University of Southern California Dissertations and Theses

Improved myocardial arterial spin labeled perfusion imaging

(USC Thesis Other)

Improved myocardial arterial spin labeled perfusion imaging

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content IMPROVED MYOCARDIAL ARTERIAL SPIN LABELED PERFUSION
IMAGING
by
Hung Phi Do
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PHYSICS)
May 2017
Dissertation Committee:
Krishna S. Nayak, Ph.D. (Advisor)
Aiichiro Nakano, Ph.D.
Stephan Haas, Ph.D.
Moh El-Naggar, Ph.D.
Justin Haldar, Ph.D.
Eric Wong, M.D., Ph.D. (University of California at San Diego)
Copyright 2017 Hung Phi Do
Dedication
To Phuong, Sophia, and Cataline.
ii
Acknowledgments
First, I would like to express my sincere gratitude to my advisor, Dr. Krishna
S. Nayak for his time and guidance throughout my PhD training. Your knowledge
and enthusiasm always motivate me every time meeting with you. Thank you so
much for your patience, teaching, encouragement, and being a great role model
not only academically but also personally. I believe that “I am in good hands” as
once praised by Dr. Sebastian Kelle from the German Heart Institute Berlin.
I sincerely thank the members of my qualifying and dissertation committees,
Drs. Aiichiro Nakano, Stephan Haas, Moh El-Naggar, Justin Haldar, Danny JJ
Wang, and Eric Wong for your time serving in my committee and giving me feed-
back. Thanks Dr. Aiichiro Nakano for agreeing to chair my committees and
helping with my defense rehearsal. I am grateful to Dr. Eric Wong for your guid-
ance to my research projects and willingness to drive to USC from San Diego. My
sincere appreciations go to Dr. Stephan Haas for his support since the first day I
arrived in the United States.
I would like to thank my clinical advisor and collaborator, Dr. Andrew Yoon
for the opportunity to collaborate and involve in the clinical studies at USC Keck
Hospital. I am also grateful to Drs. Graham Wright and Nilesh Ghugre from
Sunnybrook Research Institute, Toronto, Canada. Thank you for the productive
collaboration despite of the distance and for the opportunity to involve in the
pre-clinical studies in large animal model.
iii
MyappreciationstothemembersofthecardiacASLteam, TerrenceJao, Ahsan
Javed, and Vanessa Landes for the supports, discussions, feedbacks, and many
scan parties together. Many thanks to Ahsan Javed for always being cheerful and
supportive. Appreciations to Terrence Jao for your helps with proofreading and
being my scan buddy in numerous testing scans.
I would like to thank current and past members of the Magnetic Resonance
Engineering Laboratory (MREL) those I have a privilege to meet and learn from,
Yinghua Zhu, Ziyue “Brian” Wu, Eamon Doyle, Weiyi “Wayne” Chen, Xin Miao,
YiGuo,YannickBliesener,YongwanLim,NarenNallapareddy,ChristopherSandino,
Sajan G. Lingala, Johannes Toger, Yoon-Chul Kim, Houchun Harry Hu, Travis
Smith, Samir Sharma, Mahender K. Makhijani, Marc Lebel, Zungho Zun, and
Angel R. Pineda. Thanks to Dr. Marc Lebel for your help with my very first
research project. I am also thankful for the MRELers who had helped volunteer-
ing and held their breaths for my cardiac scans. I really appreciate your time and
your help.
This work was supported by the Wallace H. Counter Foundation, the American
Heart Association, and the Whittier Foundation. I am very appreciative of the
Teaching Assistantship from the Department of Physics and Astronomy, the Merit
Fellowship from the USC Dana and David Dornsife College of Letters, Arts and
Sciences, and the Dissertation Completion Fellowship from the USC Graduate
School. I also thank the Physics and EE staff members for making my life easier
with administrative procedures.
I would like to express my gratitude to my parents-in-law Ngo Ba Do and Vu
Thi Kim Dung for their constant encouragement and emotional support. Thank
you for visiting and helping us multiple times. I am grateful to my sister-in-law’s
family: Ngo Thi Thu Ha, Duong Manh Toan, and Bim for their heart-warming
iv
kindness and support. I am also thankful the support from my uncle-in-law’s
family: Ngo Quang An and Bui Thanh Tuat.
I am deeply indebted to my parents, Do Van Phien and Tran Thi Rong for
their unconditional love, endless patience, and understanding. I am also grateful
to the kind support from my brother’s family: Do Ngoc Phuong, Doan Thi Mo,
Do Quoc Viet, and Do Thanh Nam.
Last but most importantly, I pay tribute to my wife, Phuong, for her supports,
encouragements, sacrifices, and unwavering love. Without doubt, you are a won-
derful wife and an amazing mother. Thank you for always beside me during the
PhD program and I look forward to our future together. I dedicate this disserta-
tion to you and our daughters Sophia and Cataline.
Hung Phi Do
University of Southern California
Los Angeles, March 16, 2017
v
Contents
Dedication ii
Acknowledgments iii
List of Tables ix
List of Figures x
List of Acronyms xiii
Curriculum Vitae xvii
Abstract xxi
1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Coronary artery disease and current diagnostic tests . . . . . 1
1.1.2 Introduction to arterial spin labeled MRI . . . . . . . . . . . 5
1.2 Organization of the Dissertation . . . . . . . . . . . . . . . . . . . . 7
2 Background 10
2.1 Magnetic resonance physics . . . . . . . . . . . . . . . . . . . . . . 10
2.1.1 Principle of magnetic resonance imaging . . . . . . . . . . . 10
2.1.2 Spin-lattice (T
1
) relaxation . . . . . . . . . . . . . . . . . . . 11
2.1.3 Balanced steady state free precession . . . . . . . . . . . . . 13
2.1.4 Parallel imaging . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 Cardiovascular magnetic resonance . . . . . . . . . . . . . . . . . . 17
2.2.1 General overview . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.2 Regional assessment . . . . . . . . . . . . . . . . . . . . . . 26
2.3 Arterial spin labeled magnetic resonance imaging . . . . . . . . . . 28
2.3.1 Labeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.3.2 Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.3.3 Quantification . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.3.4 Magnetization transfer effects . . . . . . . . . . . . . . . . . 36
vi
2.4 Myocardial arterial spin labeling . . . . . . . . . . . . . . . . . . . . 36
2.4.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.4.2 Physiological noise . . . . . . . . . . . . . . . . . . . . . . . 38
2.4.3 Current myocardial ASL method . . . . . . . . . . . . . . . 38
3 Improved sensitivity of myocardial ASL using parallel imaging 40
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2.1 Imaging methods . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2.2 Experimental methods . . . . . . . . . . . . . . . . . . . . . 43
3.2.3 Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4 Improved sensitivity of myocardial ASL using double-gating 56
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.2.1 Experimental methods . . . . . . . . . . . . . . . . . . . . . 58
4.2.2 Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5 Feasibility of non-contrast assessment of microvascular integrity
using ASL-CMR 71
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.2.1 Animal protocol . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.2.2 CMR imaging . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.2.3 Data collection . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.2.4 Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.3.1 MBF at rest . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.3.2 MBF at rest and stress . . . . . . . . . . . . . . . . . . . . . 77
5.3.3 MBF post-AMI . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6 Feasibility of non-contrast assessment of MBF and MBV change
using T
1
mapping 86
6.1 ASL-CMR using MOLLI . . . . . . . . . . . . . . . . . . . . . . . . 86
vii
6.1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
6.1.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.1.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.1.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6.2 Vasodilator response in heart transplant recipients using T
1
-based
MBV mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6.2.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.2.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6.2.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
7 Physiologically synchronized multi-module pulsed arterial spin
labeled (SymPASL) MRI 95
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
7.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
7.2.1 Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
7.2.2 Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
7.2.3 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
7.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
7.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
8 Conclusions 101
Bibliography 105
viii
List of Tables
1.1 Existing tests for CAD and their characteristics . . . . . . . . . . . 4
3.1 MeasuredMBFandPNfromthereferenceandtheacceleratedmethod 46
3.2 Measured MBF and PN from R1, R2, and R3 . . . . . . . . . . . . 52
4.1 Per-segment MBF, PN, and tSNR in low and high HRV group . . . 64
4.2 Global and per-segment MBF, PN, and tSNR from single-gating
and double-gating . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.1 CMR protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5.2 Rest and stress MBF measured from ASL-CMR . . . . . . . . . . . 82
6.1 Measured T
1
and MBF from MOLLI-ASL . . . . . . . . . . . . . . 90
6.2 Rest and stress T
1
and T
1
change (ΔT
1
) . . . . . . . . . . . . . . . . 94
ix
List of Figures
1.1 Description of coronary artery disease . . . . . . . . . . . . . . . . . 2
1.2 Cascade of mechanism and manifestations of CAD . . . . . . . . . . 3
1.3 Principle of ASL-MRI . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1 Saturation and inversion recovery curve . . . . . . . . . . . . . . . . 12
2.2 Relationship between the MRI data and the reconstructed image . . 15
2.3 SENSE reconstructed images and corresponding g-factor maps . . . 16
2.4 Coronary artery branches that supply oxygen-rich to the heart . . . 18
2.5 EKG and PPG waveforms . . . . . . . . . . . . . . . . . . . . . . . 20
2.6 EKG and other physiological signals from the heart . . . . . . . . . 21
2.7 CINE CMR images . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.8 LGE CMR images . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.9 First-pass CMR images . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.10 Original MOLLI sequence . . . . . . . . . . . . . . . . . . . . . . . 25
2.11 AHA 17-segment model . . . . . . . . . . . . . . . . . . . . . . . . 27
2.12 Example of AHA 6-segment in a single short-axis slice . . . . . . . . 28
2.13 Generic ASL pulse sequence . . . . . . . . . . . . . . . . . . . . . . 30
2.14 Schematic of several common ASL methods . . . . . . . . . . . . . 31
2.15 ASL signal as a function of time derived from Buxton’s model . . . 35
2.16 Major milestones in the development of human myocardial ASL . . 37
x
3.1 Myocardial ASL pulse sequence with and without parallel imaging . 42
3.2 Control images from 3 subjects . . . . . . . . . . . . . . . . . . . . 45
3.3 Global MBF measured from the reference and accelerated method
(SENSE R=2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.4 Comparison of regional MBF . . . . . . . . . . . . . . . . . . . . . . 49
3.5 Global MBF measured from the R1 (fully-sampled), R2 (SENSE
rate 2), and R3 (SENSE rate 3) . . . . . . . . . . . . . . . . . . . . 50
3.6 Regional PN and TN as a function of imaging window from 5 subjects 51
4.1 Single-gated and double-gated myocardial ASL pulse sequence . . . 59
4.2 MBF and PN maps from single-gated and double-gated myocardial
ASL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.3 PN as a function of HRV in single-gated and double-gated myocar-
dial ASL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.4 MBFcomparisonbetweensingle-gatedanddouble-gatedmyocardial
ASL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.1 Representative maps or rest and stress MBF in healthy animals . . 78
5.2 Box plot of segmental MBF at rest and stress in healthy animal . . 79
5.3 RepresentativerestingMBFmapsfromASL-CMRinpost-AMIani-
mals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.4 Regional MBF in the remote and infarct segments . . . . . . . . . . 81
6.1 T
1
maps and MBF map from a subject . . . . . . . . . . . . . . . . 88
6.2 Linear regression of T
1SS
and T
1NS
. . . . . . . . . . . . . . . . . . 89
6.3 LinearregressionandBland-AltmancomparingmeasuredMBFfrom
MOLLI-ASL and FAIR-ASL . . . . . . . . . . . . . . . . . . . . . 89
xi
6.4 T
1
and T
1
change (ΔT
1
) maps in a healthy control and a heart
transplant recipient . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6.5 Regional T
1
and T
1
change (ΔT
1
) healthy and heart transplant group 93
6.6 Regional MBF and MPR in healthy and heart transplant group . . 93
7.1 SymPASL pulse sequence . . . . . . . . . . . . . . . . . . . . . . . . 96
7.2 SNR maps from FAIR and SymPASL . . . . . . . . . . . . . . . . . 99
7.3 SNR and SNR efficiency from experiment and simulation . . . . . . 99
7.4 Regression plot comparing measured MBF from FAIR and SymPASL100
xii
List of Acronyms
AIF Arterial Input Function
AHA American Heart Association
AMI Acute Myocardial Infarction
ANL Anterolateral
ANS Anteroseptal
ANT Anterior
ASL Arterial Spin Labeling
ASL-CMR Arterial Spin Labeled Cardiovascular Magnetic Resonance
ASL-MRI Arterial Spin Labeled Magnetic Resonance Imaging
ATT Arterial Transit Time
BOLD Blood Oxygenation Level Dependent
bpm beats per minute
bSSFP balanced Steady State Free Precession
CAD Coronary Artery Disease
CASL Continuous Arterial Spin Labeling
CBF Cerebral Blood Flow
CFR Coronary Flow Reserve
CKD Chronic Kidney Disease
