE
SNR
SI
T1
T2
T2*
TD
TE
TEeff
TI
TR
Xenon-129
Helium-3
Krypton-81m
Technetium-99m
Three-dimensional
External magnetic field
Blood Oxygenation Level Dependent effect used in functional MRI
Beats per minute
Carbon dioxide
Chronic Obstructive Pulmonary Disease
Computed Tomography
Coefficient of variation, standard deviation as a % of the mean
Process by which molecules spread from areas of high concentration, to areas of low
concentration
Diethylenetriaminepentaacetic acid
Echo train length
Free induction decay
Field of view
Gradient-echo
Interactive data language, programming software for data analysis
Inversion recovery
Net longitudinal magnetization at equilibrium
Magnetic resonance
Magnetic resonance imaging
Net transverse spin magnetization
Net longitudinal spin magnetization
Number of excitations
Molecular oxygen
Oxygen-enhanced magnetic resonance imaging
Flow of blood to reach an organ or tissue
Pulmonary function test
Rapid Acquisition with Relaxation Enhancement pulse sequence
Radiofrequency
Region of interest
Standard deviation
Spin-echo
Signal to noise ratio
Signal intensity
Time-constant for longitudinal relaxation
Time-constant for transversal relaxation due to spin interactions
Time-constant for transversal relaxation due to a combination of magnetic field
inhomogeneities and spin interactions
Delay time
Echo time of pulse sequence; time between slice excitation and measurement of
signal
Effective echo time, the echo time that contributes the central segment of k-space
Inversion time
Repetition time of pulse sequence; time between two consecutive excitations of the
same slice
iv
Lung imaging using oxygen-enhanced MRI in small animals
Ventilation
Exchange of air between the lungs and the atmosphere
v
Daniel Alamidi
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
Contents
1.
2.
Introduction and aims ...................................................................................................... 7
1.1.
Introduction ................................................................................................................ 7
1.2.
Aims ........................................................................................................................... 7
Theory ................................................................................................................................ 9
2.1.
Respiratory system ..................................................................................................... 9
2.1.1.
Anatomy and physiology of the human respiratory system............................... 9
2.1.2.
Respiratory function......................................................................................... 10
2.2.
The Lung .................................................................................................................. 11
2.2.1.
Lung diseases ................................................................................................... 11
2.2.2.
Lung function measurements ........................................................................... 11
2.2.3.
Rat lung anatomy data ..................................................................................... 11
2.3.
Challenges with MRI of the lung ............................................................................. 12
2.4.
OE-MRI of the lung ................................................................................................. 13
2.4.1.
Principles.......................................................................................................... 13
2.4.2.
Background ...................................................................................................... 14
2.5.
Inversion recovery RARE pulse sequence ............................................................... 14
2.6.
T1 - relaxation .......................................................................................................... 15
2.6.1.
3.
T1 - calculation................................................................................................. 16
Material and methods .................................................................................................... 18
3.1.
Hardware and software ............................................................................................ 18
3.2.
Scan parameters ....................................................................................................... 18
3.3.
Polarity restoration ................................................................................................... 18
3.1.
Phantom studies ....................................................................................................... 19
3.1.1.
General ............................................................................................................. 19
3.1.2.
T1-calculation with Solver and IDL ................................................................ 20
3.1.3.
SNR calculation and reduction ........................................................................ 20
3.1.4.
TR evaluation ................................................................................................... 20
3.1.5.
TI evaluation .................................................................................................... 21
3.2.
In vivo studies .......................................................................................................... 21
3.2.1.
General ............................................................................................................. 21
3.2.2.
Animal preparation .......................................................................................... 21
3.2.3.
T2 measurements ............................................................................................. 22
3.2.4.
T1 measurements ............................................................................................. 22
3.2.5.
T1 measurements with cardiac triggering ........................................................ 22
3.2.6.
T1 measurements with combined cardiac and respiratory triggering .............. 22
vi
Lung imaging using oxygen-enhanced MRI in small animals
3.2.7.
4.
Oxygen-enhanced T1 measurements ............................................................... 23
Results .............................................................................................................................. 24
4.1.
Phantom studies ....................................................................................................... 24
4.1.1.
T1-calculation with Solver and IDL ................................................................ 24
4.1.2.
SNR reduction .................................................................................................. 25
4.1.3.
TR evaluation ................................................................................................... 25
4.1.4.
TI evaluation .................................................................................................... 26
4.2.
5.
Daniel Alamidi
In vivo studies .......................................................................................................... 27
4.2.1.
T2 measurements ............................................................................................. 27
4.2.2.
T1 measurements ............................................................................................. 27
4.2.3.
T1 measurements with cardiac triggering ........................................................ 28
4.2.4.
T1 measurements with combined cardiac and respiratory triggering .............. 28
4.2.5.
Oxygen-enhanced T1 measurements ............................................................... 28
Discussion ........................................................................................................................ 29
5.1.
Polarity restoration ................................................................................................... 29
5.2.
Scan parameters ....................................................................................................... 29
5.3.
Phantom studies ....................................................................................................... 30
5.3.1.
SNR reduction .................................................................................................. 30
5.3.2.
TR evaluation ................................................................................................... 30
5.3.3.
TI evaluation .................................................................................................... 30
5.4.
In vivo studies .......................................................................................................... 31
5.4.1.
T2 measurements ............................................................................................. 31
5.4.2.
T1 measurements with cardiac triggering ........................................................ 31
5.4.1.
Oxygen-enhanced T1 measurements with combined triggering ..................... 31
Acknowledgements .................................................................................................................. 32
References ................................................................................................................................ 33
vii
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
1. Introduction and aims
1.1.
Introduction
One of the most common lung diseases is chronic obstructive pulmonary disease (COPD).
COPD is characterized by limitation of airflow during expiration due to emphysema, chronic
bronchitis or both and is in a decade expected to be the third-leading cause of death
(Devereux G 2006). This reduction in outflow of air during expiration can be measured by
pulmonary function tests.
Spirometry (measuring of respiration) is the most straightforward method of measuring lung
function and investigating diseases such as COPD. Nevertheless, this technique is incapable
of identifying regional distributions and the location of pulmonary disorders. Hence, a threedimensional (3D) method is required, i.e. imaging. Inhalation of radioactive gases is a
method for lung imaging but with some significant limitations; the radiation dose from the
radioactive substances and poor spatial resolution.
The most common way of detecting lung abnormalities is with a standard chest x-ray. A
better insight of pulmonary diseases can be achieved by computed tomography (CT) imaging.
It is valuable for the evaluation of morphological changes and regional pulmonary functional
tests. However, CT-imaging does not supply any functional information of the lungs and it
exposes the body to ionizing radiation.
MRI of the lung is challenging since the lung has low proton density and therefore creating a
low signal. It furthermore contains air-tissue interfaces generating susceptibility artifacts.
Additionally, the influence of respiratory and cardiac motion has to be controlled to avoid
motion artifacts. Several methods have been proposed in order to overcome these difficulties.
They consist of breath hold imaging, respiratory and cardiac triggering procedures and use of
pulse sequences with very short echo times (TEs).
In addition, MRI with hyperpolarized noble gases such as Helium (3He) and Xenon (129Xe)
render direct MR visualization of gaseous possible. Excellent results have been achieved for
lung imaging, but the costs for additional equipment and production of noble gases currently
limit its broad application.
A new technique in development devoted to lung functional imaging is oxygen-enhanced
MRI (OE-MRI), inhaling of molecular oxygen as a contrast agent, which enhances the signal
of the protons in the pulmonary capillaries. This method provides means to directly study
oxygen uptake of the lungs and is useful for the diagnostic of respiration related diseases.
1.2.
