Lung MRI

Lung MRI
Educational Session, ISMRM 2019
May 16, 2019
Peder Larson, Ph.D.
Associate Professor
Department of Radiology and Biomedical Imaging
peder.larson@ucsf.edu

Speaker Name: Peder Larson
I have the following financial interest or relationship(s) to disclose with regard to the subject matter of this presentation:
• Consultant: Human Longevity Inc
Declaration of
Financial Interests or Relationships
May 16, 2019Lung MRI2

Outline
 Why Lung MRI?
 Challenges in Lung MRI
 Imaging Methods
• 1H MRI
• Inhaled Gasses (O2, Hyperpolarized 3He, Hyperpolarized 129Xe, 19F)
Slides: https://radiology.ucsf.edu/research/labs/larson/educational-materials
(Google: Peder Larson Group, Educational Materials link on sidebar)

Pulmonary Imaging – X-ray
 Radiography (Xray)
• Fast (~1 s)
• Widespread
• 2D projection (no
volumetric information)
• Ionizing radiation

Pulmonary Imaging – Computed Tomography (CT)
 High resolution CT is the gold standard for structural lung imaging
• Exquisite structural information
• High resolution (≤ 1 mm)
• Fast (~1-10 s)
 “Indispensible in the monitoring of CF patients” (Radiopaedia.org)
• Guide therapy and assess response
• Evaluate a range of features, including tissue thickening,
bronchiolitis, bronchiectasis, air trapping, mucus plugging
 Major drawback: Substantial Ionizing Radiation
• Repeated examinations (often every 6-18 months)
• Substantial exposure in childhood when sensitivity to radiation
is much higher
• Higher rate of radiation-induced malignancies with increases in
life span
May 16, 2019
Radiopaedia.org
Lung MRI5

Pulmonary Imaging - MRI
 Magnetic Resonance Imaging (MRI): Currently
not a major pulmonary imaging modality
because
• Slow (10s to minutes)
• Spatial resolution not as good at CT
• Prone to motion artifacts (respiration and
cardiac)
• Often use Breath-hold scanning, challenging
for patients with compromised function and
in pediatrics
• Low proton density in lung tissue
• Large susceptibility differences due to air in
lungs leads to short T2*
May 16, 2019
Wielpütz et al; Am J Respir Crit Care Med 189, 956-965; 2014.
Biederer et al; Insights Imaging. 2012 Aug; 3(4): 373–386.Lung MRI6

Why is Lung MRI so Challenging?

If Pulmonary MRI is so hard, why should we do it?
 MRI Advantages
• No ionizing radiation
• Multiple image contrasts available reflecting ventilation, perfusion, cellularity, macromolecular content
• Resolve lung motion to assess air trapping, motion defects, and tissue stiffness
 Clinical Care
• Eliminate radiation exposure in routine imaging studies (especially important in pediatrics)
• Allows more frequent imaging studies to monitor pulmonary structure & function
 Understanding of disease and Development of therapeutics
• Perform research and longitudinal studies without limitations due to ionizing radiation exposure
• Particularly valuable for pediatrics and for low acuity diseases such as asthma
 Many possible applications: bronchopulmonary dysplasia, asthma, pulmonary nodules, pulmonary embolism,
Interstitial lung diseases, cystic fibrosis, obstructive airway disease, pulmonary infections

MR Imaging Methods
Generally, these fall into two classes
1. 1H MRI methods
• Using endogenous water signal
• Address key challenges of motion and short T2*
2. Inhaled gasses
• Oxygen-enhanced MRI – 1H imaging when breathing 100% O2
• Hyperpolarized 3He (now uncommon due to 3He shortage)
• Hyperpolarized 129Xe
• 19F labeled gasses (non-hyperpolarized)

UTE and ZTE Lung Methods
 Ultrashort echo time (UTE)
MRI
• Specialized RF excitation,
readout gradients, and
image reconstruction
• Capture rapid signal decay
• Relatively robust to motion
during scan
• Inherent information for
motion estimation
May 16, 2019
Conventional UTE (2D)
10 Lung MRI

 3D radial UTE optimized for pulmonary imaging
 Variable density readout (improves SNR)
 Slab excitation (limits FOV to reduce aliasing and
motion artifacts)
 Motion correction
• Adaptive gating (e.g. bellows)
• Pseudo-random projection ordering (bit-
reversed or golden-angle for retrospective data
sorting
• Center of k-space for data-driven motion
estimation
Johnson KM, Fain SB, Schiebler M, Nagle S. MRM 2013; 70(5):1241–1250.
May 16, 2019
Kevin Johnson’s “UWUTE”
Lung MRI11