CMR Cardiovascular Magnetic Resonance
xiii
CNR Contrast-to-Noise Ratio
CT Computed Tomography
DG Double-gating
ECG Electrocardiogram
EE Electrical Engineering
EKG Electrocardiogram
EPI Echo-Planar Imaging
EPISTAR Echo-Planar Imaging-based Signal Targeting by Alternating
Radiofrequency pulses
ESRD End Stage Renal Disease
FAIR Flow-sensitive Alternating Inversion Recovery
FFR Fractional Flow Reserve
FOV Field-of-View
GRAPPA Generalized Autocalibrating Partially Parallel Acquisition
GRE Gradient Echo
HRV Heart Rate Variation
INS Inferoseptal
INF Inferior
INL Inferolateral
LAD Left Anterior Descending artery
LCX Left Circumflex Artery
LGE Late Gadolinium Contrast Enhancement
LV Left Ventricle
MBF Myocardial Blood Flow
MBV Myocardial Blood Volume
MCE Myocardial Contrast Echocardiography
xiv
MOLLI Modified Look-Locker Inversion recovery
MPR Myocardial Perfusion Reserve
MREL Magnetic Resonance Engineering Laboratory
MR Magnetic Resonance
MRI Magnetic Resonance Imaging
MT Magnetization transfer
MTC Magnetization Transfer Contrast
MTR Magnetization Transfer Ratio
MVO Microvascular Obstruction
NSF Nephrogenic Systemic Fibrosis
PASL Pulsed Arterial Spin Labeling
PCASL Pseudo-Continuous Arterial Spin Labeling
PET Positron Emission Tomography
PG Plethysmograph Gating
PI Parallel Imaging
PICORE Proximal Inversion with Control of Off-Resonance Effects
PLD Post-Labeling-Delay
PN Physiological Noise
PTT Pulse Transit Time
PPG Photoplethysmography
R Reduction Factor in Parallel Imaging
RBF Renal Blood Flow
RCA Right Coronary Artery
RF Radiofrequency
ROI Region of Interest
xv
RR R-wave to R-wave interval
SASHA Saturation recovery Single-shot Acquisition
SAR Specific Absorption Rate
SD Standard Deviation
SENSE Sensitivity Encoding
SG Single-gating
ShMOLLI Shortened Modified Look-Locker Inversion recovery
SNR Signal-to-Noise Ratio
SPECT Single Photon Emission Computed Tomography
SSFP Steady State Free Precession
STD Standard Deviation
SymPASL physiologically Synchronized multi-module Pulsed Arterial Spin
Labeling
TD Trigger Delay
TI Inversion Time
TN Thermal noise
TR Repetition Time
TS Saturation Time
tSNR temporal Signal-to-Noise Ratio
USC University of Southern California
VS-ASL Velocity Selective Arterial Spin Labeling
xvi
Curriculum Vitae
EDUCATION
2010 - 2017: Doctor of Philosophy in Physics, the University of Southern Cali-
fornia, Los Angeles
2010 - 2014: Master of Science in Electrical Engineering, the University of South-
ern California, Los Angeles
2008 - 2009: Postgraduate Diploma in Physics, the Abdus Salam International
Center for Theoretical Physics, Trieste, Italy
2007-2008: Master of Science in Physics, Institute of Physics, Vietnam Academy
of Science and Technology, Hanoi, Vietnam
2003 - 2007: Bachelor of Science in Physics, the Hanoi National University of
Education, Hanoi, Vietnam
JOURNAL PAPERS
HP Do, V Ramanan, X Qi, J Barry, GA Wright, NR Ghugre, KS Nayak. “Fea-
sibility of Non-contrast Assessment of Microvascular Integrity using Arterial Spin
Labeled CMR.” (Submitted)
AJ Yoon, HP Do, S Cen, MW Fong, F Saremi, ML Barr, KS Nayak. “Assess-
ment of Regional Myocardial Blood Flow and Myocardial Perfusion Reserve by
Adenosine-stress Myocardial ASL Perfusion Imaging.” Journal of Magnetic Reso-
nance Imaging. (Accepted 07-Feb-2017)
xvii
HP Do, AJ Yoon, MW Fong, F Saremi, ML Barr, KS Nayak. “Double-gated
Myocardial Arterial Spin Labeled Perfusion Imaging is Robust to Heart Rate Vari-
ation.” Magnetic Resonance in Medicine. 77(5):1975-1980, 2017.
HP Do, TR Jao, KS Nayak. “Myocardial Arterial Spin Labeling Perfusion Imag-
ing with Improved Sensitivity.” Journal of Cardiovascular Magnetic Resonance
16:15, January 2014.
HP Do, HL Dao, NT Do, TV Ngo, AV Nguyen. “On the New Type of Optical
Bio-sensor from DNA-wrapped Carbon Nanotubes.” Communications in Physics
18:151-156, 2008.
PATENT
HP Do, TR Jao, KS Nayak. “Synchronized Multi-Module Pulsed Arterial Spin
LabeledMagneticResonanceImaging.” InternationalPatent No. WO/2016/089895,
published June 2016.
CONFERENCE PAPERS
HP Do, KS Nayak. “Physiologically Synchronized Multi-Module Pulsed Arterial
Spin Labeled (SymPASL) MRI.” International Society for Magnetic Resonance in
Medicine 25th Scientific Meeting, Hawaii, April 2017, p1878.
NR Ghugre, HP Do, K Chu, V Ramanan, KS Nayak, and GA Wright. “Non-
contrast assessment of vasodilator response using native myocardial T
1
and T
2
mappingandArterialSpinLabeledCMR.” International Society for Magnetic Res-
onance in Medicine 25th Scientific Meeting, Hawaii, April 2017, p0544. (Power
Pitch)
VL Landes, CM Ferguson, HP Do, JR Woollard, JD Krier, LO Lerman, KS
Nayak. “ComparisonofRenalBloodFlowMeasurementsobtainedusingASL-MRI
andCTPerfusion.” International Society for Magnetic Resonance in Medicine 25th
Scientific Meeting, Hawaii, April 2017, p4856.
HP Do, AJ Yoon, MW Fong, L Grazette, F Saremi, ML Barr, KS Nayak.
“Vasodilator Response in Heart Transplant Recipients using T
1
-based Myocardial
Blood Volume Mapping.” The SCMR/ISMRM Co-provided Workshop, Washing-
ton DC, February 2017. (Oral)
xviii
NR Ghugre, HP Do, V Ramanan, KS Nayak, GA Wright. “Contrast Free
Assessment of Vasodilator Response using MyocardialT
2
BOLD and Arterial Spin
Labeled CMR.” The Society for Cardiovascular Magnetic Resonance 20th Annual
Scientific Sessions, Washington DC, February 2017. (Walking Poster)
HPDo,VRamanan,GAWright,NRGhugre,KSNayak. “Non-contrastVasodila-
tor Response Assessment in a porcine model of Acute Myocardial Infarction using
Arterial Spin Labeled CMR.” International Society for Magnetic Resonance in
Medicine 24th Scientific Meeting, Singapore, May 2016, p0999. (Oral)
HP Do, AJ Yoon, MW Fong, F Saremi, ML Barr, KS Nayak. “Double-gated
MyocardialArterialSpinLabeledPerfusionImagingprovidesInsensitivitytoHeart
Rate Variation.” International Society for Magnetic Resonance in Medicine 24th
Scientific Meeting, Singapore, May 2016, p3145.
HP Do, KS Nayak. “Myocardial Arterial Spin Labeled Perfusion Imaging using
Modified Look-Locker Inversion Recovery (MOLLI).” International Society for
Magnetic Resonance in Medicine 24th Scientific Meeting, Singapore, May 2016,
p3142.
HP Do, V Ramanan, TR Jao, GA Wright, KS Nayak, NR Ghugre. “Non-contrast
Myocardial Perfusion Assessment in Porcine Acute Myocardial Infarction using
Arterial Spin Labeled CMR.” The Society for Cardiovascular Magnetic Resonance
Scientific Sessions, Los Angeles, January 2016, o007. (Oral)
TR Jao, HP Do, KS Nayak. “Myocardial ASL-CMR Perfusion Imaging with
Improved Sensitivity using GRAPPA.” The Society for Cardiovascular Magnetic
Resonance Scientific Sessions, Los Angeles, January 2016, p100.
HP Do, A Javed, TR Jao, HW Kim, AJ Yoon, KS Nayak. “Arterial Spin Label-
ing CMR Perfusion Imaging is Capable of Continuously Monitoring Myocardial
Blood Flow during Stress.” The Society for Cardiovascular Magnetic Resonance
and European Cardiovascular Magnetic Resonance Joint Scientific Sessions, Nice,
February 2015, p145.
TRJao,HPDo, KSNayak. “TemporalSNRofMyocardialASLdoesnotIncrease
with Improved Spatial Consistency of Background Suppression.” Biomedical Engi-
neering Society 2014 Annual Meeting, San Antonio, 2014, p646.
xix
HP Do, TR Jao, KS Nayak. “Myocardial Arterial Spin Labeling with Improved
Sensitivity to Myocardial Blood Flow using Parallel Imaging.” International Soci-
ety for Magnetic Resonance in Medicine 22nd Scientific Meeting, Milan, 2014,
p2379.
TR Jao, HP Do, KS Nayak. “In Vivo Performance of Myocardial Background
Suppression.” International Society for Magnetic Resonance in Medicine 21st Sci-
entific Meeting, Salt Lake City, 2013, p4525.
HP Do, RM Lebel, KS Nayak. “Magnetization transfer effects in wideband steady
state free precession (wb-SSFP).” International Society for Magnetic Resonance in
Medicine 19th Scientific Meeting, Montreal, 2011, p2787.
HP Do, HL Dao, NT Do, TV Ngo, AV Nguyen. “On the New Type of Optical
Bio-sensor from DNA-wrapped Carbon Nanotubes.” Vietnam National Conference
on Theoretical Physics 32nd Scientific Meeting, Nha Trang, 2007. (Oral)
xx
Abstract
Coronary artery disease (CAD) affects more than 15.5 million Americans and
causesapproximately310,000deathsperyear. Oneofthemostcommondiagnostic
tests is perfusion stress testing, primarily performed using single photon emission
computed tomography (SPECT) or first-pass cardiovascular magnetic resonance
(CMR). These methods require the use of ionizing radiation or exogenous con-
trast agents that carry associated risks to patients, especially those who require
frequent assessment or have kidney dysfunction. Myocardial arterial spin labeling
(ASL) is a promising MRI-based perfusion imaging method that can quantita-
tively measure myocardial tissue perfusion without the use of ionizing radiation
or exogenous contrast agents. Feasibility of CAD detection using ASL has been
previously demonstrated, however, several challenges remain, including low sensi-
tivity, coarse spatial resolution, and limited spatial coverage. The contributions of
this dissertation are (1) improving sensitivity, (2) exploring clinical applications,
and (3) developing a new and advantageous labeling method for myocardial ASL.
Low sensitivity is one of the major limitations of current myocardial ASL meth-
ods. I found that cardiac motion is one of the dominant sources of measurement
variability and that shortening the acquisition window from 300 ms to 150 ms
using parallel imaging significantly reduces measurement variability, i.e. improves
sensitivity. I also implemented double-gating which further improves sensitivity
xxi
of myocardial ASL method. Furthermore, I demonstrated that double-gating is
robust to heart rate variation.
I demonstrated a potential application of myocardial ASL in assessment of
microvascular obstruction (MVO) in a pig model of acute myocardial infarction
(AMI). This study also demonstrated that myocardial ASL is able to assess clin-
ically relevant increase in blood flow during vasodilation. That potentially allows
myocardial ASL to assess microvascular dysfunction as studies have shown that
microvascular function is impaired not only in the infarcted but also in the remote
territories.
Myocardial T
1
mapping with appropriate flow-sensitive preparation is able to
assess myocardial blood flow, myocardial blood flow change, and myocardial blood
volume change under vasodilation. I demonstrated the feasibility of using myocar-
dialT
1
mapping for assessment of MBF in human subjects. I also explored poten-
tial application of T
1
-based myocardial blood volume change mapping in heart
transplant recipients using myocardial T
1
mapping.
I developed a new and more efficient labeling method named physiologically
synchronized multi-module pulsed ASL (SymPASL), which may enable myocardial
ASL with volumetric coverage. SymPASL utilizes several labeling pulses that are
synchronized with blood flow and image acquisitions are synchronized with motion
of the target organ. From simulation and in vivo experiment in human kidneys,
SymPASL provides superior ASL signal and signal-to-noise ratio (SNR) efficiency
compared to the pulsed ASL method with 1 heart beat post-labeling-delay and has
similar ASL signal and SNR efficiency to pseudo-continuous ASL (PCASL) with
1.5-second labeling duration and 1.0-second post-labeling-delay.
xxii
Chapter 1
Introduction
1.1 Motivation
1.1.1 Coronary artery disease and current diagnostic tests
Coronary artery disease (CAD) is the most common type of heart disease that
affects more than 15.5 million Americans and leads to roughly 310,000 deaths
per year. In CAD, coronary arteries that supplies oxygen-rich blood to the heart
muscle is gradually narrowed due to deposited plaque over many years in a process
knownasatherosclerosisasseenintheFigure1.1. SymptomaticCADcanmanifest
as discomfort and chest pain i.e. angina, which may lead to heart attack, and
sudden death. Heart attack i.e. acute myocardial infraction (AMI) is a severe
consequence of untreated CAD when the coronary vessel is completely blocked.
There are approximately 116,000 deaths occur out of 750,000 people who have
heart attack each year. Out of those people, approximately 550,000 have their
first heart attack while approximately 200,000 have a recurrent attack. In 2015,
the total cost of CAD in the United States was $182 billion and projected total
cost in 2030 is predicted to be $322 billion [1].
1
Figure 1.1: (A) Location of the heart in the body, (B) normal coronary artery
with normal blood flow, and (C) coronary artery narrowed by plaque that leads
to abnormal blood flow. Heart attack i.e. acute myocardial infraction (AMI) is
a severe consequence of untreated CAD when the coronary vessel is completely
blocked. (Source: NHLBI, NIH [2])
The cascade of CAD and its progression are shown in Figure 1.2. Studies
have shown that endothelial dysfunction precedes development of atherosclero-
sis [3]. The endothelium is a thin and smooth layer of cells lining the inside
of arteries that is responsible for the blood vessel tone in terms of contraction
and relaxation. Endothelium also regulates hemostasis, acts as barrier to poten-
tially toxic materials, and regulates inflammation [4]. Endothelial dysfunction can
lead to microvascular dysfunction, which are the first signs of CAD. Microvascular
2
dysfunction is eventually followed by decreased sub-endocardial perfusion, altered
metabolism/abnormal ST segment, and others as shown in Figure 1.2.
Figure 1.2: Cascade of mechanisms and manifestations of ischemia having an
impact on CAD risk. (Source: Shaw et al., 2009 [5])
Early assessment of CAD not only reduces mortality and morbidity in patients
with CAD, but also helps to reduce healthcare costs. CAD is a preventable disease
and early detection of CAD allows effective treatment and monitor avoiding pro-
gression to serve stages. X-ray angiography is the gold standard for assessment of