Aims
The overall aim of this project was to perform oxygen-enhanced MR imaging on living
rodents during free breathing. Throughout the task, a T1-measurement protocol for the lung
was established. To achieve these goals, the project in developing this new technique for
rodents was divided into three main aspects:
Initially gel phantoms were used to optimize the parameters of a pulse sequence made for T1relaxation time measurements. The signal to noise ratio (SNR) in vivo in lung is lower than in
phantoms due to the low proton density. The SNR was thus virtually reduced in the phantoms
7
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
by adding noise to the images with help of simulations to investigate the results in a realistic
environment.
In the second part of the project the optimized sequence was verified in vivo in lung tissue of
rats. The accuracy and precision of T1 was improved with cardiac triggering and a respiratory
technique that blocked the acquisition of the signal during the inspiration of the animal.
The final part of the project considered the examination of oxygen induced changes in T1
relaxation time of lung parenchyma in rodents.
8
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
2. Theory
2.1.
Respiratory system
2.1.1. Anatomy and physiology of the human respiratory system
The human body is dependent on a constant supply of oxygen (O2) to every single cell of the
organism. Respiration is the physiological process by which organisms supply O2 to their
cells and the cells use that O2 to produce high energy molecules. A respiratory system
consists of conducting airways: trachea, bronchi, bronchioles and terminal bronchioles. The
acinar airways, where actual gas exchange occurs, includes transitional bronchioles,
respiratory bronchioles, alveolar ducts and alveolar sacs (Faller A et al. 2004) (Figure 1).
Figure 1. Detailed illustration of the lung and the respiratory system with conducting and
acinar airways (Mayo M L 2009).
The alveolar sacs are closely packed air sacs, like individual grapes within a bunch (Figure
2). The lung can be regarded as a collection of these small 500 million bubbles. The
individual alveoli are tightly wrapped in blood vessels, allowing gases in the alveoli to easily
diffuse into the blood.
Figure 2. A schematic view of alveoli (Faller A et al. 2004).
9
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
Fresh air is taken in via the upper airways (the nasal/oral cavities, pharynx and larynx)
through the lower airways (trachea, primary bronchi and bronchial tree) and into the small
bronchioles and alveoli within the lung tissue. The upper airways are the means of
transportation of the inspired and expired air, and they warm it, humidify it, purify it and
regulate it (sense of smell). The trachea is divided into two equally sized bronchi, which in
turn diverge into two daughter branches (Hlastala M P et al. 2001).
Additionally, the lung filters unwanted materials from the circulation and acts like a reservoir
for blood. The lung structure has an intricate structure; it is elastic and remarkably durable
(Schwartzstein R M et al. 2005, West J B 2008).
2.1.2. Respiratory function
The breathing mechanism is driven by the diaphragm; the thin, dome-shaped muscle at the
base of the thoracic cavity. During inspiration, the volume of the thoracic cavity increases
and air is drawn into the lung. The volume increase causes the internal pressure of the chest
to become lower than atmospheric pressure, resulting in a flow of air into the airways. The
driving forces for gas exchange between the lung and the environment are the pressure
differences.
The processes of internal respiration concern the exchange of O2 and carbon dioxide (CO2)
between blood and cells in different tissues. On the contrary, external respiration is the
process by which outer air is drawn into the body in order to supply lungs with O2, and
“used” air is expelled from the lungs in order to remove the CO2 from the body.
The oxygen transport can be divided into three steps: (1) Ventilation, or breathing, involves
the physical movement of air in and out of the lungs; (2) Gas diffusion, exchange of gases
between the alveoli and the pulmonary capillaries (Figure 3); (3) Perfusion, circulation of
blood between the lungs and organs.
Figure 3. Gas exchange in the lung. The exchange of gases in the lung
transforms deoxygenated (poor in oxygen) venous blood that is rich in
CO2 into oxygenated (rich in O2) arterial blood with low CO2 content
(Faller A et al. 2004).
10
Lung imaging using oxygen-enhanced MRI in small animals
2.2.
Daniel Alamidi
The Lung
2.2.1. Lung diseases
Many of the properties and mechanics governing lung function are impaired in case of a
disease. One of the most common lung diseases is COPD. It is currently the fourth leading
cause of mortality in the western world. With an increase in smoking in developing nations;
COPD is expected to be the third-leading cause of death worldwide within ten years. The
disease is associated with several risk factors where cigarette smoking clearly is the most
significant (Devereux G 2006).
COPD is characterized by limitation of airflow during expiration due to either emphysema,
chronic bronchitis or both. Emphysema is a condition of the lung that destroys the alveolar
walls, resulting in loss of elasticity of lung tissue. Chronic bronchitis is the inflammation of
the bronchi in the lungs and destruction of structures supporting the alveoli. These symptoms
lead to a reduction in the outflow of air during expiration that can be measured by pulmonary
function tests.
In pulmonary diseases, like COPD, alterations in the ventilation and/or perfusion progresses
and impaired oxygen diffusion from the alveoli into the capillaries are the main cause of
respiratory disorders. These abnormalities can be detected by pulmonary function test
measurements (Devereux G 2006).
2.2.2. Lung function measurements
In case of a disease in the lung, pulmonary function tests (PFTs) can be vital for measuring
lung function. Spirometry is the most common way of measuring lung function and
investigating diseases such as these. The volume and/or flow of air that can be inhaled and
exhaled is examined. However, this method only provides information on a global scale,
when disorders of the lung have reached more advanced stages. PFT is a relative insensitive
method.
Lung ventilation scintigraphy is an imaging method that uses inhalation of radioactive gases
such as 133Xe, 81mKr or 99mTc- labeled diethylenetriaminepentaacetic acid (DTPA). The
pulmonary ventilation function is evaluated with a gamma camera but the radioactive dose
and poor spatial resolution limits the use of this method.
Another common image modality of choice for detecting lung abnormalities is a chest x-ray.
A better insight of the lung disease is attained with CT-imaging as it rapidly provides more
detailed information. Nonetheless, CT-imaging does not provide any functional information
of the lungs and there are restrictions related to CT, as the body is exposed to ionizing
radiation with each imaging session (Mills et al. 2003).
MRI of the lung has recently provided excellent results for evaluation of ventilation by
utilization of hyperpolarized noble gases (Ohno Y et al. 2007). An alternative approach to
hyperpolarized gas MRI is a new technique devoted for functional lung imaging, OE-MRI.
2.2.3. Rat lung anatomy data
The respiratory system of rats is similar to that of humans, but with smaller dimensions and
higher breathing frequency. Rodents have high respiratory rates, 71-146 breaths per minute,
and high cardiac rates, 320-480 beats per minute (BPM), compared to human beings with 1211
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
70 breaths per minute and 70 BPM (Grant K 2009, Pass D et al. 1993). The total lung
capacity (TLC) of the rat is about 10 ml compared to 6 l of a human. There are 4 lobes in the
right rat lung (3 in human right lung) and only a single lobe in the left lung (2 lobes in human
left lung).
Further, the parenchyma of the rat lung occupies double as large fraction of the total lung
than the human (rat: 24%, human: 12% lung volume). The blood-gas barrier thickness,
diffusion distance in the rats (0.40 μm) is somewhat smaller than that of the human (0.62
μm). The thickness might have important consequences for both gas exchange and lung
mechanics. In case of an inflammation of the lung tissue, plasma leakage into the alveoli
increases the diffusion distance and impairs alveolar gas exchange.
Lastly, rat lungs have fewer respiratory bronchioles and airway generations than human lungs
do. Lung anatomy comparison between rodents and humans are presented in Table 1 (Grant
K 2009, Irvin C G et al. 2003, Lindstedt S L et al. 2002, Sahebjami H 1992).
Table 1. Lung anatomy comparison between rodents and humans.
Respiratory rate
Cardiac rate
Total lung capacity (TLC)
Right lobes in lung
Left lobes in lung
Fraction of parenchyma of total lung
Blood-gas barrier thickness
Airway generations
2.3.
