The presence of air in the lungs makes magnetic resonance imaging (MRI) particularly challenging; hence, the development and clinical incorporation of pulmonary MRI has trailed its use in other body regions. The use of newer perfusion imaging techniques and development of hyperpolarized gas MRI techniques have made this an effective modality for studying pulmonary structure and function noninvasively and without ionizing radiation. The purpose of this article is to review the technical aspects of structural and functional pulmonary MRI and its potential use in clinical imaging.
is a second-year Resident in Radiology at Yale University Medical
Center, New Haven, CT. He graduated from Dr. D.Y. Patil Medical
College, Bombay University, India in 1998. Since then, he has
completed a Nuclear Medicine Residency at Yale. He plans to begin
a Fellowship in Musculoskeletal Imaging following completion of
There are 2 major components of lung function, ventilation and
perfusion, which are traditionally evaluated with a combination of
chest radiography, computed tomography (CT), ventilation-perfusion
scintigraphy (VQ), and spirometry. Magnetic resonance imaging (MRI)
has been used for imaging both structure and function in many other
parts of the body. The purpose of this article is to review the
technical aspects and briefly describe the role of pulmonary MRI in
The growth of pulmonary MRI has been limited by at least 3
obstacles: 1) inflated normal lungs consist of approximately 20%
water and 80% air, leading to low proton-density and MR signals; 2)
macroscopic (cardiac and respiratory) and microscopic (blood
perfusion and thermal diffusion) motion limits the resolution of
images obtained by using longer conventional spin and gradient-echo
sequences; and 3) the differing diamagnetic properties of air and
water result in an inhomogeneous magnetic field. These latter 2
factors decrease the T2* to approximately 2 msec, and,
collectively, all three make imaging challenging. Nevertheless,
technical advances have largely overcome these obstacles.
MRI techniques for pulmonary imaging include perfusion imaging,
with and without contrast agents; structural vascular imaging with
MR pulmonary angiography (MRA); time-of-flight (TOF) imaging; and
ventilation imaging using hyperpolarized gases, oxygen or sulfur
There are 2 major techniques for MR perfusion imaging:
First-pass contrast agent techniques and arterial spin labeling
First-pass contrast agent technique
Two-dimensional (2D) T1-weighted ultrashort repetition time (TR)
(approximately 6 msec) and echo time (TE) MRI with contrast agent
forms the basis of pulmonary perfusion imaging. A short TE
(approximately 1.3 to 1.4 msec) is necessary to overcome the
inhomogeneous magnetic susceptibility of the lung, which becomes
more evident with longer TEs. Three-dimensional (3D) techniques
have been used to obtain datasets of the entire pulmonary tree
within seconds. Images are obtained after the administration of a
low dose of contrast (5 mL); the pulmonary arterial tree distal to
the subsegmental branches followed by a gradual diffuse increase in
signal intensity in the pulmonary parenchyma is seen (Figure 1).
Arterial spin labeling
In ASL techniques, saturation, spatial, or adiabatic inversion
pulses are applied to water molecules outside the imaging field,
and their signal is imaged as they enter the field.
The small signal differences between ASL and control images are
detected on a difference image. This reflects tissue perfusion and
forms the basis of echoplanar imaging signal targeting with
alternating frequency (EPISTAR) (Figure 2).
The spin-labeling pulse is applied at the level of the pulmonary
arteries, and imaging is performed over the lungs. Continuous
arterial spin labeling (CASL) techniques establish a steady state
of spin labeling, allowing for steady-state acquisitions with
respiratory triggering. Pulsed arterial spin labeling (PASL)
techniques result in intermittent spin labeling, and so
steady-state acquisitions are not entirely possible. Imaging using
PASL techniques requires the use of breath-holding and cardiac
Spin-labeling techniques are inherently sensitive to changes in
both perfusion as well as blood volume, and high-resolution
perfusion maps of the lungs can be obtained. They obviate the need
for exogenous contrast agent administration and have the advantage
of providing perfusion quantification, because temporal image
degradation seen with MRA is not seen.
ASL perfusion maps, however, are not entirely accurate, as water
molecules are labeled outside the imaging slice, and there is
signal loss during the transit time between labeling and imaging.
The signal loss cannot be accurately quantified due to different
vascular paths, which vary transit times unpredictably.
Pulmonary vascular imaging
Pulmonary MR angiography
Gadolinium chelate paramagnetic agents are used for pulmonary
MRA. Gadolinium shortens the T1 relaxation time allowing the use of
fast 3D gra-dient-echo sequences, which have very short TE and TR.