UWUTE: Pulmonary Nodule Detection
 Goal: Detect clinically-
relevant pulmonary nodules
to enable metastatic disease
screening in the lung
• Routine detection with CT
• Conventional MR
sequences fail
 5-minute Free breathing
 Adaptive bellows-based
gating
May 16, 2019
UTECT
CT UTE MRI
Burris, Larson, Johnson, Hope et al; Radiology 2015.Lung MRI12

Neonatal Pulmonary MRI
 Cincinnati Children’s
Hospital
 Dedicated Neonatal
1.5 T MRI
 3D UTE (UWUTE)
 Self navigation
May 16, 2019Lung MRI13 Higano, et al. J. Magn. Reson. Imaging. doi: 10.1002/jmri.25394,
10.1002/jmri.25643

UTE Cones: Pulmonary Nodules in Pediatrics
May 16, 2019
(a) 12 year-old with Bilateral nodules
CT UTE MRI
CT UTE MRI
CT UTE MRI
CT UTE MRI
(b) 6 year-old with partially calcified nodule
(c) 11 year-old with lung base nodule (d) 3 year-old with lung base nodule
Acquired with UTE Cones sequence, fat suppression, no
motion correction, in sedated subjects
Zucker, Cheng, Vasanawala, et al, J Magn Reson Imaging 2018.Lung MRI14

MRI Sequence comparison
 Sedated 5 year-old
 Only UTE shows signal from the lung parenchyma, and has the most clear
visualization of the pulmonary vasculature without motion artifacts.
May 16, 2019
Coronal SSFSE Axial SSFSE Axial GRE UTE
Lung MRI15
Zhu, Larson, et al. Proc. ISMRM 2019, # 4492.

Other UTE Trajectories
 Twisted Projection Imaging (Boada MRM
1997 doi:10.1002/mrm.1910380624)
 Cones (Gurney et al. MRM 2006 doi:
10.1002/mrm.20796)
 FLORET (Robison et al MRM 2018 doi:
10.1002/mrm.26500)
 Stack of Spirals
 Stack of Stars
Madelin JMRI 2013 doi: 10.1002/jmri.24168

Major Challenge: Motion
Image quality substantially compromised in up to 30% of patients due to uncorrected motion (Irregular
breathing, Coughing, Bulk motion)
Particularly problematic for children and patients with compromised pulmonary function
May 16, 2019
UTE MRI in Pediatric Cystic Fibrosis(CF) patients
Lung MRI17

Motion Compensation and Reconstruction
 With UTE and ZTE,
center of k-space or
low-resolution
images for motion
estimation
 Motion Estimation
can be used for
hard-gating
(binning), soft-
gating or XD
motion-resolved
reconstructions
May 16, 2019
Jiang, Johnson, Lustig, Larson et al, Magn Reson Med 2018.
[1] Cheng et al. JMRI 2014 .[2] Johnson et al. MRM 2014. [3] Feng et al. MRM 2015Lung MRI18

[1] Addy, N O, et al, MRM, 2016
Motion Estimation: Dynamic 3D Navigator with Local
Low-Rank
• Create Image-based navigator [1] from low-frequency
portion of k-space data
• Challenge: high-frame rate (>2Hz) and spatial resolution
(<1cm) -> ~200-fold undersampling!
• Solution: Enforce a local low-rank constraint across time
Jiang, Johnson, Lustig, Larson et al. Magn Reson Med 2018.
May 16, 2019
Dynamic 3D Navigator with Local Low-Rank Recon
Time
…
t1 t2 t3 t4
Gridding
Lung MRI19

Motion Estimation: Dynamic 3D Self-navigator
Use only center of –
space for dynamic
reconstruction with
local low-rank
constraint and
parallel imaging
(>200-fold
undersampled)
6mm isotropic
resolution
300ms temporal
resolution
May 16, 2019Jiang, Johnson, Lustig, Larson et al, Magn Reson Med 2018.
Cystic Fibrosis
(CF) patients
Lung MRI20

Scan time 4 m 18 s
TR 3.4 ms
TE 80 μs
Temporal Res 515.2 ms
Matrix Size 327 x 183 x 396
Spatial Res
1.25 x 1.25 x 1.25
mm3
“Extreme MRI”
 Super-high-resolution dynamic
reconstruction
• 1.25mm isotropic
• 515 ms temporal resolution
across 4 minute scan
 Extreme undersampling,
extreme computational and
memory cost
• Use multi-scale low-rank
factorization
• Stochastic optimization
Ong, Zhu, Larson, Lustig, et al. ISMRM 2019
#1176 (Thursday 1:57pm 710B)