CAD but it is invasive and typically done after other non-invasive diagnostic tests
such as blood tests, resting EKG, functional CINE CMR, functional echocardiog-
raphy, EKG stress testing, and perfusion stress testing, respectively [6].
3
Perfusion abnormality is an early indication of CAD so perfusion stress test-
ing plays an important role in early detection of CAD. Several modalities have
been used clinically, including single photon emission computed tomography
(SPECT), position emission tomography (PET), myocardial contrast echocardio-
graphy (MCE), cardiac computed tomography (CT), and first-pass cardiac MRI
(CMR) [7]. SPECT is the most widely used modality for perfusion stress test-
ing with approximately 10 million procedures per year in the United States [8].
First-pass CMR has gained increasing attention due to its improved sensitivity and
specificity compared to SPECT [9]. The current perfusion stress testing methods
require the use of ionizing radiation and/or exogenous contrast agent that pose
associated risks to patients, especially those with kidney disease. The use of ion-
izing radiation may have long-term effects and limit the repeatability of the tests
[10, 11, 12]. The use of Gadolinium in first-pass CMR are associated with nephro-
genic systemic fibrosis (NSF) in patients with end stage renal failure [13]. Further-
more, recent studies reveal that Gadolinium contrast agent can be deposited in the
brain even in subjects without severe renal dysfunction [14, 15, 16, 17]. Table 1.1
lists existing perfusion stress tests and their characteristics.
SPECT PET MCE Cardiac First-pass Myocardial
CT CMR ASL
Non-contrast 7 7 7 7 7 3
Non-radiation 7 7 3 7 3 3
ESRD-safe 3 3 7 7 7 3
Quantitative 7 3 3 3 3 ?
Repeatable 7 7 7 7 7 ?
Table 1.1: Non-invasive tests for CAD and their characteristics. ASL-MRI is a
novel test that has potential advantages compared to existing ones.
In renal disease, kidneys are damaged and cannot filter blood properly. This
leads to accumulation of toxic wastes inside the body and consequently leads to
4
other health conditions. Kidney disease patients have a high-risk of stroke and car-
diovascular disease. More than 20 million Americans have chronic kidney disease
(CKD). There are approximately 678,000 Americans that are in the final stage of
kidney disease called end-stage renal (kidney) disease (ESRD) (2016 Atlas of CKD
& ESRD) [18]. Kidney disease patients have a higher risk or CAD so frequent
screening of CAD is necessary. Existing perfusion tests may not suitable for kid-
ney disease patients because exogenous contrast agents can be toxic and the tests
associated with ionizing radiation cannot be performed frequently.
1.1.2 Introduction to arterial spin labeled MRI
Arterial spin labeled MRI (ASL) is a non-contrast and non-invasive MR-based
imaging technique that is able to quantitatively assess tissue perfusion without the
use of any exogenous contrast agents [19, 20]. Instead ASL uses blood itself as an
endogenous contrast agent for quantitative assessment of tissue blood flow. The
principle of ASL is described in Figure 1.3. Arterial blood is “labeled” using one or
more radiofrequency (RF) pulses that modify the magnetic state of arterial blood
from its equilibrium by either inversion of saturation. Once the labeled blood
is delivered to the target tissue the first image acquired is called the “labeled
image”. The labeled image reflects the MR signal from the target tissue and the
“labeled” blood that is delivered to the target tissue. The second image called
the “control image” is then acquired without the preceding radiofrequency pulses.
Subtraction of the labeled image from the control image yields an image that
is directly proportional to blood perfusion in the target tissue. By appropriate
scaling, the subtracted image is converted to a quantitative map of myocardial
blood flow in the physical unit of ml-blood/gram-tissue/minute (ml/g/min).
5
Figure 1.3: Principle of ASL-MRI. Arterial blood is labeled upstream (red box)
using RF pulses. After post-labeling-delay time of 1-3 seconds that allows labeled
blood to reach the target tissue (blue box) the first image is acquired called
“labeled” image. The second image is then acquired without the preceding labeling
pulses called “control” image. Subtraction of the labeled image from the control
image yields perfusion weighted image that is directly proportional to tissue blood
perfusion. With appropriate quantification, the perfusion weighted image can be
converted to a perfusion map in the physical unit of ml-blood/gram-tissue/min.
(Source: Ferre et al., 2013 [21])
ASL is a promising technique for quantitative assessment of tissue perfusion
because it does not involve the use exogenous contrast agents or ionizing radiation.
Therefore, itcanbeperformedfrequently, repeatedly, orevencontinuouslywithout
incremental risk to the patients. Arterial spin labeling (ASL) is used extensively in
the brain for clinical quantification of cerebral blood flow (CBF) in cerebrovascular
diseases, neuro-oncology, and neuropsychiatric diseases [22, 23, 24, 25, 26, 27].
Another novel application of ASL is vessel encoded ASL in which quantitative map
of CBF supplied by an individual artery can be achieved [28]. The MR community
hasfosteredthetranslationofASLintotheclinicbycreatingconsensusaimstogive
a standardized recommendation for performing ASL in clinical applications in the
6
brain [29]. In the consensus, pseudo-continuous ASL (PCASL) is recommended as
the method of choice for labeling and segmented 3D image acquisition is preferred.
ASL in the heart is still an active area of research because of several chal-
lenges that includes cardiac and respiratory motion, field inhomogeneity, pulsatil-
ity of blood flow, and complex geometry of coronary arteries and path of blood
flow. The current state-of-the-art in the heart is single-slice flow-sensitive alter-
nating inversion recovery (FAIR) [30, 31]. Using single-slice FAIR ASL, Zun et al.,
demonstrated that myocardial ASL is able to assess myocardial blood flow (MBF)
and detect angiographically significant CAD [32, 33]. However, several limitations
remain including low sensitivity, coarse spatial resolution, and limited spatial cov-
erage [34]. There is an opportunity to develop myocardial ASL into a safe and
repeatable test for CAD. This dissertation explores this opportunity.
1.2 Organization of the Dissertation
This dissertation is organized as following:
Chapter 1 introduces the motivation of this dissertation. This dissertation is
motivatedfromtheunmetneedindevelopmentofanon-contrastandradiation-free
test for CAD. This would have an immediate impact on management of patients
with end-stage renal disease (ESRD).
Chapter 2 provides background materials that are relevant to this dissertation.
The background includes principles of MRI physics, spin-lattice (T
1
) relaxation,
balanced steady state free precession (bSSFP) image acquisition. This chapter
also describes cardiovascular magnetic resonance (CMR) and several of its clinical
applications. I also briefly review the development of myocardial ASL from late
7
90s, the current state-of-the-art, and the limitations that motivate the work in this
dissertation.
Chapter 3 demonstrates that cardiac motions is one of the dominant sources
of physiological noise in myocardial ASL. Shortening the acquisition window from
300 ms to 100-150 ms using parallel imaging significantly reduces physiological
noise that in turn significantly increase sensitivity of myocardial ASL.
Chapter 4 demonstrates the robustness of double-gated myocardial ASL to
heartratevariation, bysystematicallycomparingwiththemorewidelyusedsingle-
gated myocardial ASL method. This work demonstrated that double-gating is
robust to heart rate variation with provides higher sensitivity to measured MBF
compared to single-gating.
Chapter 5 demonstrates the feasibility of non-contrast assessment of microvas-
cularintegrityusingmyocardialASLinaporcinemodelofacutemyocardialinfarc-
tion (AMI) i.e. “heart attack”. Specifically, the results show that myocardial is
able to detect microvascular obstruction (MVO), which is confirmed by contrast
based methods including first-pass CMR and LGE. Myocardial ASL is also able
to assess regional vasodilator response that may play an important role in assess-
ment of microvascular function since previous studies suggested that microvascular
function is impaired not only in the infarcted but also in remote territories.
Chapter 6 demonstrates the feasibility of non-contrast assessment of myocar-
dial blood flow and myocardial blood volume (MBV) change using T
1
mapping.
Feasibility of quantitative assessment of myocardial blood flow in two breath-
holds was demonstrated. This work also explored potential use of T
1
-based MBV
change mapping to study microvascular function in heart transplant recipients.
The results show that heart transplant recipients has significantly higher resting
T
1
due to fibrosis and but lower vasodilator response compared to healthy control
8
due to myocardial fibrosis and cardiomyocyte hypertrophy. The significant lower
vasodilator response may be an indication of microvascular dysfunction in heart
transplant patients.
Chapter 7 describes a new labeling method named physiologically synchronized
multi-modulepulsedarterialspinlabeled(SymPASL)MRI.SymPASLutilizesmul-
tiple labeling pulses that are synchronized with pulsatile blood flow of the coronary
arteries and image acquisition is synchronized to cardiac motion. Multiple labeling
pulses will provide higher ASL signal and less sensitivity to arterial transit time
(ATT). This may potentially allow SymPASL to be compatible with volumetric
imaging. SymPASL is validated by simulations and in vivo experiments in human
kidneys.
9
Chapter 2
Background
2.1 Magnetic resonance physics
2.1.1 Principle of magnetic resonance imaging
Magneticresonanceimaging(MRI)isapowerfulnon-invasiveimagingmodality
that has superior soft tissue contrast compared to nuclear medicines, computed
tomography, and ultrasound. An MR image is generated by manipulating nuclear
spins of the hydrogen atoms in the imaging object using external magnetic fields.
Hydrogen nucleus possesses an intrinsic magnetic moment called “spin”. In case
of no external magnetic field, there is no net magnetization formed in the imaging
object since spins are randomly oriented. Under a strong external magnetic field
B
0
, spins have a tendency to align in the direction of the applied external field.
ThatproducesasmallnetmagnetizationmomentMthatisparalleltothedirection
(longitudinal direction) of B
0
.
If M is in the different direction from B
0
it will precess about the direction of
B
0
at the Larmor frequency. The Larmor frequency depends on the strength of the
applied external magnetic field and the gyromagnetic ratio of the nucleiω
0
=γ.B
0
.
10
Gyromagnetic ratio of hydrogen nucleus γ is 42.575 MHz/T. Larmor frequency is
also called resonant frequency.
A radiofrequency (RF) pulse applied in the transverse direction and rotating at
the resonant frequency can perturb the net magnetization M from its equilibrium.
This process is called “excitation”. After an excitation, M will continue to precess
about the direction of B
0
at the Larmor frequency. The precession of the net
magnetization causes a change in the magnetic flux that in turns induces a small
voltage (an “electromotive force” (EMF)) in the receiver coil known as Faraday’s
law of induction.
Spatial selection and spatial encoding in MRI is realized by using gradient
fields (G). There are three sets (G
x
, G
y
, G
z
) of gradient field that can produce
linearly varying magnetic field in three directions. Gradient fields can be used for
spatial selective excitation as well as spatial encoding. Spatial selective excitation
allows localization of spins in the region of interest and spatial encoding allows
differentiation MR signal from different physical locations from acquired MR data
producing a 2D or 3D MR image.
2.1.2 Spin-lattice (T
1
) relaxation
After the perturbation of a RF pulse, the magnetization starts to relax to
its thermal equilibrium state via the process called spin-lattice (T
1
) relaxation. T
1
relaxation is a process mediated by interaction of excited protons and its surround-
ing magnetic moments from other protons, other nuclei, and unpaired electrons.
Each excited proton experiences a time varying local field due to tumbling motion
of molecule in which the proton resides. T
1
relaxation is most effective (short T
1
)
when the tumbling frequency is near Larmor frequency. Tumbling frequency that
is further away from the Larmor results in longerT
1
. In the liquid phase, tumbling
11
frequency is higher than the Larmor frequency since water molecule is freely move
and tumbling. Therefore,T
1
of liquid such as water is long in the order of 4-5 sec-
onds. In the solid phase, tumbling frequency is lower than the Larmor frequency
since water molecule is more restriction. That also results in longer T
1
in solid
phase. For example, T
1
of ice is also in the order of 3-5 second. T
1
is depending
on field strength, temperature, age, tissue type and its pathological states [35, 36].
0 2000 4000 6000 8000 10000
−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1
Time (ms)
Mz/M0
0 2000 4000 6000 8000 10000
−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1
Time (ms)
Mz/M0
Saturation Recovery Inversion Recovery
Figure 2.1: T
1
relaxation recovery curve of the net magnetization after a saturation
(left) and an inversion (right) pulse. BloodT
1
is used for this example that is 1932
ms as in Stanisz et al., 2005 [37].
Macroscopically,T
1
relaxation can be simply modeled as an exponential recov-
ery of the net magnetization towards equilibrium:
M
z
(t) =M
z0
·exp(−t/T
1
) +M
0
·