Rodent
71-146
320-480
10 ml
4
1
24 %
0.40 μm
Single generation
Human
12-70
70
6,000 ml
3
2
12 %
0.62 μm
Several generations
Challenges with MRI of the lung
MR Lung imaging is facing many difficulties because of the morphology, physiology and
composition of the lung. Due to the air within the lung, it has inherently low average proton
density (approximately 20-30 % of the soft tissue) generating weak signal, resulting in low
SNR (Dietrich O 2009).
Additionally, the signal is hampered by multiple air-tissue interfaces within the alveoli in the
lung. The heterogeneous microstructure of lung parenchyma generates large local variations
of susceptibility within small spatial scales between paramagnetic oxygen in air and
diamagnetic tissue. These susceptibility variations influence the magnetic fields within the
lung that rapidly dephase the already low MR signal, resulting in a very short T2*. Therefore
sequences other than conventional Gradient-Echo (GRE) should be used for lung MRI, for
example a Turbo Spin-Echo (TSE) sequence with a short echo spacing (Kauczor H-U et al.
1999).
Moreover, signal distortions due to cardiac pulsation and respiration motion are a major
problem in lung MRI. This is particularly apparent in small rodents, because of their higher
cardiac and respiratory rates. To overcome these difficulties many strategies have been used.
They consist of breath hold imaging, respiratory and cardiac triggering procedures and use of
pulse sequences with extremely short TE (Beckmann N et al. 2007).
12
Lung imaging using oxygen-enhanced MRI in small animals
2.4.
Daniel Alamidi
OE-MRI of the lung
2.4.1. Principles
OE-MRI was first proposed in 1996 for evaluation of regional ventilation using molecular O2
as a contrast agent (Edelman R R et al. 1996). When compared with hyperpolarized gas
imaging (see theory section 2.2.2), OE-MRI offers several benefits: the oxygen ventilation
technique provides ways to directly study oxygen uptake from the air space to the pulmonary
blood system; oxygen is easily reached as a part of emergency equipment in most MR suites;
OE-MRI is very cost-effective since it does not require any supplementary expensive
equipment (Mai V M et al. 2005).
The underlying principle of OE-MRI is the weakly paramagnetic property of molecular
oxygen caused by the presence of two unpaired electrons in their outer shells. The contrast
mechanism works similar like a gadolinium-based contrast agent, but with a smaller
magnitude: i.e., the T1 of the protons in blood is shortened depending on the O2 concentration
(Edelman R R et al. 1996).
Accordingly, oxygen-enhanced lung imaging can be regarded as imaging of lung function, as
it provides information about three physiological parameters: The inhaled oxygen must be
transferred to the lung region; subsequently adequate ventilation of the area is an essential
condition for oxygen-induced reduction of T1 relaxation. The intermediate step regards the
exchange of oxygen between the alveoli and the pulmonary capillaries, i.e. diffusion, is
required for signal enhancement. As a final point, fresh capillary blood must be supplied in
which the oxygen can be solved; thus, lung perfusion is for that reason a third constraint for
the observation of reduced T1 values (Edelman R R et al. 1996, Loffler R et al. 2000).
MRI with hyperpolarized gases and perfusion measurements with Gadolinium-DTPA are
performed to provide information about ventilation and perfusion, respectively. OE-MRI has
the potential to bring out information of these two physiological parameters as well as
providing diffusion information of the gas exchange in lung.
Oxygen-induced relative T1 reductions between 7 % and 14 % have been observed after
inhalation of pure oxygen. The signal within each single voxel is averaged over various kinds
of tissue such as blood, vessels, alveolar cells and surrounding tissue (Nakagawa T et al.
2001, Dietrich O et al. 2009).
Subsequently, during gas exchange oxygen diffuses across the alveolar membrane into the
pulmonary capillary blood and initially dissolves into blood plasma as molecular oxygen.
Oxyhemoglobin is generated when the oxygen molecules couples with hemoglobin. It is
known that oxyhemoglobin is diamagnetic and that deoxyhemoglobin is paramagnetic. This
means that deoxygenated blood has a shorter T2* and hence lower MR signal than fully
oxygenated blood in gradient-echo (GRE) sequences with its strong T2* weighting. Blood
oxygenation level dependent (BOLD) contrast is based on this effect of deoxyhemoglobin
and is frequently used in functional MRI (McRobbie D W et al. 2003).
After pure oxygen inhalation the molecular oxygen will dissolve and the hemoglobin will be
saturated with oxygen. Consequently the concentration dissolved oxygen in the blood
increases approximately fivefold. Hence the most important component for the T1 effect is
caused by the increased concentration of dissolved oxygen in the capillary blood of the lung.
13
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
This effect can be detected by T1-weighted MRI sequences as regions of increased signal
intensity (Tadamura E et al. 1997, Watt K N et al. 2008).
In addition, an increase in oxygen concentration within the blood results in a prolongation of
T2* when T1 is measured in lung. However, this has only a minor impact on the signal
intensity in T1 weighted sequences (Ohno Y et al. 2007).
2.4.2. Background
Throughout the last decade several researchers have declared the potential of using OE-MRI
in clinical applications for evaluation of lung function. Oxygen induced MRI for patients with
pulmonary diseases has been successfully established which demonstrates the significance of
oxygen enhancement. This includes ventilation abnormalities and emphysema, as well as
assessing parameters of pulmonary function measurements for patients with lung cancer
(Loffler R et al. 2000, Edelman R R et al. 1996, Chen Q et al. 1998).
Moreover, OE-MRI was found to be as effective as quantitative CT for smoking-related lung
functional loss assessment and stage classification concerning patients with smoking-related
COPD (Ohno Y et al. 2008).
In contrast, regarding preclinical animal studies the publications are very limited and further
experimental and theoretical work is required. In mice only one study can be found (Watt K
N et al. 2008), and not a single work on rats.
Furthermore, external magnetic fields (B0) over 3 T are preferred to improve the SNR when
small animals are studied. This induces problems for lung studies since the susceptibility
variations are proportional to B0. In addition, small animals have high breathing and cardiac
frequencies, which leads to signal distortions.
2.5.
Inversion recovery RARE pulse sequence
A basic Spin-Echo (SE) sequence has a long acquisition time since only one line of k-space
data is collected after an excitation. However a SE image can be acquired with dramatic time
saving by utilization of a Turbo SE (TSE) sequence. TSE is a commercial version of Rapid
Acquisition with Relaxation Enhancement (RARE) sequence, which collects more than one
line of data for each excitation by producing a train of echoes. The echo train is formed by
multiple refocusing, 180˚, radiofrequency (RF) pulses (Figure 4).
The acquisition time is proportional to the echo train length (ETL), i.e. the number of echoes
acquired. For instance if 10 echoes are individually phase-encoded in RARE, the total
acquisition time is decreased by a factor of 10. This echo train formation can be continued as
long as sufficient transverse magnetization remains to form an echo, i.e. as long the T2
relaxation permits (McRobbie D W et al. 2003, Vlaardingerbroek M T et al. 1996).
For enhancement of T1 contrast and T1 relaxation measurements an inversion pulse is
implemented to the RARE sequence (Vaninbroukx J et al. 2003).
14
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
Figure 4. Pulse sequence diagram of a RARE sequence with a train of echoes
formed by multiple refocusing pulses. The top line, RF, shows the applied
radiofrequency pulses, GS the slice-selective gradient, GR the readout gradient
(frequency-encoding). Centrically reordered phase encoding is used to produce
the highest SNR (Stock W K et al. 1999).
2.6.