Gadolinium increases the vessel-to-background contrast-to-noise
ratios and mitigates flow artifacts. It also has the advantage of
causing fewer adverse reactions compared with iodinated contrast
agents used in CT angiography.
The MRA sequence should be a rapid T1-weighted sequence with
small voxel size, which covers the vascular system under
consideration. Three-dimensional gradient-recalled echo (GRE)
sequences with rapid radiofrequency (RF) pulses are ideal for this.
Short TR and TE sequences are used (TR <2 msec, TE <1 msec);
however, it has not been possible to entirely implement this on
existing clinical systems. Rapid imaging sequences allow for quick
serial data acquisition, which can be used in perfusion imaging
K-space phase ordering methods have also improved image quality.
Centric k-space ordering methods acquire essential data for image
interpretation at the beginning of the acquisition as opposed to
conventional methods, which place this data in the middle of the
acquisition; thus reducing motion artifacts.
Gadolinium-enhanced pulmonary MRA provides excellent depiction of
pulmonary perfusion, but results in venous enhancement that limits
the use of this technique in assessing dynamic aspects of
Two-dimensional and 3D TOF techniques have also been used to
image pulmonary vasculature.
These methods depend on blood flow to produce contrast between
vessels and surrounding tissues, and have been used to evaluate the
vasculature (Figure 3).
The techniques are limited by longer than optimal TR and TE, which
lead to increased image degradation from motion and air-tissue
interfaces and are now seldom used.
Hyperpolarized gas imaging
Inert gases, such as helium 3 (He-3) and xenon 129 (Xe-129), can
be polarized by using high-intensity laser illumination. The
resulting polarization produces an MRI signal that is far greater
than the signal produced by proton MRI. Broadband MRI systems (such
as those used for spectroscopy) and special RF coils are used for
imaging hyperpolarized gases. The nuclear magnetic moments (µ) for
He-3 and Xe-129 are lower than those of hydrogen 1 (H-1), and
resonant frequencies of 48.7 MHz and 17.6 MHz, respectively, are
used for them at 1.5T, as opposed to 63.9 MHz for H-1.
Since there is no recoverable polarization, multiple short TR
sequences are applied to obtain complete images of the lungs in a
single breath-hold while conserving the magnetization available
from the gas.
Briefly, there are 2 major methods of hyperpolarization: 1)
optical pumping and spin exchange, in which rubidium atoms are
polarized by using a circularly polarized laser light and the
polarization is transferred to either He-3 or Xe-129 by a process
called collisional spin exchange (Figure 4); and 2) metastability
exchange, in which a layer of metastable He-3 atoms is created by
laser illumination and adjacent He-3 atoms are polarized by
collisional spin exchange.
The differences between He-3 and Xe-129 are summarized in Table 1.
Due to the numerous differences between hyperpolarized gas imaging
and conventional proton MR, there are different conditions for
imaging them (Table 2).
There are 4 approaches to hyperpolarized gas imaging: static,
dynamic, diffusion, and intrapulmonary oxygen (O
Short TR and TE sequences with initial low flip angle and 10-mm
section thickness are adequate for imaging the lungs in a single
breath-hold of 10 to 20 seconds. Normally ventilated lungs show
homogeneous distribution of signal. Areas that have poor or absent
ventilation, such as areas of airway destruction or obstruction, do
not show signal. Signal is also not obtained from areas that have
increased oxygen concentration, which hastens gas depolarization.
Static imaging is limited in providing information about areas of
air trapping, which is useful in evaluating chronic obstructive
pulmonary disease (COPD) (Figure 5).
The key to dynamic ventilation pulmonary imaging is balancing
temporal and spatial resolution while conserving the fixed
magnetization available from the inhaled hyperpolarized gas. Low
flip angle gradient-echo sequences allow for inspiratory and
expiratory imaging but have limited spatial and temporal resolution
as more RF pulses are required and only a single line of k-space is
filled per excitation.
Echoplanar imaging (EPI) sequences require fewer RF pulses and have
excellent temporal resolution. However, their spatial resolution is
limited to about 5 mm due to the high diffusion coefficient and
relatively short T2* of these gases at 1.5T. Interleaved spiral
pulse sequences are preferred as they represent a compromise
between conventional gradient-echo imaging and EPI by limiting the
number of RF pulses and motion artifact (Figure 6).