1−exp(−t/T
1
)

(2.1)
where, M
0
is the equilibrium magnetization, M
z0
is the initial z component
of the net magnetization, and T
1
is longitudinal relaxation time. In general, the
initial magnetizationM
z0
can be arbitrary raging from−M
0
toM
0
. There are two
special cases that are saturation and inversion recovery when the M
z0
is equal to
zero and−M
0
, respectively.
12
If the radiofrequency pulse brings the net magnetization to the transverse plane
(M
z0
= 0) that is called saturation and the relaxation equation is as following:
M
z
(t) =M
0
·

1−exp(−t/T
1
)

(2.2)
If the radiofrequency pulse brings the net magnetization to the -Z direction,
(M
z0
=−M
0
) that is called inversion and the relaxation equation is as flowing:
M
z
(t) =M
0
·

1− 2·exp(−t/T
1
)

(2.3)
2.1.3 Balanced steady state free precession
Balanced steady state free precession (bSSFP) is an imaging technique in which
a train of RF pulses are repeated with a repetition time (T
R
) with balanced [38].
Balanced gradient means net area of gradient waveforms in one T
R
is zero in all
three directions. Although bSSFP was proposed by Carr et al., in 1958 [38], it has
gained increasing attention since the recent advancement in gradient hardware.
Stronger gradient strength and higher slew-rate allow shortT
R
(3 ms - 6 ms), which
reduces sensitivity to off-resonance artifacts. That enables high quality bSSFP
image acceptable for clinical use. With an assumption that T
1
and T
2
is much
larger than T
R
the steady state bSSFP signal is derived as follow [39]:
M
SS
=M
0
·
(1−E
1
)·sin(α)
1− (E
1
−E
2
)·cos(α)−E
1
·E
2
, (2.4)
where M
0
is the equilibrium magnetization, E
1
= exp(−T
R
/T
1
), E
2
=
exp(−T
R
/T
2
), T
R
is repetition time T
R
, and α is the flip angle. The optimal
flip angle for highest M
SS
is as following:
13
α
opt
=cos
−1

E
1
−E
2
1−E
1
E
2

≈cos
−1

T
1
/T
2
− 1
T
1
/T
2
+ 1

, (2.5)
and the corresponding peak signal is given by:
M
SS
(α =α
opt
) =M
0
·
v
u
u
t
1−E
2
1
1−E
2
2
≈M
0
·
s
T
2
T
1
. (2.6)
The above derivation of Mss demonstrates that bSSFP has superior SNR effi-
ciency compared to all other gradient sequences and has intrinsic T
2
/T
1
weighting
[39, 40].
Superior SNR efficiency and T
2
/T
1
contrast make bSSFP a preferred imaging
sequence for cardiovascular magnetic resonance (CMR) imaging including func-
tional, flow, and quantitative parametric mappings. In some cardiac applications,
bSSFP is performed in transient state because the available time for image acquisi-
tion is only fraction of a second. That does not sufficient for bSSFP to reach steady
state since it would take more than 5xT
1
for steady state to establish [41]. bSSFP
signal at transient state is oscillating that can lead to artifacts due to inconsistent
weighting in k-space data. Catalyzation is typically used to mitigate the transient
artifacts by playing one or several dummy RF pulses before image acquisition. Sev-
eral methods have been proposed that efficiently reduce transient artifact allowing
snapshot bSSFP with relative short catalyzation time [41, 42, 43, 44, 45].
One major limitation of bSSFP is that it is highly sensitivity to off-resonance
[46, 47]. Off-resonance can lead to signal nulls known as “banding artifact”, at
off resonance frequency of every 1/T
R
. Off-resonance can also cause frequency
dependent flow artifact [48, 49]. Much of effort was devoted to minimize off-
resonance sensitivity of bSSFP including using minimal T
R
[50], frequency scout
14
[51], improved shimming [52], improved sequence design such as wideband SSFP
[53], combination of images with difference RF phase cycling [54, 55, 56].
2.1.4 Parallel imaging
MR is a slow imaging modality compared to CT and ultrasound. In some
applications, it can take few minutes to complete. The lengthy scan can cause
discomfort to the patient and is susceptible to motion artifacts due to movement
during scan. Lengthy scan also hinders or even prevents successful application of
MRI in certain applications for example imaging of moving organs like the heart.
iFT
Acquired MRI data in k-space Reconstructed MR image of the
heart in the short-axis view
k
x
k
y
2.k
xmax
k
x
Figure 2.2: Acquired MRI data and reconstructed image. Extend of k-space data
determines spatial resolution of the reconstructed image via a relationship δ =
1/(2·k
max
) while sampling frequency Δk determines the Field-of-view (FOV) via
a relationship Δk = 1/FOV. Given a desire spatial resolutionδ certain amount of
k-space data (up to 2·k
max
) is required and sampling frequency Δk≤ 1/FOV to
satisfy Nyquist criterion allowing reconstruction of an MRI image free of artifacts.
Coarser sampling (under-sampled) in k-space will result in aliasing artifacts after
an inverse Fourier reconstruction.
15
Acquired MR data is in Fourier space (k-space), a reciprocal space of an imag-
ing object via Fourier transform. An MR image is obtained by inverse Fourier
transform of the k-space data [57]. Acquired MR data is typically digitized and
satisfied Nyquist sampling theorem i.e. the sampling frequency must be at least
twice the highest frequency contained in the signal. The MR data that is acquired
bellow Nyquist criterion (under-sampled data) results in aliasing artifacts after an
inverse Fourier transform.