T1 - relaxation
T1 relaxation (also known as thermal, longitudinal or spin-lattice relaxation) describes the
recovery of the longitudinal magnetization to its initial value at thermal equilibrium (M0)
along B0, just after applying an RF excitation pulse. It results from interactions amongst the
water protons and protons attached to the surrounding molecules (the lattice) with fluctuating
magnetic dipole moments at the larmor frequency.
There are two energy states the hydrogen nucleus can occupy in the presence of B0; up, the
lower energy state where the magnetic dipole moment points along B0, or down, the higher
energy state where the magnetic dipole moment points opposite B0. At M0, thermal
equilibrium, a slightly increase of dipoles is observed in the lower energy state along B0. In
the excitation process, after a 90° pulse, there is an equal population of hydrogen dipoles in
the lower and higher energy states.
To give away energy and return to the lower energy state, the protons interact with the lattice,
which can absorb the energy. In order to enable this energy transfer, the magnetic dipole
moments of the neighboring protons or other nuclei or molecules has to fluctuate at the
larmor frequency and thereby satisfy the resonance condition. Due to the fluctuating fields
the spins can change from high to low energy states through interaction with the lattice, and
contribute to a relaxation in magnetization. T1 is defined as the time it takes for Mz to recover
to a value about 63 % of M0 after a 90˚ excitation of the longitudinal magnetization.
Accordingly, longitudinal relaxation can occur only when a proton encounters another
magnetic field fluctuating near the Larmor frequency and therefore, different amounts of
spins will contribute to T1 relaxation in different types of tissue. Information about the
mobility of molecules, particularly water molecules, and hence the binding of water
molecules, for example macromolecules can be generated from T1.
In addition there are several factors affecting T1: B0; free water content; lipid and
macromolecule content; molecular motion; viscosity; temperature. For instance, the T1
relaxation time is prolonged as B0 increases.
15
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
2.6.1. T1 - calculation
The gold standard to measure T1 is by a series of inversion recovery (IR) acquisitions with
varying inversion times (TI). The IR method may be describes as:
Prepare (180˚) - TI - 90˚ (detection) - TD
In its simplest form the first part within an IR sequence involves inverting the initial
magnetization (which is initially in its equilibrium state) with an 180˚ preparation pulse. After
this pulse the longitudinal magnetization will recover exponentially during TI at a rate
described by T1. Subsequently after TI, acquisition of the signal is performed during a
readout time, d. The readout sequence is typically a SE such as RARE (Figure 4), which is
used in this study. Hence a 90˚ RF pulse is applied, which tips the current longitudinal
magnetization into the transverse plane, where it after the refocusing pulses (180˚ pulses)
gives rise to an echo and a free induction decay (FID). The sequence is then repeated after a
delay time TD (Figure 5).
Figure 5. Pulse sequence scheme of the inversion recovery RARE
sequence where d, is the readout time during the readout and Td, is
the delay time after the acquisition before the next IR pulse.
The sequence is repeated several times with different TIs. In this way the recovery curve can
be sampled where the signal intensity (SI) is plotted versus the TI (Figure 6). The time
dependence of the SI and longitudinal magnetization (Mz) is described as:
𝑆𝐼 ∝ 𝑀 = 𝑀 [1 − (1 − 𝑀 /𝑀 ) exp (− 𝑇𝐼/𝑇 )]
[2.1]
where 𝑀 is the magnetization at the beginning of the 𝑇𝐼 period. This equation can be
developed and simplified if TR is held constant:
𝑆𝐼(𝑇𝐼) ∝ 𝑀 = 𝐴 − 𝐵𝑒
/
[2.2]
where 𝑆𝐼(𝑇𝐼) is the signal intensity (~𝑀 ) at time 𝑇𝐼, 𝐴 is a constant for the offset (𝑀 ) and
𝐵 is a constant for the proton density and 𝑀 (Kingsley P B 1999, Gupta R K 1980). The
three constants 𝐴, 𝐵 and 𝑇 are obtained by minimizing the difference between the observed
data and the calculated values, i.e., a three-parameter nonlinear least square method fit
generates the values of the three constants. For perfect pulses only a two-parameter fit is
needed to fit the data because 𝑀 /𝑀 = −1 for fully relaxed IR (Levy G C et al. 1975).
16
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
100000
SI [a.u]
50000
0
2000
4000
6000
TI [ms]
-50000
-100000
Figure 6. Inversion recovery plot of SI within a circular region of interest
(ROI) at a gel phantom with a T1 of 1500 ms. 8 TI points (of 100, 300, 400,
500, 1200, 2000, 3500 and 5000 ms) were used to sample the curve. The
standard deviations (SDs) are smaller than symbols.
17
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
3. Material and methods
3.1.
Hardware and software
All the experiments were performed on a 4.7 T Bruker MRI system (Bruker Biospin 47/40,
Ettlingen, Germany) and operated from Linux computers running Paravision 5.0 (Bruker,
Ettlingen, Germany). The gel-containing vials and living rodents were scanned using a 72
mm quadrature coil (Model no: 1P TP9455, Serial no: S 0027).
3.2.
Scan parameters
An inversion recovery RARE pulse sequence (section 2.5) was used for the phantom and in
vivo studies with following parameters: TE = 3.52 ms; bandwidth = 200 kHz; field of view
(FOV) = 7 x 6.5 cm; matrix size = 196 x 256; number of excitations (NEX) = 2; and slice
thickness = 5 mm. The frequency and phase encoded pixel sizes were 0.36 mm/pixel and
0.25 mm/pixel respectively. Initially, an ETL of 4 was used and a non-selective inversion
pulse was applied.
Centrically reordered phase encoding was utilized that collects the center of the k-space
during the first part of the readout interval, resulting in a short effective echo time (TEeff)
(Figure 4). The lines around the center of the k-space determine the contrast of the acquired
image, while the outer part of the raw data matrix provides information on details of the
image (Zhou X J 2004a). A short TEeff is desired to avoid T2 weighting. Centrically reordered
phase encoding provides higher signal intensity, a shorter TEeff and a higher SNR.
3.3.
Polarity restoration
The phase information of the transverse magnetization is ignored when magnitude MR
images are used. As a result the sign is lost from the inversion recovery curve (Figure 7). It is
preferable to restore the polarity of the inversion recovery data, as this will reduce the
variance in the estimation of T1 (Gowland P A et al. 2003, Zhou X J 2004b). In this project
magnitude images were sampled, hence a straightforward polarity restoration technique was
implemented for the T1 calculations.
To begin with, the minimum point at a magnitude curve was obtained, the lowest signal value
A in Figure 7. The polarization of the TIs less than or equal to this minimum point was
switched, resulting in a polarity-restored IR curve, hence point A was moved to point B in
Figure 7. This procedure was also performed for the points ± 1 steps from the minimum
point, the values within the circle in Figure 7. Accordingly three different inversion recovery
plots were acquired. The sum of squared error at the three recovery curves was calculated,
and the curve with smallest amount of inaccuracy was selected.
Data analysis with standard interactive data language (IDL) (version 6.4, ITT Visual
Information Solutions, Boulder, CO, USA) software was performed. ROIs were positioned
on a set of IR magnitude images acquired with different inversion times (Figure 8) and
polarity-restored IR plots were performed.
18
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
100000
Polarity-restored data
Magnitude data
Signal [a.u]
50000
0
A
B
-50000
-100000
0
2 000
4 000
6 000
TI [ms]
Figure 7. Polarity-restored and magnitude IR curves for a gel phantom. The arrow
points at the lowest signal value (A) on the magnitude curve. TIs less than or equal to
the minimum point (A) were polarity-restored; the sign of the minimum point A was
switched, resulting in a new position B. The points within the circles, ± 1 steps from the
minimum point were also polarity restored, to decide the optimal IR curve for the T1
assessment. The SD is less than 1 % and is therefore too small to present graphically.
3.1.