“Extreme
MRI”
 1.6mm
isotropic
 560 ms
temporal
resolution
across 5
minute scan
Scan time 4 m 40 s
TR 2.8 ms
TE 80 μs
Temporal Res 560 ms
Matrix Size 320 x 185 x 189
Spatial Res 1.6 x 1.6 x 1.6 mm3
May 16, 2019
Ong, Zhu, Larson, Lustig,
et al. ISMRM 2019 #1176
(Thursday 1:57pm 710B)
Lung MRI22

Motion Compensation Strategies
1. Data organization
• Bin data
• Reject bulk motion periods
2. Image Reconstruction
• Single phase with hard or
soft-gating (SG)
• Motion-resolved
reconstruction with EXtra
Dimensional (XD) methods
May 16, 2019
Jiang, Lustig, Larson et al, Magn Reson Med 2018
Feng et al, JMRI. 2019 doi:10.1002/jmri.26245.
Lung MRI23

Motion-corrected Image Reconstruction
 Motion-resolved
(MR)
 Soft-gating(SG)
 Motion-Correction
(MoCo)
 Iterative Motion-
Correction (I-
MoCo)
May 16, 2019
5 year-old, unsedated, 5-minute scan with bulk motion
Zhu, Lustig, Larson et al.
Proc ISMRM 2019 #4492
Thursday 8:15am Computer #120
Lung MRI24

Pediatric Lung Pathologies Observed with UTE MRI
and Motion Compensation
May 16, 2019
Pneumatoceles (8yo)
Bronchiolitis Obliterans (11 yo)
CTUTE
Ground Glass in chILD (4 yo)
UTE CT
Pulmonary Nodules (5 yo)
Zhu, Larson, et al. Proc. ISMRM 2019, # 1898, # 4492.
CTUTE
CTUTE
Lung MRI25

To UTE or not to UTE: other 1H approaches
 Ultrafast SSFP (Bieri MRM 2013
DOI: 10.1002/mrm.24858)
• Short TE (< 1ms) with partial
Fourier readout
• Signal gain from T2* refocusing
from SSFP
• With very short TR (< 2ms), can
achieve no banding artifacts at 1.5T
• Derive ventilation and perfusion via
Fourier Decomposition (more in
later slides)

Quantitative Lung Physiology with GRE MRI (Hopkins,
Prisk et al UCSD)
 2D sagittal dynamic imaging
with BODY coil; 1.5cm slice,
~1.5mm in-plane
 Fast 2D gradient-echo
sequence
• Use partial k-space for short
TE (~1ms) and short TR (<5
ms)
• Fast 2D scanning with
parallel imaging
• Can freeze respiratory
motion
 Estimate proton density based on 2 TEs
and external phantom
 Derive Ventilation and Perfusion
 Combine with Pulmonary ASL methods
and oxygen-enhanced imaging
Henderson et al. J Appl Physiol 2013 doi:10.1152/japplphysiol.01531.2012

Fourier Decomposition MRI
 Fourier Decomposition (FD) methods derive
ventilation (V) and perfusion (Q) in the lung
without contrast agents
 Variations: Matrix pencil (MP), and Phase-
resolved functional lung (PREFUL) methods
 Based on fast 2D dynamic scanning (e.g. FLASH or,
for 1.5T, ultrafast-SSFP), ~0.5 s temporal resolution
 Register dynamic images and analyze temporal
frequency amplitudes
 Ventilation changes with amplitude of respiration
frequency due to tissue density changes
 Perfusion changes with amplitude of cardiac
frequency due to in-flow/time-of-flight effects
ISMRM 2019: #1, 6, 14, 875, 1880, 1895,
1896, 1897, 4079, 4081, 4093
Bauman et al MRM 2009. doi: 10.1002/mrm.22031
Bauman Bieri MRM 2017. doi: 10.1002/mrm.26096
Voskrebenzev Vogel-Claussen et al MRM 2018. doi: 10.1002/mrm.26893
Ventilation
Weighted
Perfusion
Weighted
MatrixPencil
Fourier
Decomposition

PREFUL MRI Processing
Voskrebenzev Vogel-Claussen et al MRM 2018. doi: 10.1002/mrm.26893
Perfusion changes with
amplitude of cardiac
frequency due to time-of-
flight effects
Ventilation changes with
amplitude of respiration
frequency due to tissue
density changes