0.0
0.5
1.0
1.5

0.0
0.5
1.0
1.5
Fully-sampled
(R = 1)
SENSE rate-2
(R = 2)
SENSE rate-3
(R = 3)
SNR = 80 SNR = 50 SNR = 31
g-factor = 1.00 g-factor = 1.08 g-factor = 1.43
ROI
Figure2.3: ReconstructedMRimagefromfully-sampled,rate-2,andrate-3SENSE
parallel imaging (top row) and corresponding g-factor maps (bottom row). From
a region of interest (ROI), SNR are 80, 50, 31 and average g-factor are 1.00,
1.08, and 1.43 in fully-sampled, rate-2, and rate-3, respectively. As expected, SNR
reduction in SENSE reconstructed image is approximately equal to
√
R·g−factor
in comparison to fully-sampled image.
16
Parallel imaging (PI) is a reconstruction technique that can reconstruct under-
sampled k-space data free of aliasing artifacts [58]. PI can significantly reduce scan
time (2-5 folds in practice) by under-sampling k-space data. PI uses information
from multiple receiver coils to resolve aliasing artifacts arising from under-sampled
data. Different receiver sees an imaging object differently i.e. different sensitivity.
That allows PI to unwrap aliasing artifacts from under-sampled data using sensi-
tivity information from multiple receivers. Sensitivity information from multiple
receivers can be used directly in the image domain or indirectly in the k-space
domain for PI reconstruction that corresponds with two classes of PI technique:
image based e.g. SENSE [59] and k-space based e.g. GRAPPA [60], respectively.
MR image reconstructed from PI has lower signal-to-noise ratio (SNR) com-
pared to fully-sampled image. The reduction in SNR is proportional to square-root
of reduction factor (R) and the noise amplification factor called g-factor [59, 61].
g-factor is merely depending on the sensitivity maps of receiver coils and under-
sampled pattern. Example of reconstructed images and g-factor maps are seen in
Figure 2.3.
2.2 Cardiovascular magnetic resonance
2.2.1 General overview
The human heart is a muscular organ that is located within the thoracic cavity
between the lungs in the space called mediastinum. The heart is a powerful pump
that constantly pumps blood through blood vessels of the circulatory system. At
rest, the heart pumps approximately 60-100 beats per minute. If the heart rate is
75 beats per minute, it would be 108 thousand beats per day, and more than 39
million beats per year. The amount of blood that is pumped from the left ventricle
17
per heartbeat is called stroke volume. The stroke volume is approximately 70 ml
in healthy adults. It means that the heart pumps 5.25 litters of blood per minute,
315 liters per hour, more than 7,560 million liters per day, and 2,759,400 litters
per year, making the heart the most active muscle in the human body.
Figure 2.4: Picture of the heart with coronary arteries. There are three mains
coronary branches that supply oxygen-rich to the heart muscle. That are right
coronary artery (RCA), left anterior descending artery (LAD) and left circumflex
artery (LCX). (Source: Wikipedia [62])
While working hard to supply oxygen-rich blood to the entire body via arterial
vessel system, the heart also needs oxygen-rich blood itself to maintain its own
functionality. The arteries that supply blood to the heart itself are called coronary
arteries. Coronary arteries originate from aortic roots before branching out into
several branches to supply oxygen-rich blood for all area of the heart as shown
in Figure 2.4. In a right-dominant heart, coronary arteries are divided in three
18
main branches called left circumflex artery (LCX), left anterior descending artery
(LAD), and right coronary artery (RCA).
The heart is located within the thoracic cavity which makes respiratory motion
one of the primary factors to introduce motion artifacts during a cardiovascular
MR (CMR) data acquisition. For this reason, breath-hold is typically required
for most of the clinical CMR applications to reduce or eliminate the effects of
respiratory motion. However, breath-holding is often not feasible in patients with
severe cardiovascular disease. In such case, respiratory triggering/gating [63, 64,
65], navigator triggering/gating [66], motion compensation by phase ordering [67]
may be used. Furthermore, radial imaging can be used to mitigate the effect of
motion since it is less sensitive to motion compared to Cartesian imaging [68].
Respiratory self-gating is increasingly used because of advanced reconstruction
that allows extraction of respiratory signal from the acquired data itself [69, 70].
The extracted respiratory signal then can be used for reconstruction of respiratory
motion-resolved images.
Besides respiratory motion, the heart is moving itself. Cardiac gating (trigger-
ing)isthemostwidelytechniqueusedforCMRtosynchronizeMRdataacquisition
to the cardiac motion; more specifically to the electrical signal obtained from an
electrocardiogram (EKG) or photoplethysmography (PPG). EKG is a represen-
tation of electrical activity of the heart muscle measured by electrodes attached
on the skin. PPG, however, is measured based on optical properties of the blood
instead of direct electrical activity of the heart muscle. PPG is typically attached
in the peripheral area where there is high degree of superficial vasculature such as
fingers, toes, and ears. There is a delay time between EKG and PPG called pulse
transit time (PTT) because of the pressure wave propagation from the heart to
peripheral location as shown in Figure 2.5. PPT is subject specific and depending
19
Figure 2.5: Electrocardiogram (EKG/ECG) and photoplethysmography (PPG)
signals. Peak voltage in EKG waveform is labeled as R-wave. An CMR acquisi-
tion is typically triggered based on the detection of the R-wave. The time interval
between two consecutive R-waves is the duration of one heartbeat and it is some-
times called RR interval. Note that there is a delay time called pulse transit
time (PPT) in PPG waveform compared to EKG waveform. PTT needs to be
considered when PPG is used instead of EKG (source: Allen et al., 2002 [71]).
on the location of the probe [71]. It is important to calibrate for PTT when PPG
is used instead of EKG.
A typical EKG is shown in Figure 2.6 along with other physiological signals.
The MRI pulse sequence is typically triggered based on the R-wave i.e. the highest
voltage peak. The duration of one heart beat is the time interval between two
successive R-waves called R-R interval. The cardiac cycle is divided into two parts:
systole and diastole. The former corresponds to the contraction phase, while the
latter to the dilatation phase of the heart muscle. Systolic duration is the time
interval from the QRS complex to the T-wave and the rest is diastolic duration.
There are two quiescent periods at approximately 35-40% and 70-75% of cardiac
cycle located during systole and diastole with an average duration of 187 msec
20
(range 66-330 msec) and 118 msec (range 0-223 msec), respectively [72, 73, 74].
The quiescent periods are subject and heart rate dependent. Higher heart rate
leads to shorter duration of these quiescent periods. The data acquisition window
for MRI is typically triggered to occur at these two quiescent periods.
Isovolumic contraction
Ejection
Isovolumic relaxation
Rapid inflow
Diastasis
Atrial systole
Aortic pressure
Atrial pressure
Ventricular pressure
Ventricular volume
Electrocardiogram
Phonocardiogram
Systole Diastole Systole
1st 2nd 3rd
P
R
T
Q
S
a c v
Pressure (mmHg)
120
100
80
60
40
20
0
Volume (mL)
130
90
50
Aortic valve
opens
Aortic valve
closes
Mitral valve
closes
Mitral valve
opens
Figure 2.6: EKG and other physiological signals measured from the heart. EKG is
a representation of the electrical signal of the heart muscle. R-wave is the highest
peak in the EKG that is typically used to trigger an CMR acquisition. In one
heartbeat, the heart has contraction and relaxation phases that are called systole
and diastole, respectively. (source: Wikipedia [75])
Cardiac gating can be used in such a way that the data acquisition is synchro-
nized to occur only at a specific phase of cardiac cycle and the data acquisition
process can be repeated over one or several heartbeats until the entire k-space
21
data is acquired. That allows acquisition of MRI images at specific cardiac phases
that are free of cardiac motion. Similarly, a series of cardiac images at different
phases throughout the cardiac cycle, so-called CINE images, can be acquired by
cardiac gating and continuous data acquisition throughout the cardiac cycle. In
practice, cardiac gating is typically used in conjunction with respiratory gating.
Besides cardiac gating, cardiac self-gating has been emerging as a promising tech-
nique [70, 76, 77], where EKG signal can be estimated directly from the data itself.
Cardiac self-gating can reduce complexity of the CMR exam, improve throughput
by avoiding the problems of miss-gating, and artifacts in EKG signal due to inter-
action with radiofrequency and gradient fields [78, 79, 80].
Short-axis view 2-chamber view 3-chamber view 4-chamber view
Figure 2.7: Example CINE CMR images that are in short-axis (SA), 2-chamber
(2C),3-chamber(3C),and4-chamber(4C)view. ExcellentSNRandCNRbetween
myocardium and ventricular blood pools is obtained from bSSFP CINE sequence.
(Source: Wong et al., 2012 [81])
CMR is an increasingly important tools for assessment of cardiovascular dis-
eases. The application of CMR ranges across morphology, cardiac function, blood
flow, myocardial perfusion, myocardial viability, and coronary MR angiography
[82, 83]. Figure 2.7 shows a set short-axis and long-axis slices acquired from a
balanced SSFP CINE CMR. CINE CMR is routinely used in clinic for assessment
of global and regional cardiac function by taking a series of images of the heart at
different cardiac phases. CINE CMR allows accurate quantification of myocardial
22
masses, blood volume, ejection fraction, wall thickening, and wall motion, which
areimportantfactorsforassessmentofhearthealthanddiagnosisofheartdiseases.
Sub-endocardium scar Transmural scar
Figure 2.8: Example of LGE images. (A) non-transmural scan of approximately
25% of myocardial width. (B) Transmural scar. Hyperintensity myocardium corre-
sponds to scar tissue, which has slower Gadolinium wash-out compared to normal
myocardium. Gadolinium hasT
1
shortening effects that explain the hyperintensity
in scar tissue in T
1
-weighted LGE images. (Source: Marra et al., 2010 [84])
Late Gadolinium contrast enhanced (LGE) CMR is a gold-standard for assess-
ment of myocardial viability. LGE is performed using an inversion recovery
sequence a few minutes (3-5 min) after the administration of the Gadolinium con-
trast agent. The inversion time (delay time from the inversion pulse to the center
of k-space) is chosen such that the normal myocardium is nulled i.e. it appears
dark in the LGE images. The infarcted myocardium appears brighter due to the
delay of wash-out of Gadolinium contrast agent. Gadolinium situated in the scar
tissue enhances spin-lattice relaxation rate i.e. shortensT
1
of neighboring protons,
which leads to the hyperintensity in the scar tissue as seen in Figure 2.8. Both
LGE and SPECT are able to detect transmural myocardial scars with comparable
success rates. LGE, however, is able to detect sub-endocardial scars that are some-
times undetectable by SPECT [85, 86]. Figure 2.8 shows example of LGE images
23
in two patients. Patient in panel A shows non-transmural scar that is about 25%
of myocardial width while patient in the panel B shows a transmural scar.
First-pass CMR image Arterial input & tissue
function
Impulse response
Figure 2.9: An image taken from a first-pass perfusion CMR image series. The
green curve is the arterial input function (AIF) taken from left ventricular blood
pool. The blue and yellow curves are signal intensity evolution taken from non-
ischemic and ischemic myocardium, respectively. After deconvolution of the AIF
from tissue curves, impulse responses are generated. The initial amplitude of the
impulse response (which is proportional to myocardial blood flow) is higher in
normal myocardium compared to ischemic myocardium. (Source: Patel et al.,
2008 [87])
First-pass CMR is an emerging tool for quantitative assessment of myocardial
blood flow and diagnosis of CAD. It was demonstrated that first-pass CMR pro-
vides high accuracy in diagnosis of CAD and has superior sensitivity compared
to SPECT [9]. In first-pass CMR, series of T
1
-weighted images are dynamically
acquired (typically 1 image per heartbeat) during the first-pass arrival of the intra-
venous Gadolinium contrast administration. Normal tissue and ischemic tissue
have different tissue perfusion rates; especially during stress which results in dif-
ferent signal intensity time curves as shown in Figure 2.9. Myocardial blood flow
couldbequantifiedusingdeconvolutionofarterialinputfunctions(AIF),i.e. signal
intensity time curve of the left ventricular blood pool, from tissue signal intensity
24
time curves. Discussions regarding imaging sequences and quantitative analysis of
first-pass CMR can be found at Kellman et al., [88] and Jerosch-Herold et al., [89]
respectively.
Figure 2.10: The original MOLLI sequence: acquisition strategy and cardiac syn-
chronization (top panel), acquired image series sorted by inversion time (middle
panel), and non-linear fitting that results in aT
1
map of the heart (bottom panel).
(source: Kellman et al., 2014 [90])
T
1
relaxation time is a tissue specific parameter that dictates how fast the net
magnetization recover to the equilibrium state after a perturbation. T
1
is different
in different tissue types and their respective pathological states, which can be
exploited for diagnosis. The process of measuring map of tissue T
1
is referred
to as T
1
mapping in MRI. Myocardial T
1
mapping could be used for myocardial
tissue characterization [91]. Quantitative assessment of myocardial T
1
could be
potentially used for detection of diffuse myocardial fibrosis that is not possible with
LGE [92, 93]. In the case of focal diseases,T
1
can be a complementary to LGE as a
non-contrastassessment. MyocardialT
1
mappinghasalso beendemonstratedtoas
25
noveltechniqueforassessmentofischemia detection withouttheuseof Gadolinium
contrast [94, 95].
Inversion recovery spin echo is the gold standard for T
1
mapping. However, its
prohibitively long makes it inapplicable for most of clinical purposes. Inversion
recovery Look-Locker [96, 97] was introduced in early seventies that significantly