Hyperpolarized gases, particularly He-3, have a diffusion
coefficient larger than water and have displacements of up to a few
millimeters in unrestricted spaces during the TE used in
gradient-echo sequences for lung imaging. In more confined distal
airways, displacements are smaller, resulting in a lower apparent
diffusion coefficient (ADC) measured by MRI. Normal lungs show
relatively uniform diffusion values with a low mean ADC, while
patients with emphysema have increased and inhomogeneous ADC
Diffusion imaging provides us with the opportunity to indirectly
study lung microstructure (Figure 7).
Intrapulmonary oxygen concentration imaging-
Hyperpolarized gases have a greater rate of depolarization in the
presence of molecular oxy-gen.
The primary mechanisms of loss of hyperpolarization are the
application of RF pulses and the presence of molecular oxygen.
Intrapulmonary oxygen concentration can be determined from the rate
of magnetization decay in successive images obtained by varying the
RF pulse flip angle or interimage delay time. At present, these
factors cannot be varied during a single breath-hold, hence there
are difficulties in reproducing lung position in successive
Oxygen is a weakly paramagnetic agent that affects the magnetic
properties of the lung primarily due to its large surface area of
distribution. Oxygen shortens T1 with an increase in its
concentration in the air spaces and blood vessels. Alternately
breathing room air (20% oxygen) and 100% oxygen results induces T1
shortening and a 20% to 30% signal change, which can be detected
with inversion recovery or multiple inversion recovery (MIR)
sequences. Average images during room air and 100% oxygen
inhalation are subtracted to obtain qualitative oxygen enhanced
ventilation maps. This method is also a means to study gas exchange
from alveoli to the pulmonary vasculature (Figure 8).
Sulfur hexafluoride imaging
Sulfur hexafluoride (SF
) has low blood solubility and high density and has been used to
study gas distribution in the lung. The most abundant isotope of
has a spin of -
⁄2 and is seen with MRI: SF
can be imaged at equilibrium with repeated RF pulses.
This method has not been used in humans as yet (Figure 9). The
techniques for pulmonary MRI are summarized in Figure 10.
Perfusion and vascular imaging
Thrombo-occlusive disease of the thoracic veins accounts for
significant morbidity in patients with in-dwelling catheters,
coagulopathies, and underlying malignancies. The veins can be
imaged without contrast using 2D and 3D TOF methods, which are
limited by long acquisition times and associated artifacts.
Better images are obtained by using intravenous gadolinium either
directly with first-pass imaging after injection into the affected
or indirectly after injection into the antecubital vein of the
nonaffected extremity and imaging during equilib-rium.
The latter method requires larger quantities of contrast but
overcomes the difficulty of injecting into a swollen extremity. MR
venography has been found to be extremely sensitive and specific
for the evaluation of central thoracic veins with the advantage of
imaging a larger field-of-view for a more comprehensive overview of
Pulmonary hypertension is characterized by increased pulmonary
artery pressure, which may be idiopathic or secondary to congenital
heart disease, coagulopathies, COPD, and so forth. It has been
noted that patients with severe pulmonary hypertension who are
imaged with axial dual spin-echo MRI with cardiac gating have
increased signal in the pulmonary arteries, and the signal in the
right pulmonary artery correlates directly with vascular
Dilated central pulmonary arteries with attenuated peripheral
vessels and loss of the normal systolic-diastolic distension and
collapse of the right pulmonary artery are also seen on MRI.
MRA is excellent for evaluating structural vascular
abnormalities. Contrast-enhanced MRA has been used to image
vasculitides, such as Takayasu's arteritis and Behçet's disease,
that are characterized by pulmonary arterial aneurysms and vascular
malformations that may affect the pulmonary arteries. If
thrombosed, these malformations may be missed at conventional
Pulmonary arteriovenous malformations in conditions, such as
hereditary hemorrhagic telangiectasia, are also seen using
gadolinium-enhanced MRA. They are poorly seen on spin-echo images
due to flow-related signal void, but are clearly seen on
breath-hold cine gradient-echo images (Figure 11).
MR has also been useful in characterizing rare pulmonary arterial
diseases, such as sarcoma and dissection.
Congenital conditions, such as total and partial anomalous
pulmonary venous connection are seen in 96% to 100% of the time on
Pulmonary embolism remains an underdiagnosed and potentially
fatal disease that accounts for approximately 50,000 deaths per
Multiple imaging methods, including ventilation perfusion
scintigraphy, CT pulmonary angiography, and conventional pulmonary
angiography, have been used to diagnose this, each with known
limitations. MRI provides the advantage of imaging pulmonary emboli
without ionizing radiation and providing ventilatory information.
Initial studies with spin-echo MRI showed thrombi as low or
intermediate signal areas, which are mimicked by slow-flow states.