Phantom studies
3.1.1. General
Phantom studies were performed to optimize and establish timing constrains for a pulse
sequence scheme. The sequence design was subsequently used for the in vivo experiments.
The particular aim was to keep the scan time as low as possible and acquiring T1 values with
good accuracy and precision. The main parameters optimized in the sequence were the length
of the repetition time (TR) and the number of TI values used for the T1 assessment.
Figure 8. Circular ROIs placed at a 5 mm single slice
axial MRI image of 8 gel phantoms with T1 values of
1000 – 1800 ms. The rectangular ROI in the right corner
represent the noise area for the SNR calculation.
19
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
Eight nickel-doped agarose gel phantoms in vials with T1 values of 1000 – 1800 ms were
used (Alamidi D et al. 2009). Earlier experiments in rat lung confirmed this range of T1values. The vials were scanned in an axial projection with a slice thickness of 5 mm (Figure
8).
3.1.2. T1-calculation with Solver and IDL
To assure that the estimation in IDL was acceptable, T1 calculations at the eight gel
phantoms with the Solver tool in Microsoft Excel software were done. The function of the
Solver is to adjust parameters with the least square method, similar to the IDL software. The
Solver tool use modulus data, hence the absolute value of equation [2.2] is applied. Mean
signal intensity values were obtained from circular ROIs placed on the gel phantom images
(Figure 8).
Estimation of T1 over the eight vials with T1 values of 1000 – 1800 ms was also done with
the IDL software for evaluation between the two methods. T1 maps were generated with the
polarity-restored data in the IDL software. A pixel-by-pixel three-parameter nonlinear least
square T1 method fit was applied (section 2.6.1), i.e., T1 was calculated for each pixel.
Circular ROIs were drawn on the T1 map to obtain the mean T1 values of the pixels.
The inversion recovery RARE sequence with a constant TR was used and the TI values were
100, 300, 400, 500, 1200, 2000, 3500 and 5000 ms for the both methods.
3.1.3. SNR calculation and reduction
For the T1 assessments, a set of TI values were used. The image with longest TI, i.e. with
highest signal, was selected for the SNR calculation. The noise was determined by placing a
rectangular ROI in the noise area of the image (Figure 8). A single circular ROI to obtain the
signal intensity was placed in the phantom with a T1 value of 1500 ms (earlier results have
shown that T1 in rat lung is around 1500 ms). The SNR was calculated by dividing the mean
signal in the circular ROI with the mean signal in the noise.
The SNR in vivo is lower than in phantoms because of the low proton density and signal in
lung. To investigate the results in a realistic situation SNR was reduced in the phantom
images using simulations with standard IDL software. The purpose of the simulations was to
optimize the sequence for in vivo studies, and to keep the scan time as low as possible.
Stochastic noise that decreased the existing SNR to a ratio of 40 and 20 was added to the
images. The noise was defined as a Gaussian distribution of random numbers, with a mean of
zero and a SD of one. T1 was estimated by a three-parameter nonlinear least square method
fit (section 2.6.1) in the IDL software.
3.1.4. TR evaluation
T1 measurements with different TRs of 3000, 4000, 6000 and 8000 ms were performed to
obtain the relationship of TR and the precision of T1. The TIs were adjusted with a range
from 100 ms up to the respective TR to receive a full relaxation curve (Table 2). Setting TR
significantly longer than the longest TI increases the measurement time without improving
accuracy or precision (Kingsley P B et al. 2001). Accordingly, TR was set just slightly longer
than the longest TI.
20
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
Table 2.
TR and TI values for the T1 phantoms measurements.
TR [ms]
8000
6000
4000
3000
TI [ms] --->
7000
6000
5000
5000
3500
3500
3500
2000
2000
2000
2000
1200
1200
1200
1200
500
500
500
500
400
400
400
400
300
300
300
300
100
100
100
100
3.1.5. TI evaluation
The correlation of the number of TI values with the T1 precision was studied when the TR
parameter was established. With fewer TI points the scan time is improved. The experiment
started with an acquisition of 8 TI points and was consecutively reduced to 3 TI points. A
constant TR of 6000 ms was used for all measurements. After the acquisitions, simulations
and T1 calculations in IDL were performed.
3.2.
In vivo studies
3.2.1. General
The optimized inversion recovery RARE pulse sequence design was verified in vivo in rat
lungs to establish the timing constrains for the sequence. The in vivo studies were performed
with the same geometry and parameters as the phantom studies, i.e. in an axial projection
with a slice thickness of 5 mm. The slice was carefully positioned slightly over the diaphragm
to include as much lung parenchyma as possible.
T1 maps were calculated in the same way as for the phantom studies. Mean T1 values from
ROIs placed on the relaxation maps in lung were obtained. The position and shape of the
ROIs in which the maps were determined were carefully chosen to avoid contributions from
large vessels.
The relation between ETL and SNR was examined with the anticipation to decrease the scan
time. To improve the accuracy and precision of the T1 calculations cardiac triggering was
performed.
3.2.2. Animal preparation
Wistar rats (280 ± 35 g, mean ± SD) were prepared in accordance with the guidelines
established by the Animal Ethics Committee of Göteborg’s University, with the ethical
approval number 400-2008.
Rats were lightly anaesthetized with isoflurane (5 %, Abbot Scandinavia, Solna, Sweden) for
1-2 min and immediately injected with anesthetic solutions Zoletil (40 mg/kg, VIRBAC S.A.
Titulaire de I’AMM, 06516 Carros, France) and Domitor (0.4 mg/kg, Orion Pharma, Espoo,
Finland). Next the animals were transferred to a dedicated rat holder (Bruker, Ettlingen,
Germany) in supine position, head up, and allowed to breathe room air or pure oxygen
spontaneously. The bed was inserted to the center of the scanner with the lungs in the center
of the coil.
21
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
Cardiac and respiratory monitoring systems (SA instrument, Inc) were used throughout the
imaging. A pneumatic pillow was taped on the rats’ abdomen for respiratory monitoring. The
core temperature was measured with a rectal thermo sensor and was maintained at 37˚C ±
1˚C by means of heated water tubes.
The bed was equipped with a fluid based (fluorocarbon) heating system, which prevented the
animals from cooling down during imaging. By running a set of conventional protonlocalizers it was ensured that the lungs were positioned in the center of the coil and magnet.
All measurements were performed with spontaneously breathing animals.
3.2.3. T2 measurements
A long ETL without letting the signal totally decay due to T2 decay is desirable to minimize
the scan time. Therefore T2-measurements in vivo in rat lung were made to examine the ETL
relation to the SNR. The same sequence and scan parameters as for the T1 studies was used
(section 3.2), but with 8 echoes (3.52, 7.04, 10.56, 14.05, 17.6, 21.12, 24.64, 28.16 ms). A
constant TR of 2000 ms and no IR pulse was used. The filling of k-space was also changed to
multi echo sampling, each echo filling a different k-space.
The measured signal versus TE was fitted to an exponential decay Mxy = M0exp(-TE/T2). For
the evaluation of the T2 values, T2 maps were generated with the IDL software. A T2 method
fit to the exponential decay (see above) was applied for the corresponding pixels. Mean T2
values of the pixels from ROIs placed on the T2 maps in lung were obtained.
3.2.4. T1 measurements
T1 measurements were performed on 4 rats breathing air with an ETL of 4 and 6 with a
constant TR of 6000 ms and six TI values of 100, 500, 1200, 2000, 3500 and 5000 ms. The
same inversion recovery RARE sequence and scan parameters was used as for the phantom
studies (section 3.2).
3.2.5. T1 measurements with cardiac triggering
To increase the accuracy and precision of T1, motion artefacts were reduced using cardiac
triggering, where no respiratory triggering was used. T1 measurements were performed on 4
new rats breathing air with the IR RARE sequence and scan parameters as for the phantom
studies (section 3.2). The TR was 6000 ms and six TI values of 100, 500, 1200, 2000, 3500
and 5000 ms were acquired with an ETL of 6.