reduces scan time since data can be acquired continuously after an inversion pulse
. The Look-Locker method is mainly applied in the brain and other stationary
organs. Direct application of Look-Locker in to the heart is not feasible because
of respiratory and cardiac motion. This led to the development of MOLLI (Mod-
ified Look-Locker inversion recovery) [98]. MOLLI is a clever design such that
data acquisition is cardiac triggered to occur at the same cardiac phase typically
during quiescent diastole. Furthermore, the entire data acquisition is within a
single breath-hold. The modifications allow acquiring pixel-by-pixel T
1
maps of
the heart. MOLLI became the most robust and widely used T
1
mapping method
in the heart until today. The sequence diagram, raw images, and T
1
map of the
heart are shown in Figure 2.10. After MOLLI several other T
1
mapping meth-
ods have been developed such as variants of MOLLI [90], ShMOLLI [99], SASHA
[100], SAPHIRE [101], SMARTT
1
[102], and others. The comprehensive compar-
ison of several common myocardial T
1
mapping methods can be found at Roujol
et al., [103]. These methods are based on a similar principle as MOLLI: that is to
acquire images at the same cardiac phase (stable mid-diastole) at different time
points along recovery curve after the inversion and/or saturation pulses.
2.2.2 Regional assessment
AHA 17-segment model is a standardized recommendation of American Heart
Association for regional assessment cardiac imaging in general and CMR [104].
26
The AHA 17-segment model provides a standardized representation of left ventric-
ular myocardial segments and corresponding supplied coronary arteries based on
anatomy landmark and coronary topology defined by coronary angiography. The
AHA 17-segment can be obtained by acquiring three short axis slices and one long
axis slice. The name of myocardial segments and assigned coronary territories is
shown in Figure 2.11.
3 short-axis slices Vertical
long-axis slice
Coronary Artery Territories
Bullseye Representation
Figure 2.11: The AHA 17-segment model and assignment of supplied coronary ter-
ritories (A). The bullseye representation of the AHA 17-segment model (B). LAD:
left anterior descending; RCA: right coronary artery, and LCX: left circumflex
coronary artery. (Source: Cerqueira et al., 2002 [104])
Segmentation of left ventricular myocardium is an essential step for regional
assessment of CMR. The segmentation can be performed manually, semi-
automatically, or automatically. Manual delineation is the most accurate way for
myocardial segmentation given the knowledge of cardiac anatomy. It is, how-
ever, laborious, time-consuming. Semi-automatic and automatic segmentation
are desired for routine clinical practice. The successfulness of semi-automatic
and automatic segmentation methods strongly depends on contrast to noise ratio
(CNR) between myocardium and ventricular blood pools. Several commercial seg-
mentation packages are available at MEDVISO [105], CMRSegTools [106], Medis
27
[107], and Circle cardiovascular imaging [108]. CINE using bSSFP is one of the
applications where semi-automatic and automatic segmentation is increasingly
used because of excellent CNR between myocardium and ventricular blood pools.
Anatomy, MR image with segmentation contours, and corresponding regional rep-
resentation based on the AHA 17-segment model is shown in Figure 2.12.
A
B C D
Figure 2.12: (A) the heart location. (B) anatomy of the middle short axis of the
heart. (C) a corresponding MRI image with contours that delineated myocardium.
(D) Regional representation based on the AHA 17-segment model of a single-slice
middle short axis. (Sources: (A) Wikipedia [109], (B) Wikipedia [110], (D) Voigt
et al., 2014 [111])
2.3 Arterial spin labeled magnetic resonance
imaging
Arterial spin labeling (ASL) is an MR-based technique that is able to quantita-
tively assess tissue perfusion without the use of any exogenous contrast agent. ASL
is a subtractive technique, in which tissue perfusion is measured from a “labeled”
and “control” image. Upstream arterial blood is “labeled” using one or more satu-
ration or inversion radiofrequency pulses. That saturate or invert magnetization of
the arterial blood. After a post-labeling-delay (PLD) of 1-2 seconds allowing the
labeled blood to reach the target tissue, the first image is acquired called “labeled”
28
image. This labeled image reflects MR signal from labeled blood and static tis-
sue. The “control” image is acquired without the preceding labeling pulses. The
labeled image is subtracted from the control image to yield a perfusion weighted
image that is directly proportional to tissue perfusion. By appropriate scaling,
the perfusion weighted image can be converted to a perfusion map in the physical
unit of ml-blood per gram-tissue per minute (ml/g/min). The labeled bolus signal
decays with T
1
relaxation of blood which ranges from 1.5-2 second [112, 113, 37]
depending on field strength. Therefore, T
1
recovery of the labeled bolus needs to
be considered for the selection of PLD. PLD must be chosen such that it is suffi-
ciently long to account for arterial transit time (ATT) but also relatively short to
minimize T
1
decay.
ASL was first demonstrated in vivo in the rat brain early nineties by Detre
et al., and William et al., [19, 20]. Today, ASL is extensively used in the brain
for quantitative assessment of cerebral blood flow (CBF) in neurological diseases
[24, 23, 114, 115]. A consensus statement of brain ASL has been formed that
guides the standardization of brain ASL in clinical applications [29].
A generic pulse sequence of the ASL method is shown in Figure 2.13. The ASL
pulse sequences typically has two main components: labeling and imaging. After
data acquisition, the acquired control and labeled image are processed to form a
perfusion map. Multiple averages are used to overcome ASL’s low-SNR. Labeling,
imaging, and quantification details are discussed next.
2.3.1 Labeling
Labeling methods in ASL can be categorized in three groups: 1) continuous
labeling [19, 20, 116], 2) pulsed labeling [117, 118, 30, 31, 119, 120, 121], and 3)
velocity selective labeling [122]. Unlike the continuous and pulsed groups, which
29
CASL/PCASL Labeling
Image
Acquisition
Image
Acquisition
Labeling Duration PLD
Inversion Time (TI)
Labeled/Control
Pulse
CASL/
PCASL
PASL
Figure 2.13: General schematic the continuous ASL (CASL) and pulsed ASL
(PASL) pulse sequence. In the PASL, inversion time (TI) and post-labeling-delay
(PLD) are sometimes used interchangeably. Figure is adapted from Alsop et al.,
2015 [29]. CASL: continuous ASL; PCASL: pseudo-continuous ASL; PLD: post-
labeling-delay; TI: inversion time.
label blood based on spatial location, velocity selective methods label blood above
a certain velocity.
Continuous ASL (CASL) was first proposed in the early nineties by Deter et
al., and William et al., [19, 20]. Continuous labeling can be achieved by flow
driven adiabatic inversion [123]. CASL suffers from high SAR levels because a
radiofrequency pulse is applied continuously for 2-3 seconds during labeling. Con-
tinuous application of RF power also challenges and so requires modifications to
existing RF amplifiers. In response to the limitations of CASL, Dai et al., pro-
posed pseudo-continuous ASL (PCASL) [116]. PCASL uses a series of RF pulses
in the presence of a slice selective gradient to achieve flow-driven adiabatic inver-
sion. To produce the labeled image, all RF pulses have the same polarity (i.e. all
positive) and inflowing arterial blood is continuously labeled. To produce the con-
trol image, the polarity of RF pulses are alternated and inflowing arterial blood
receives no/minimal perturbation. PCASL implementation is compatible with
existing RF amplifiers and magnetization effects from control and labeled images
30
are balanced, making volumetric coverage possible. PCASL is a method of choice
as recommended by the consensus [29].
Labeled
Control
CASL/PCASL EPISTAR PICORE FAIR
Figure 2.14: Configuration of several common labeling methods. Blue indicates
imaging location, red indicates labeling/control plane/slab, and yellow arrows indi-
cate the direction of arterial blood flow. CASL and PCASL belong to continuous
labeling and EPISTAR, PICORE, and FAIR belong to pulsed labeling. CASL:
continuous ASL; PCASL: pseudo-continuous ASL; EPISTAR: echo planar imag-
ing and signal targeting with alternating RF; PICORE: proximal inversion with a
controlforoff-resonanceeffects; FAIR:flow-sensitivealternatinginversionrecovery.
Several considerations PCASL are high SAR, sensitivity to motion, off-
resonance, blood flow, and blood vessel geometry. PCASL uses a long RF pulse
train that is SAR intensive especially at high field. Flow-driven adiabatic inver-
sion occurs at a thin plane called labeling plane. That make PCASL sensitivity to
B
1
inhomogeneity, off-resonance, motion, pulsatile blood flow, and orientation of
vessel with respect to the labeling plane. These lead to reduced labeling efficiency
of PCASL [124, 125, 126].
Pulsed ASL (PASL) is a group of ASL methods where a single RF pulse is used
for labeling. In PASL, an inversion slab of 20-30 cm is used to label upstream
31
blood. There are several variations of PASL including echo planar imaging and
signaltargetingwithalternatingRF(EPISTAR);proximalinversionwithacontrol
for off-resonance effects PICORE); flow-sensitive alternating inversion recovery
(FAIR); and others. These variations differ from each other in the way they control
for magnetization transfer effects. PASL is simpler for implementation, has lower
SAR. However, PASL has lower SNR and repeatability in comparison to PCASL
[127].
VelocityselectiveASL(VS-ASL)differsfromtheothertwogroupssinceitlabels
arterial bloods based on its velocity instead of spatial location. VS-ASL utilizes
a motion-sensitive module to saturate blood above certain cut-off velocity (V
c
ut).
VS-ASL is known to be insensitive to ATT since V
c
ut can be chosen such that
VS-ASL labels blood that already presents in the imaging location. Insensitivity
to ATT is very attractive in certain patient population where ATT could be long.
PCASL is an optimal labeling method in the brain ASL. However, the opti-
mal labeling method in the heart may be different due to cardiac and respiratory
motion, pulsatile blood flow, complex geometry of coronary branches, and field
inhomogeneity. PASL is the most widely use method in the heart specifically
flow-sensitive alternating inversion recovery (FAIR) [30, 31].
2.3.2 Imaging
Segmented 3D acquisition is recommended by the consensus as a preferred
imaging method however it is not widely adopted and validated yet. In such case,
2D multi-slice single-shot EPI or spiral can be used. 2D multi-slice has benefits of
high SNR efficiency and low sensitivity to motion. bSSFP could be used but less
scan time efficient compared to EPI and spiral.
32
Intheheart, bSSFPisthemostwidelyusedimagineacquisitionforquantitative
CMR imaging and for myocardial ASL. In first-pass perfusion CMR, bSSFP was
demonstrated to have superior SNR and SNR compared to GRE and EPI [128].
Single-shot EPI could be challenging at 3.0 Tesla because of off-resonance and
geometrical distortion. Similarly, single-shot spiral is sensitive to off-resonance.
Off-resonance causes blur in spiral image that may result in spurious signal in
myocardium that interfere with ASL signal.
2.3.3 Quantification
Buxton’sgeneralkineticmodelisusedforquantitationofperfusion[129]. When
the longitudinal magnetization of tissue spins is identical in control and labeled
image acquisition the direct subtraction of the labeled image from the control
image ΔM(t) is directly proportional to longitudinal magnetization difference of
arterial blood between control and labeled condition. Immediately after an inver-
sion 2·α·M
0b
is the arterial magnetization difference, whereM
0b
is the equilibrium
magnetization of arterial blood and α is the labeling efficiency. ΔM(t) can be
model as:
ΔM(t) = 2·α·M
0b
Z
t
0
c(t
0
)·r(t−t
0
)·m(t−t
0
)dt
0
= 2·α·M
0b
·{c(t)∗[r(t)·m(t)]} (2.7)
where * denotes convolution, f is tissue perfusion in the unit of ml-blood per
ml-voxel volume per second, c(t) is the delivery function, r(t) is the clearance
function (or residue function), and m(t) is the magnetization relaxation function.
Product of r(t) and m(t) represents the fraction of the magnetization that arrives
at time t’ still remains at time t.
33
If arterial blood flow is assumed to be uniform plug flow. No labeled blood is
arrived at the imaging location whent< Δt; fromt = Δt tot = Δt+τ the labeled
blood is uniformly arrived; and for t > Δt +τ unlabeled blood is arrived. Δ t is
called arterial transit time (ATT) and τ is bolus duration for PASL and labeling
duration for CASL. PASL labels all blood once at t = 0 while CASL continuous
labels inflow blood hence c(t) is non-zero in the time interval Δt < t < Δt +τ.
c(t) =exp(−t/T
1b
) for PASL and c(t) =exp(−Δt/T
1b
) for CASL.
If single-compartment model is assumed for water exchange between blood and
tissue the residual function can be express as r(t) = exp(−f.t/λ) where λ is the
equilibrium tissue/blood partition coefficient of water. And if the water is assumed
to completely extract from the vascular space once it arrives the imaging location
the magnetization function m(t) =exp(−t/T
1
) where T
1
is relaxation time of the
tissue.
With above assumptions of c(t), r(t), and m(t) the equation 2.7 can be solved
analytically. From analytical expression of equation 2.7 ASL signal as a function
of time can be depicted as shown in Figure 2.15.
For quantification, multiple or single post-labeling-delay (PLD) can be used.
Multiple PLD strategy has benefit of simultaneous estimation of accurate tissue
perfusion and also arterial transit time (ATT). Accurate tissue perfusion is impor-
tant for quantitative imaging and longitudinal follow-up. Furthermore, ATT can
be a physiological parameter that would be useful for pathology assessment. How-
ever, multiple PLD is time consuming since multiple ASL acquisitions at different
PLD (typically 3 to 5) are required. That may not be feasible in clinical practice.
When single PLD is used, ATT is need to be considered for correct quantification
of tissue perfusion. Long PLD is typically used in both PASL and CASL to reduce
the effect of varying in ATT [130].
34
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
0
0.005
0.01
0.015
0.02
0.025
0.03
Time (ms)
ASL signal (%)