Gradient-echo imaging techniques are more sensitive to flow and
have fewer slow-flow artifacts. Three-dimensional
gadolini-um-enhanced pulmonary MRA with perfusion imaging is most
sensitive, and detailed pulmonary vascular anatomy, including
subsegmental vessels and perfusion patterns, can be seen (Figure
Ventilation imaging with 100% oxygen and perfusion imaging with ASL
have been used as a noninvasive approach to ventilation-perfusion
imaging (Figure 8).
Excellent resolution of the pulmonary air spaces is now possible
with the use of single breath-hold hyperpolarized gas MRI. These
images are representative of the ventilation pattern in the lungs
of healthy subjects imaged when positioned supine; there is
preferential ventilation posteriorly with scattered ventilation
defects (>2 cm diameter) from dependent atelectasis (Figure 5).
Smoking is known to cause chronic inflammation, resulting in
significant airway damage. The changes affect the smaller
respiratory bronchioles at first and may not produce significant
changes in the mean forced expiratory volume in the first second
). Asymptomatic smokers with normal FEV
have been found to have more ventilation defects than healthy
nonsmokers, showing that hyperpolarized gas MRI is extremely
sensitive in detecting early airway disease.
Mucous plugging, chronic infection, and bronchial wall
thickening are the hallmarks of cystic fibrosis. These lead to
bronchiectasis and impaired ventilation of distal air spaces. In
studies performed on patients with cystic fibrosis, images obtained
after inhalation of hyperpolarized gas show far more defects than
expected from the morphologic abnormalities seen on proton-density
images. Thus, MRI with hyperpolarized gases may have utility in the
early detection of changes and monitoring disease progression
Asthma is characterized by hypersensitivity of the airways,
which causes airway inflammation and reversible obstruction.
Hyperpolarized gas MRI is extremely sensitive to these changes and
can detect areas of abnormal ventilation even in asymptomatic
patients. Resolution of the areas of obstruction after
bronchodilator therapy has been documented (Figure 5).
Chronic rejection of lung transplants is manifested by
bronchiolitis obliterans, a slow process that may be asymptomatic
and undetected by CT. Ventilation studies using hyperpolarized gas
MRI have shown multiple filling defects in these patients and
performed better than ventilation perfusion scintigraphy.
Early detection of this process may facilitate prompt treatment and
prolonged transplant survival.
More recently, various sequences have been used to study models
of lung inflammation. These authors have postulated a correlation
between the various components of inflammation and changes in the
appearance of the MRI signal.
Studies with dynamic imaging of hyperpolarized gases have shown
that functional ventilation can be directly assessed. There is
rapid wash-in, wash-out, and uniform distribution of the gas in
healthy subjects. Smokers with centrilobular emphysema show
nonuniform distribution of the gas initially, which becomes uniform
with rebreathing and shows air trapping in abnormal areas during
This overcomes the limitation of static imaging, which is unable to
detect air trapping. Dynamic imaging has also been used in the
evaluation of cystic fibrosis, interstitial pulmonary fibrosis, and
lung transplants (Figure 6).
Healthy volunteers show homogeneous diffusion and ADCs with
relatively low variation. Patients with emphysema have
inhomogeneous diffusion with centrilobular distribution and
heterogeneous ADC values with greater variation. Diffusion changes
correlate well with pulmonary function tests (Figure 7).
In the developing human lung, alveolar number remains stable
from approximately 8 years of age, and subsequent lung growth is
due to enlargement of the alveoli. Recent diffusion studies confirm
this finding by showing an increase in mean ADCs of older patients,
but no significant change in the air space variability.
Pulmonary diffusion imaging is a novel approach to imaging lung
microstructure in vivo and may help detect developmental lung
Intrapulmonary oxygen concentration imaging
Intrapulmonary oxygen concentration can be quantified based on
the rate of decay of the T1 signal of hyperpolarized gas. The
initial oxygen concentration in the lung also correlates well with
the rate of magnetization decay. Because changes in ventilation and
perfusion affect the oxygen concentration in the lungs, this is a
novel, noninvasive approach to imaging pulmonary oxygenation.
Novel imaging techniques have helped overcome obstacles to
pulmonary MRI, and we now have reliable, reproducible methods for
structural and functional imaging of the lungs. While MRI has been
found to be extremely sensitive in early detection of ventilation
changes, the treatment implications of these findings are not
completely understood, and this is likely to be the greatest
obstacle to incorporating pulmonary MRI into clinical imaging.
The author thanks Dr. Jeffrey Weinreb for his guidance in
writing this paper.