3.2.6. T1 measurements with combined cardiac and respiratory triggering
In order to keep TR constant and increase the accuracy and precision of T1, combined cardiac
and respiratory triggering was used. Cardiac triggering was on together with the respiratory
triggering technique where the acquisition of the signal was blocked during the inspiration of
the animal. The combined triggering method with a constant TR of 6000 ms was tested on
four new animals and six TI values of 100, 500, 1200, 2000, 3500 and 5000 ms were
acquired with an ETL of 6. The scan parameters were the same as for the phantom studies
(section 3.2).
22
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
3.2.7. Oxygen-enhanced T1 measurements
Two T1 measurements on four rats were performed with the combined cardiac and
respiratory triggering technique. The animals were breathing air during the first acquisition
and inhaling pure oxygen on the second. To ensure high oxygen concentration in the lung, a
delay time of minimum 2 minutes was allowed between air and oxygen acquisitions.
23
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
4. Results
4.1.
Phantom studies
4.1.1. T1-calculation with Solver and IDL
T1-calculations on eight gel phantoms in the Solver and IDL software were done where an
excellent correlation (r = 1, p < 0.0001) between the two methods is observed (Table 3). The
estimated T1 values are practically the same for the two methods. ROIs are the mean signal
within the ROI for the Solver and averaged T1s for pixels within the ROI for the IDL
software. An obtained T1 map used for the T1 calculation in the IDL software is shown in
Figure 9.
Table 3.
T1 relaxation times [ms] on eight gel phantoms calculated with the
Solver tool and IDL software.
Solver
IDL
Nr
Estimated T1
Average T1 ± SD
1
974
973 ± 11
2
1076
1076 ± 15
3
1168
1169 ± 13
4
1246
1247 ± 15
5
1354
1356 ± 19
6
1434
1437 ± 20
7
1546
1550 ± 20
8
1784
1790 ± 32
Figure 9. T1 map generated with the polarityrestored data in the IDL software on 8 gel
phantoms with T1 values of 1000 – 1800 ms.
24
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
4.1.2. SNR reduction
The stochastic noise that decreased the existing SNR to a ratio of 40 ± 1 and 20 ± 1 was
added to the original phantom images. The SNR reduced images are shown in Figure 10.
Figure 10. Axial MRI images on a 5 mm single slice of 8 gel phantoms with T1 values of 1000 –
1800 ms. Image A has the original SNR of 114, for B and C is the SNR decreased to a ratio of 40 ± 1
and 20 ± 1, respectively.
4.1.3. TR evaluation
The results from the TR evaluation are presented in Figure 11 where the coefficient of
variation (SD/mean) is plotted versus mean T1. Artificial noise was added to the images and
reduced the SNRs to 40 ± 1 and 20 ± 1. For the phantom with a T1 value of 1500 ms and a
SNR of 20 the CV increased (0.10, 0.15, 0.27 and 0.30) with shorter TRs (8000, 6000, 4000
and 3000 ms, respectively). This pattern is observed over the whole range of phantoms,
where the CV is low with longer TRs (6000 and 8000 ms) compared to the shorter TRs (3000
and 4000 ms).
The difference in the CV in the phantom with a T1 of 1500 ms between a TR of 6000 ms and
8000 ms was 33 % for a SNR of 20. In order to decrease the scan time a TR of 6000 ms was
chosen for the sequence design.
25
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
A
B
0.5
0.5
SNR = 20
SNR = 40
0.4
SNR = 78
CV (SD/mean T1)
CV (SD/mean T1)
SNR = 20
SNR = 40
0.4
0.3
0.2
0.1
SNR = 105
0.3
0.2
0.1
0.0
0.0
1000
1500
2000
1000
2500
1500
2000
2500
Mean T1
Mean T1
C
D
0.5
0.5
SNR = 20
SNR = 20
SNR = 40
SNR = 40
0.4
SNR = 114
CV (SD/mean T1)
CV (SD/mean T1)
0.4
0.3
0.2
0.1
SNR = 116
0.3
0.2
0.1
0.0
0.0
1000
1500
2000
1000
2500
Mean T1
1500
2000
2500
Mean T1
Figure 11. The CV over mean T1 for T1 measurements on gel phantoms with different TRs of 3000,
4000, 6000 and 8000 ms from graph A to D, respectively. Circular ROIs were placed on the 8 vials as
in Figure 8. The original and reduced SNRs to a ratio of 40 ± 1 and 20 ± 1 for respectively study are
presented. The CV is low with longer TRs (6000 and 8000 ms) compared to the shorter TRs (3000
and 4000 ms).
4.1.4. TI evaluation
In Figure 12, eight diagrams with 8-3 TI values (A-D) for the T1 assessment are shown. The
CV is plotted versus mean T1 for the original and reduced SNRs. A constant TR of 6000 ms
was used.
For the phantom with a T1 value of 1500 ms the CV was between 0.12 - 0.15 (0.15, 0.13,
0.12, 0.14) for the T1 assessment with a SNR of 20 and TI values of 8, 7, 6 and 5,
respectively. With a reduced amount of TI values, 4 and 3, the CV increased to 0.36 and 475.
As a result, less than five TI points results in a large error for the T1 estimation where the CV
curves increases.
26
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
B
A
0.5
0.5
SNR = 20
SNR = 20
SNR = 40
0.4
SNR = 40
CV (SD/mean T1)
CV (SD/mean T1)
0.4
SNR = 114
0.3
0.2
SNR = 114
0.3
0.2
0.1
0.1
0.0
0.0
1000
1500
2000
1000
2500
1500
2000
2500
Mean T1
Mean T1
C
D
0.5
0.5
SNR = 20
SNR = 40
0.4
SNR = 114
CV (SD/mean T1)
CV (SD/mean T1)
SNR = 20
SNR = 40
0.4
0.3
0.2
0.1
SNR = 114
0.3
0.2
0.1
0.0
0.0
1000
1500
2000
2500
1000
Mean T1
1500
E
2500
F
0.5
1000
SNR = 20
SNR = 20
SNR = 40
0.4
SNR = 40
800
SNR = 114
CV (SD/mean T1)
CV (SD/mean T1)
2000
Mean T1
0.3
0.2
0.1
SNR = 114
600
400
200
0.0
0
1000
1500
2000
2500
1000
Mean T1
1500
2000
2500
Mean T1
Figure 12. The CV over mean T1 assessments on gel phantoms for different TI values of 8, 7, 6, 5, 4
and 3, starting with 8 TI points (A) down to 3 points (F) is shown. Circular ROIs were placed in the 8
vials (Figure 8). Noise was added to the images and decreased the original SNR to ratios of 40 ± 1 and
20 ± 1 for all experiments. Less than five TI points results in a large error for the T1 estimation where
the CV curves increases (E, F). Note the large y-axis scale at the last diagram (F) with 3 TIs.
4.2.
In vivo studies
4.2.1. T2 measurements
Images on eight TEs between 3.52 ms and 28.2 ms were acquired. The average T2 in rat lung
was estimated to 9 ± 2 ms (22 %). Based on an assumption that keeping the SNR over 4, the
sixth echo is still tolerable, thus an ETL of 6 was chosen.
4.2.2. T1 measurements
The measured mean T1 times were 1592 ± 308 ms (19%) and 1622 ± 294 ms (18%) for the
right and left lung, respectively, without triggering. The results are calculated from a set of
four rats and are presented in Table 4.
27
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
Table 4.
T1 relaxation times [ms] in rodent lung with an ETL of 4
without triggering, during breathing of room air. The individual
ROIs are averaged T1s for pixels within the ROI.