PASL
CASL
Figure 2.15: ASL signal derived from Buxton’s general kinetic model. Red line is
PASL signal and black line is CASL signal. In this plot, ATT is assumed to be 500
ms, bolus duration for PASL and labeling duration for CASL is 2000 ms. Perfect
inversion is assumed. Blood flow is 1 ml/g/min, T
1b
= 1932 ms, and T
1tissue
=
1475 ms.
For myocardial ASL, multiple PLDs acquisition was demonstrated by Wang
et al., [131]. It, however, time consuming and my not possible within the time
constrain of 3 min during peak adenosine vasodilation. Single PLD is typically
used and myocardial blood flow (MBF) is quantified as the follow equation:
F =
λ· (C−L)
2·B·T
D
·exp(−T
D
/T
1blood
)
(2.8)
where F is myocardial blood flow. λ is blood-tissue partition coefficient, C, L,
and B are control, labeled, and baseline image that is acquired in the absence of
label/control pulse. T
1blood
is T
1
relaxation time of arterial blood. T
D
is the post
labeling delay.
Alternatively, apparent T
1
-based ASL uses the bellow formula for quantifying
MBF. That was first derived by Belle et al., [132]:
35
F =
λ
T
1blood
·

T
1NS
T
1SS
− 1

(2.9)
where F is myocardial blood flow. T
1NS
and T
1SS
are T
1
maps generated from
labeled (non-selective inversion) and control (slice-selective inversion) image series.
2.3.4 Magnetization transfer effects
Magnetization transfer (MT) is an MRI technique that is able to enhance con-
trast between different tissue types based on the application of an off-resonance
radiofrequency pulse [133, 134]. MT can be used for clinical applications [135, 136].
However, magnetization transfer is an un-wanted effect in labeling methods where
RF and slice selective gradient are used such as CASL, EPISTAR, and PICORE,
and others. In such cases, the labeling pulse acts as an off-resonance RF excitation
with respect to imaging location. Off-resonance RF excitation is typically does
not alter magnetization of free protons but it does saturate of restricted protons in
proteins and membranes. Those restricted protons are typically not measurable by
MRI but they can exchange magnetization state free protons by physical, chem-
ical, and/or quantum mechanical exchange. Magnetization transfer alters signal
intensity of the labeled image that is needed to compensate in the control image
allowing accurate quantification of perfusion map.
2.4 Myocardial arterial spin labeling
2.4.1 Overview
Although ASL is extensively used in the brain, its application to the heart
remainschallengedduetocardiacandrespiratorymotion, complexbloodflowpath
36
and coronary geometry, pulsatile blood flow, and field inhomogeneity. Feasibility
ofmyocardialASLinhumanheartwasfirstdemonstratedin1999byfromPoncelet
et al., and Wacker et al., [137, 138]. For more than two decades, most of research
in myocardial ASL are feasibilities study where all used flow-sensitive alternating
inversion recovery (FAIR) except the recent development of steady-pulsed ASL
(spASL) [139]. In FAIR method, label and control are realized by using global
inversion and slice selective inversion, respectively. There are two type implemen-
tations of FAIR ASL that are subtraction-based [137, 32, 33, 131, 140, 141] and
apparentT
1
-based [142, 143, 144] approach. In the subtraction based FAIR, post-
labeling-delay(PLD)iskeptconstantincontrolandlabeledimagessuchthatstatic
tissue is exactly canceled after the subtraction. In the apparent T
1
-based FAIR,
control and labeled series can be acquired with varying PLD. Then the control and
labeled image series are fitted to generate control and labeled T
1
maps. MBF can
be estimated from the two apparent T
1
maps as in equation 2.9.
Myocardial ASL in human
1
1999 2003 2009 2010 2011
Poncelet et al., MRM 1999
• First demonstration of
ASL in human heart
Wacker et al., JMRI 2003
• First clinical study of ASL
in CAD patients
Zun et al., MRM 2009
• Stable performance
• w/ noise analysis
Zun et al., iJACC 2011
• MBF at rest and
adenosine stress
• CAD detection
Current myocardial ASL uses Flow-sensitive Alternating Inversion Recovery (FAIR)
with single-slice coverage
Detre et al., MRM 1992
• ASL was first developed in brain in 1992
Wang et al., MRM 2010
• MBF and ATT estimation
• Free-breathing using
navigator-echo
Figure 2.16: Major milestones in the development of human myocardial ASL.
Wacker et al., in 2003 [142] was demonstrated that myocardial is able to assess
CAD in patients with suspected CAD. The study used apparent T
1
-based FAIR
with saturation with single-shot FLASH image acquisition at 2.0 Tesla. Each series
of ten images can be acquired in a single breath-hold of 14-18 seconds depending
on individual heart rate. A more recent clinical application of myocardial ASL
37
in patient with suspected CAD was performed by Zun et al., in 2011 [33]. In
this study FAIR with bSSFP was used at 3.0 Tesla. The study demonstrated
that myocardial ASL is able to detect MBF increase with adenosine vasodilator
and angiographically significant CAD in patients with suspected CAD (Zun et al.,
2011 [33]).
2.4.2 Physiological noise
ASL is a poor-SNR imaging method. Multiple averages are needed to produce
meaningful map of MBF. Physiological noise (PN) is used as a quality metric for
myocardial ASL method. PN is a measure of short term variability in the unit of
ml/g/minandisdefinedasstandarddeviationofrepeatedmeasurements(typically
six measurements) [32]. In apparentT
1
-based myocardial ASL, residual of fit could
be used as a surrogate for PN. Lower PN means higher sensitivity to measured
MBF. PN is one of the major limitations of myocardial ASL as acknowledged by
Zun et al., [32, 33] and Epstein et al., [34].
2.4.3 Current myocardial ASL method
Currently, FAIR labeling with bSSFP image acquisition is most widely used for
myocardial ASL. Each pair of control and label image is acquired in approximately
10-12 second breath-hold. That is repeated six times for signal averaging resulting
inapproximately3minscantimeforsingle-slicecoverage. Atrest, thatprocesscan
be repeated when an increased spatial coverage is needed. However, there is time
constrain during stress of adenosine vasodilation. The peak adenosine vasodilation
only lasts about 3-4 min. Current ASL method only able to acquire a single mid-
short-axissliceoftheheartthatisnotsufficientforroutineclinicaluse. Myocardial
ASLhasapotentialtoevolveintoaclinicallyusefultestforCAD.However, current
38
limitations of myocardial ASL includes low sensitivity (high physiological noise),
coarse spatial resolution, and limited spatial coverage [34]. This dissertation aims
to (1) improve sensitivity, (2) explore clinical applications, and (3) develop a new
and advantageous labeling method for myocardial ASL.
39
Chapter 3
Improved sensitivity of myocardial ASL using
parallel imaging
3.1 Introduction
Arterial spin labeling (ASL) is a quantitative, contrast-free MRI technique for
measuring tissue perfusion. It is most commonly used in the brain for the clinical
quantification of cerebral blood flow in cerebrovascular disease and neuro-oncology
[22, 26, 25]. ASL outside the brain is an immature technology but shows great
promiseintheheart. SeveralgroupshavebeenabletouseASLtomeasuremyocar-
dial blood flow (MBF) in both animal models [132, 145, 146, 147, 148, 149] and
humans [137, 143, 32, 131]. Myocardial ASL has even been demonstrated to be
compatible with pharmacological stress testing and is able to detect clinically rele-
vant increases in MBF with vasodilation [142, 33], making it a potential diagnostic
tool for detecting ischemic heart disease.
However, myocardial ASL perfusion imaging faces several challenges. The ASL
signal has low sensitivity to blood flow and only produces a 1-8% signal change
under normal physiological myocardial blood flows of 0.5-4 ml/g/min [34]. ASL
also requires the subtraction of multiple image pairs and is particularly sensitive
40
to imperfect subtraction caused by either respiratory or cardiac motion. In this
study, we aim to reduce physiological noise (PN), which refers to temporal fluc-
tuations in the ASL signal. We hypothesize that motion within the acquisition
window is a major contributor to PN, which can be reduced by accelerating the
image acquisition using parallel imaging (PI). We chose parallel imaging for accel-
eration because of its widespread clinical adoption over the past decade and its
well-understood noise and artifact behavior.
In this study, we compare the performance of an accelerated myocardial ASL
method using rate-2 sensitivity encoding (SENSE) [59] with a reference, unaccel-
erated myocardial ASL method [32] in healthy volunteers at rest.
3.2 Methods
3.2.1 Imaging methods
All experiments were performed on a Signa Excite HDxt 3 Tesla scanner (GE
Healthcare, Waukesha, WI, USA) with an eight-channel cardiac receiver coil.
Myocardial ASL was performed on a single short axis middle ventricular slice using
flow-sensitive alternating inversion recovery (FAIR) [30, 31]. The pulse sequence
shown in Figure 3.1 is composed of either a slab selective inversion pulse for control
images or a global inversion pulse for tagged images, a fat saturation module, a 19-
TR Kaiser-Bessel ramp up of flip angles to minimize transients [150], a snapshot
balanced steady state free precession (b-SSFP) image acquisition, and a 19-TR
Kaiser-Bessel ramp down of flip angles.
The reference ASL scan was implemented as described previously by Zun et
al. [32, 33]. All image acquisitions had 3.2 ms TR, 1.5 ms TE, 50
0
prescribed flip
angle, 62.5 kHz receiver bandwidth, and 24-26 cm isotropic field-of-view (FOV).
41
TI = RR (Post Labeling Delay)
Fat Sat Ramp
TD (Trigger Delay)
a
b
FAIR Labeling
SSFP
R=1
SSFP
R=2
Figure3.1: SchematicofthemyocardialASLpulsesequence. (a)Referencemethod
[32] and (b) accelerated method, using rate-2 SENSE acceleration. The pulse
sequence consists of a FAIR labeling pulse (red), a fat saturation pulse (green),