Oxygen enhanced MRI
 O2 shortens T1 and T2* (See
ISMRM 2019 #1889 for details)
• T1w MRI sensitive to oxygen
uptake & ventilation
• UTE beneficial to remove
confounding T2* losses
 Compare images with room air and
breathing 100% O2
 Signal Changes reflect ventilation
Kruger et al, NMR in Biomed 2014. doi:10.1002/nbm.3158%oxygenenhancementsinHealthysubjects

Oxygen enhanced MRI Relaxation
Triphan et al. JMRI 2014. doi:10.1002/jmri.24692

Hyperpolarized Gasses
 Noble gases (e.g. 3He, 129Xe) can be
hyperpolarized via Spin Exchange Optical
Pumping
 129Xe is most common (3He is in global shortage)
• 10-40% polarization!
• Anesthetic
• Tissue solubility
 Subjects breathe in hyperpolarized gas, and
image
 Phase 3 Clinical Trial of Hyperpolarized 129Xe
 Clinical trial consortium (lead: Jason Woods,
Cincinnati Children’s Hospital)
Walker et al. Rev Mod Phys 1997, doi:10.1103/RevModPhys.69.629
Roos et al. Magn Reson Imaging Clin N Am. 2015, doi: 10.1016/j.mric.2015.01.003

Hyperpolarized 129Xe Ventilation
 Image gas distribution to measure ventilation
 Fast imaging (breath-hold, T1 ~ 20s after inhalation)

Hyperpolarized 129Xe Diffusion-weighted Imaging
 Diffusion-weighted imaging to estimate
alveolar sizes
 Image enlarged airways as in
emphysema

 Large chemical shift
~200 ppm between
gas and dissolved
phases of 129Xe
 Allows for imaging
of gas exchange
within the lung
Hyperpolarized 129Xe Dissolved Phase Imaging

19F Gasses
 Inhale inert fluorinated gasses
e.g. tetrafluoromethane (CF4), sulfur
hexafluoride (SF6), hexafluoroethane (C2F6)
and perfluoropropane (C3F8 or PFP)
 Ventilation imaging
 Flourine-19 has a similar frequency to 1H so
no additional hardware needed
 100% natural abundance, No background
signal
 Non-toxic, inexpensive, and no
hyperpolarization required
 Short T2* ~1-2ms, use UTE
May 16, 2019
Couch et al NMR in Biomedicine. 2014;27(12):1525–1534. doi:10.1002/nbm.3165
Kruger et al. JMRI 2016 doi: 10.1002/jmri.25002
Lung MRI36

Summary
 Lung MRI is challenging due to motion, short-T2*, and low signal, but there are many
advances in imaging methods and inhaled contrasts
 1H Imaging Methods
• UTE and ZTE MRI – Offer ability to capture rapid signal decay and motion. Provides
arguably highest resolution anatomical information.
• Fast 2D gradient-echo methods – Derive functional measurements (ventilation and
perfusion) without the need for custom sequences or hardware
 Inhaled gasses
• Oxygen-enhanced MRI – Assess ventilation and perfusion as a result of relaxation effects
• Hyperpolarized 129Xe – multiple functional measurements of ventilation, perfusion, and gas
exchange
• 19F – alternative to hyperpolarized gas for background-free ventilation imaging
• Recent review in Kruger et al. JMRI 2016 doi: 10.1002/jmri.25002

Software Resources
https://github.com/PulmonaryMRI/pulmonary-MRI-reconstruction
 Motion management methods for 3D center-out acquisitions
(e.g. UTE radial, cones)
ANTs https://github.com/ANTsX/ANTs
 ANTs Lung MRI Segmentation:
https://github.com/ntustison/LungAndLobeEstimationExample
( http://onlinelibrary.wiley.com/doi/10.1002/mrm.25824/full )
https://github.com/fumguo/Pulmonary-MRI-and-CT-biomarker-
framework
https://github.com/UCSDPulmonaryImaging/Deforminator
PREFUL processing – contact Jens Vogel-Claussen, Hannover
Medical School

Acknowledgements
 UCSF: Xucheng Zhu, Wenwen Jiang,
Marilynn Chan, Ngoc Ly, Thomas Hope,
Michael Hope
 UC-Berkeley: Miki Lustig, Frank Ong
 UW-Madison: Kevin Johnson, Sean Fain,
Scott Nagle
 Stanford University: Joseph Cheng, Shreyas
Vasanawala
Funding
 NIH NHLBI R01HL136965
May 16, 2019
peder.larson@ucsf.edu @pezlarson http://www.radiology.ucsf.edu/research/labs/larson
Lung MRI39

Lung MRI

More Related Content

What's hot (20)

Similar to Lung MRI (20)

More from Peder Larson (15)

Recently uploaded (20)

Lung MRI