Subject
Right lung
Average T1 in air ± SD
Left lung
Average T1 in air ± SD
1
2
3
4
Mean
1617 ± 331
1495 ± 337
1730 ± 362
1472 ± 200
1592 ± 308 (19 %)
1811 ± 394
1412 ± 182
1715 ± 408
1550 ± 193
1622 ± 294 (18 %)
4.2.3. T1 measurements with cardiac triggering
The mean T1 relaxation times in lung parenchyma were 1682 ± 203 ms (12 %) and 1769 ±
188 ms (11 %) for right and left lung with a new set of four rats. The individual results are
presented in Table 5.
Table 5.
T1 relaxation times [ms] in rodent lung with an ETL of 6
with cardiac triggering, during breathing of room air. The individual
ROIs are averaged T1s for pixels within the ROI.
Subject
Right lung
Average T1 in air ± SD
Left lung
Average T1 in air ± SD
1
1644 ± 248
1671 ± 161
2
1645 ± 159
1832 ± 209
3
4
Mean
1595 ± 126
1843 ± 278
1682 ± 203 (12 %)
1576 ± 183
1995 ± 200
1769 ± 188 (11 %)
4.2.4. T1 measurements with combined cardiac and respiratory triggering
Artifacts were still observed in the images in spite of the combined cardiac and respiratory
triggering. Since the artifact was not related to cardiac and respiratory motion it was not
possible to evaluate if the combined triggering technique improved the accuracy and
precision of the T1 measurements.
4.2.5. Oxygen-enhanced T1 measurements
No consistent oxygen induced changes were observable (see section 4.2.4.).
28
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
5. Discussion
5.1.
Polarity restoration
The developed polarity restoration technique with the IDL software worked fine. A
comparison with the Solver tool was done to ensure the T1 estimation of the IDL software.
The results on eight vials with different T1 times correlated very well (r = 1, p < 0.0001)
which supports the T1 calculation in IDL. Moreover, sometimes the Solver has difficulties to
find the appropriate start values in order to calculate a minimum. The program may need
manual adjustment of the start values to get the solver going in right direction.
It was necessary to make three different calculations to acquire a good accuracy and precision
of T1, especially for noisy images. For images with low SNRs, the minimum value of the
magnitude curve is not always the accurate point to switch the polarity. When the polarisation
of a point is incorrect, the IR curve as well as the estimated T1 value will be inaccurate. By
calculating three IR curves one is not restricted to a single answer. The sum of squared errors
from the three plots was calculated and the curve with smallest error was selected. In that
way the precision of T1 was optimized for images with different SNRs.
An alternative method to the polarity restoration is to use the relative phases of the magnitude
images to reduce the potential systematic errors due to noise. The sign of the transverse
magnetization is in that way determined and signed magnitude images are created. This
method was tested but excluded due to complications and shortage of time.
5.2.
Scan parameters
The SNR in vivo in lung is low and the primary aim with the parameter optimization was to
increase the SNR with appropriate choices of slice thickness, NEX, bandwidth and pixel size.
However, the scan parameters ought to work practically for the oxygen-induced acquisitions
where two scans on the same animal are performed. Both air and oxygen measurements have
to be completed during the examination to achieve oxygen-induced changes. The scan time is
consequently a finite as it is not permitted anesthetizing the animals for an extended time.
Thus, the SNR should be kept high and the acquisition time short.
A relative thick slice thickness of 5 mm was chosen to increase the SNR, as the SNR is
directly proportional to the slice thickness. A NEX of 2 was reasonable, since the SNR
increases with the squared root of the NEX and motion artifacts are reduced for NEX >1. For
NEX higher than 2, the scan time would be longer and impractical for the oxygen-enhanced
measurements.
A short TE is required for MRI lung measurements since the T2 in lung is very short, in this
study a T2 of 9 ± 2 ms was measured. Preferably, the T2 decay should not influence the T1
assessments. To achieve a short TE, the bandwidth must be increased, in this case to 200
kHz. Unfortunately a higher bandwidth reduces the SNR, but this sacrifice had to be done to
attain a short TE of 3.52 ms.
The SNR increases with large voxels, as the SNR is proportional to the voxel size.
Nevertheless, to manage separating objects in in vivo images the pixel size has to be less than
the size of the examined object. When ROIs were placed in the in vivo images, contributions
from large vessels were avoided. The diameter of a vessel in rat lung was measured to 1-3
29
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
mm2. With the selected pixel sizes of 0.36 mm/pixel, 0.25 mm/pixel for the frequency and
phase resolutions, it was possible to separate the parenchyma from the vessels.
A scan time of 8.5 minutes per TI value and six acquisitions resulted in a total scan time of
about one hour. For oxygen-induced measurements this time would be doubled since one air
and one oxygen measurement is required, resulting in a total acquisition of at least two hours.
5.3.
Phantom studies
5.3.1. SNR reduction
The use of reducing the SNR by adding artificial noise to the gel phantom images was
essential to achieve relevant information. It would not be possible studying the original
phantom data with a SNR around 100 to prepare the parameters for the sequence used in vivo,
since no significant conclusions can be drawn from the images.
5.3.2. TR evaluation
A constant TR of 6000 ms was chosen since it was the most suitable parameter regarding the
scan time and CV (Figure 11). The longest TR, 8000 ms, had a 33 % lower CV than a TR of
6000 ms for the T1 assessment. However a TR of 8000 ms would also increase the scan time
with 33 % in comparison with a TR of 6000 ms. For a TR of 4000 ms the CV would increase
with 80 % in contrast to 6000 ms and was therefore excluded. The acquisition time should be
kept low for the oxygen-induced measurements in vivo with a good accuracy and precision of
T1; therefore the TR of 6000 ms was selected.
For each TR the range of TI values were changed. This could possibly affect the result of the
TR evaluation since additional TI points were used for the longer TRs (Table 2). It would be
interesting to examine the TR evaluation with a constant range of TIs.
5.3.3. TI evaluation
Reducing the number of TI values from 8 to 5, resulted in no large change of the CV (0.12 –
0.15) (Figure 12). The CV increased with more than a 100 % when 4 TI values were used for
a SNR of 20. Therefore, 5-6 TI values should be enough to sample for the T1 calculation
regarding the scan time and accuracy and precision of T1.
One interesting issue is the influence of the position of TI values chosen for the T1
calculation. In this study the starting points of 8 TI values were selected by intuition.
However, the choices of TI values may be critical for the T1 precision. Which are the
“perfect” points to sample for the inversion curve? The more points the better, but if you are
restricted to only six points where should they be positioned at the curve? In the beginning of
the IR curve, around the x-axis, or in the last part where the highest signal is achieved?
A speculation is that the noise contribution is the same over the whole span of inversion times
and the points should be chosen equidistantly over the IR curve. If the T1 is previously
known, the “perfect” points can be calculated with the T1 equation [1-2exp(-TI/T1)], where
perfect pulses are used. By starting with two endpoints, say 100 and 5000 ms for a constant
TR of 6000 ms the rest of the TIs can be calculated. By knowing the signals for each TI, one
can go backwards and calculate the TI for each signal. The equidistantly TIs in signal would
30
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
in that case be 100, 438, 868, 1458, 2401 and 5000 ms. Future studies will be done to see if
the choice of TIs will affect the T1 assessment.
5.4.
In vivo studies
5.4.1. T2 measurements
The sequence to measure T2 was not optimal, since the RARE sequence could only have
eight echoes. However, the aim was not to accurately measure T2, but only to get an
approximation of how long ETL that could be used. Because of the outflowing blood from
the slice, the T2 most likely was underestimated, meaning that T2 should be longer in reality
which resulted in a safe decision of the ETL.
5.4.2. T1 measurements with cardiac triggering
For the T1 measurements, use of cardiac triggering was necessary to increase the accuracy
and precision of T1. The precision improved with up to 8 % in lung parenchyma. The subject
number four was an outsider and differed from the other three animals with a longer T1 due
to imperfect trigger signal. But it is not totally clear why it differed and thus, will be further
investigated.
5.4.1. Oxygen-enhanced T1 measurements with combined triggering
A detailed examination was performed to characterize the artifact (ghosting). The artifact was
only observed in certain acquisitions and it was not stable by time. We attribute the ghosting
effect to shortcomings of the hardware. No conclusions can therefore be drawn about the
oxygen-enhanced measurements since the magnitude of the artifact contribution was too
large to detect any oxygen-induced changes. Further investigations will be performed.
31
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
Acknowledgements
I would like to express my gratitude to my supervisor Lars E. Olsson for giving me the great
opportunity to work with this project, and all the excellent help and support. I would also like
to thank Frank Risse for the guidance through the programming work and the valuable
discussions improving my thesis, Jelena Pesic for the assistance and handling with the
animals in the lab and Sven Månsson for the support with the MR sequence.
32
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
References
Alamidi D and Olsson L E 2009 Manual for making nickel-doped agarose gel phantoms
for MRI Imaging Centre Technical Report 3023 DECS Imaging AstraZeneca Mölndal 4-23
Beckmann N, Cannet C, Karmouty-Quintana H, Tigani B, Zurbruegg S, Blé F-X, Crémillieux Y,
Trifilieff A 2007 Lung MRI for experimental drug research Eur J Radiol 64 381-396
Chen Q, Jakob P M, Griswold M A, Levin D L, Hatabu H, Edelman R R 1998 Oxygen enhanced MR
ventilation imaging of the lung Magn Reson Materials in Physics, Biology and Medicine
7 153-161
Devereux G ABC of chronic obstructive pulmonary disease: Definition, epidemiology, and risk factors
2006 BMJ 332 1142-1144
Dietrich O 2009 Proton MRI: Oxygen-enhanced lung MRI and alternative approaches. In: MRI of
the Lung eds. Kauczor H-U, Baert A L, Knauth M, Sartor K Berlin Heidelberg: SpringerVerlag 76-87
Edelman R R, Hatubu H, Tadamura E, Li W, Prasad P V 1996 Noninvasive assessment of regional
ventilation in the human lung using oxygen-enhanced magnetic resonance imaging Nat Med
2(11) 1236-1239
Faller A, Schünke M, Schünke G 2004 The Human Body Stuttgart: Georg Thieme Verlag
333-373
Gowland P A and Stevenson V L 2003 T1: the longitudinal relaxation Time. In: Quantitative MRI
of the Brain Measuring Changes Caused by Disease ed. Tofts P West Sussex, England: John
Wiley & Sons Ltd 114-126
Grant K 2009 Vital statistics in rats September 2009
http://ratguide.com/health/basics/vital_statistics_in_rats.php (February 25, 2010)
Gupta R K, Ferretti J A, Becker E D, Weiss G H 1980 A modified fast inversion-recovery technique
for spin-lattice relaxation measurements J Magn Reson 38 447-452
Hlastala M P and Berger A J 2001 Physiology of respiration Oxford University Press New York 1-4
20-38
Irvin C G and Bates J HT 2003 Measuring the lung function in the mouse: the challenge of size
BioMed central Resp res 4:4
Kauczor H-U and Kreitner K-F 1999 MRI of the pulmonary parenchyma Eur Radiol 9 1755-1764
Kingsley P B 1999 Signal intensities and T1 calculations in multiple-echo sequences with imperfect
pulses Concepts in Magnetic Resonance 11(1) 29-49
Kingsley P B and Monahan W G 2001 Effect of increased repetition time TR on precision of
inversion-recovery T1 measurements Magn Reson Imaging 19 279-282
33
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
Levy G C and Peat I R 1975 The experimental approach to accurate carbon-13 spin-lattice
relaxation measurements J Magn Reson 18 500-521
Lindstedt S L and Schaeffer P J 2002 Use of allometry in predicting anatomical and physiological
parameters of mammals Laboratory Animals Ltd. 36 1-19
Loffler R, Muller C J, Peller M, Penzkofer H, Deimling M, Schwaiblmair M, Scheidler J, Reiser M
2000 Optimization and evaluation of the signal intensity change in multisection oxygenenhanced MR lung imaging Magn Reson Med 43 860-866
Mai V M and Chen Q 2005 Oxygen-enhanced ventilation imaging In: Functional Lung Imaging
eds. Lipson D A and van Beek J R Boca Raton, Florida: Taylor & Francis Group 182-205
Mayo M L 2009 Hierarchical model of gas exchange within the acinar airways of the human lung
(PhD Thesis, University of Missouri: Columbia, USA) 8-9
McRobbie D W, Moore E A, Graves M J, Prince M R 2003 MRI from picture to proton Cambridge:
Cambridge University Press 30-36 137-161 222-235
Mills G H, Wild J M, Eberle B, Van Beek E J R 2003 Functional magnetic resonance imaging of the
lung Br J Anaesth 91(1) 16-30
Nakagawa T, Sakuma H, Murashima S, Ishida N, Matsumura K, Takeda K 2001 Pulmonary
ventilation-perfusion MR imaging in clinical patients J Magn Reson Imaging 14 419-424
Ohno Y and Hatabu H 2007 Basics concepts and clinical applications of oxygen-enhanced MR
imaging Eur J Radiol 64 320-328
Ohno Y, Iwasawa T, Seo J B, Koyama H, Takahashi H, Oh Y-M, Nishimura Y, Sugimura K 2008
Oxygen-enhanced magnetic resonance imaging versus computed tomography Am J Respir
Crit Care Med 177 1095-1102
Pass D and Freeth G 1993 The rat ANZCCART news vol 6 no 4
Sahebjami H 1992 Aging of the normal lung. In: Treatise on pulmonary toxicology: Comparative
biology of the normal lung ed. Parent R A Boca Raton, Florida: CRC Press, Inc 351-363
Schwartzstein R M and Parker M J 2005 Respiratory Physiology: A Clinical Approach Philadelphia:
Lippincott Williams & Wilkins 1-4
Stock W K, Chen Q, Morrin M, Hatabu H, Edelman R R 1999 Oxygen-enhanced magnetic
resonance ventilation imaging of the human lung at 0.2 and 1.5 T J Magn Reson Imaging
9 838-841
Tadamura E, Hatabu H, Li W, Prasad P V, Edelman R R 1997 Effect of oxygen inhalation on
relaxation times in various tissues JMRI 7 220-225
Vaninbroukx J, Bosmans H, Sunaert S, Demedts M, Delcroix M, Marchal G, Verschakelen J 2003
The use of ECG and respiratory triggering to improve the sensitivity of oxygen-enhanced
proton MRI of lung ventilation Eur Radiol 13 1260-1265
34
Lung imaging using oxygen-enhanced MRI in small animals
Daniel Alamidi
Vlaardingerbroek M T and den Boer J A 1996 Magnetic Resonance Imaging Berlin Heidelberg:
Springer-Verlag 117-121
Watt K N, Bishop J, Nieman B J, Henkelman R M, Chen X J 2008 Oxygen-enhanced MR Imaging of
mice lungs Magn Reson Med 59 1412-1421
West J B 2008 Respiratory Physiology: the essentials 8th ed Philadelphia: Lippincott Williams &
Wilkins 1-26
Zhou X J 2004a Basic of physiologic gating, triggering, and monitoring. In: Handbook of MRI pulse
sequences eds. Bernstein M A, King K F, Zhou X J Burlington USA: Elsevier Inc. 485-486
Zhou X J 2004b Basic pulse sequences. In: Handbook of MRI pulse sequences eds. Bernstein M A,
King K F, Zhou X J Burlington USA: Elsevier Inc. 606-630
35