a Kaiser-Bessel ramp up (yellow), balanced SSFP image acquisition (blue), and a
Kaiser-Bessel ramp down (yellow). The FAIR labeling inversion pulse is triggered
to occur at mid-diastole. The center of the acquisition window is set to occur
precisely one heartbeat after the tagging pulse. These diagrams are to scale for a
heart rate of 60 beats per minute.
Slice thickness was 10 mm, and the selective inversion slab for control images was
30 mm thick to ensure that the imaging slice was within the inversion slab. The
reference ASL scans had a 96x96 matrix size resulting in an acquisition window
of 307 ms. This was preceded by a 10 ms fat saturation module and a 60 ms
Kaiser-Bessel ramp up of flip angles and followed by a 60 ms Kaiser-Bessel ramp
down of flip angles. The total imaging window of the reference method was 440
ms.
The accelerated ASL scan used sensitivity encoding (SENSE) [59] with a reduc-
tionfactorof2toshortentheacquisitionwindowto153ms, asshowninFigure3.1.
This was also preceded by a 10 ms fat saturation module and a 60 ms Kaiser-Bessel
ramp up, and followed by a 60 ms Kaiser-Bessel ramp down. The total imaging
42
window of the accelerated method was 286 ms. We hypothesized that motion
during the long acquisition window was a major source of physiological noise and
could be reduced by shortening the acquisition window from 307 ms to 153 ms.
Breath-holds and cardiac triggering were used to minimize respiratory and
cardiac motion, respectively. The FAIR labeling inversion pulse was timed to occur
at mid-diastole through plethysmograph gating (PG). Mid-diastole was estimated
to be at 77% of the R-R interval duration [151] and the PG trigger delay was
set to this value minus 200 ms to account for circulation time from the R-wave
to the fingertip [71] . The center of the acquisition window was set to occur one
heartbeat (T
D
= 1R-R) after the FAIR labeling pulse. Control and tagged images
were acquired in the same breath-hold.
The ASL protocol was comprised of 7 breath-holds and took roughly between
2Ð3 minutes. In the first breath-hold, baseline and noise images were acquired
in 2 seconds. The baseline and noise images were later used for estimating the
blood equilibrium magnetization (M
0
) and thermal noise (TN), respectively. Six
pairs of control and tagged images were acquired for spatial temporal averaging
in the following six 12-second breath-holds. There was a 6-second waiting period
between control and tagged images within each breath-hold. The protocol was the
same for the reference and accelerated scans except the acquisition window was
shortening by using rate-2 SENSE in the accelerated scan.
3.2.2 Experimental methods
Thisstudyincluded7healthysubjects(6males, 1female, age22-29years, mean
age = 25 years). The University of Southern California Institutional Review Board
approved the study protocol and informed consent was obtained from all subjects.
The scan protocol started with a localization scan and identification of a middle
43
short axis slice [104]. A baseline image, an image without the FAIR labeling pulse,
was acquired to ensure the slice was prescribed properly without banding artifacts
over the myocardium. If banding artifacts were present on the myocardium, a fre-
quency scout scan was performed. The offset frequency that resulted in no banding
artifact in the region of interest (left ventricular myocardium) was recorded and
used for subsequent scans. A fully sampled baseline image with no banding arti-
fact was then acquired in a 1-second breath-hold to estimate the coil sensitivity
map. The reference ASL scan was performed before the accelerated ASL scan in
all subjects.
3.2.3 Data analysis
MBF, PN, and TN were calculated in the same way as in prior work by Zun et
al. [32, 33]. All data processing was performed in MATLAB (Mathworks, Natick,
MA). After image reconstruction, the myocardium was manually segmented and
resampled into polar coordinates using a spatio-temporal averaging filter [152].
MBF quantification was derived from Buxton general kinetic model [129]
F =
C−L
2M
0
·T
D
·exp(−T
D
/T
1
)
, (3.1)
where F is myocardial blood flow. C, L, and M
0
refer to the mean myocardial
signal in the control, labeled, and baseline images,T
D
represents the post labeling
delay time and was equal to the R-R interval, andT
1
is the longitudinal relaxation
time of blood, which was assumed to be 1650 ms [112].
The size of the spatial filter and the number of resampled segments could be
freely chosen. Global MBF quantification was performed with a filter size of 2π
and a single segment. Regional MBF quantification was performed with a filter
44
size of π/3 and 6 segments [104]. Septal MBF quantification was performed using
a filter size of 2π/3 with 3 segments. This measurement was performed in order
to compare these data with a prior analysis of physiological noise [32]. A paired
Students t-test was used to compare MBF and PN between the reference and
accelerated ASL methods. Agreement between regional MBF measured from the
two methods was also assessed by Bland-Altman plot.
3.3 Results
a
b
Subject 3
Subject 6
Subject 5
Figure 3.2: Control images from three representative subjects. (a) Reference
method and (b) accelerated method, using rate-2 SENSE acceleration. All images
are windowed identically. Motion artifacts, identified by red arrows, are visible
in the images acquired using the reference method whereas SENSE accelerated
images showed significantly less motion artifacts.
Representative control images acquired from the reference method are shown in
Figure 3.2a; those acquired from the accelerated method are shown in Figure 3.2b.
There was no visible residual aliasing in the SENSE accelerated images. Motion
artifacts were seen in several images acquired from the reference method, as indi-
cated by red arrows whereas SENSE accelerated images showed significantly less
motion artifacts. It is worth noting that the per-pixel SNR of baseline images was
126 and 90 for the reference and accelerated acquisitions, respectively. The SNR of
45
46
Global Septal Regional
MBF PN MBF PN MBF PN
(ml/g/min) (ml/g/min) (ml/g/min) (ml/g/min) (ml/g/min) (ml/g/min)
Reference 1.20± 0.44 0.20± 0.08 1.23± 0.56
∗
0.28± 0.11
∗
1.22± 0.68 0.34± 0.21
Accelerated 1.24± 0.25 0.08± 0.05 1.28± 0.44 0.17± 0.06 1.28± 0.46 0.21± 0.11
p-value 0.7297 0.0059 0.6806 0.0002 0.5353 0.0001
Table 3.1: Measured MBF and PN from the reference and the accelerated ASL method. Values are reported as mean
± SD across all subjects in ml/g/min unit. Values indicated by (*) are comparable to the results reported in Reference
[32].
baseline images was high in both cases, and this is not expected to be a limitation
of the approach.
Figure 3.3 shows global MBF measured from the reference (dark gray) and
accelerated (light gray) methods. Error bars reflect one standard deviation of
measured physiological noise. The mean and standard deviation of PN across
all subjects was 0.20± 0.08 ml/g/min from the reference method and 0.08± 0.05
ml/g/min for the accelerated method, corresponding to a 60% reduction. PN mea-
sured from the accelerated method was found to be significantly lower than that
from the reference method (p = 0.0059). The mean and standard deviation of
thermal noise (TN) across all subjects was 0.0175± 0.0040 ml/g/min for the ref-
erence method and 0.0241± 0.0045 ml/g/min for the accelerated method, and the
average standard deviation increase was 39% (expected based on the shortened
readout time and g-factor losses). The mean and standard deviation of global
MBF measured with the reference method were 1.20± 0.44 ml/g/min, and mea-
sured with the accelerated ASL method were 1.24± 0.25 ml/g/min. There was no
significant difference between global MBF measured from the two techniques (p =
0.7297). The measured global MBF range of 0.79-2.03 ml/g/min for the reference
method and 0.86-1.52 ml/g/min for the accelerated method are consistent with
PET literature values 0.73-2.43 ml/g/min [153].
Table 3.1 compares the accelerated and reference ASL methods. A significant
difference was found between global, segmental, and septal PN between the two
methods with p-values of 0.0059, 0.0001, and 0.0002, respectively. There was no
significant difference between global, segmental, and septal MBF with p-values of
0.7297, 0.5353, and 0.6806, respectively. Septal MBF and septal PN values for
the reference method were comparable to results reported in Ref. [12], which were
septal MBF = 1.36± 0.38 ml/g/min and septal PN = 0.23± 0.12 ml/g/min.
47
1 2 3 4 5 6 7
0
0.5
1
1.5
2
2.5
Subject number
Global MBF (ml/g/min)
Reference Accelerated
Figure 3.3: Measurements of global MBF from all subjects. (dark gray) Reference
method and (light gray) accelerated method, using rate-2 SENSE. Error bars indi-
cate physiological noise standard deviation. No significant difference was found
between global MBF measured from the two techniques (p = 0.7297). The mean
and standard deviation of PN across all subjects was 0.20± 0.08 ml/g/min from
the reference method and 0.08± 0.05 ml/g/min for the accelerated method, cor-
responding to a 60% reduction. Physiological noise from the accelerated method
was significantly lower than that from the reference method (p = 0.0059).
Figure 3.4 shows a Bland-Altman plot comparing regional MBF (6 segments
per subject) measured from the reference and accelerated method. There was no
significant bias in regional MBF measurements using the accelerated method.
An additional study was performed to investigate whether further shortening
acquisition window helps with PN reduction. The experiments were performed in
5 healthy subjects with the reference method (R1, 300 ms acquisition window),
the accelerated method with SENSE rate 2 (R2, 150 ms acquisition window), and
SENSE rate 3 (R3, 100 ms acquisition window). Figure 3.5 shows a bar plot of
the global MBF from R1, R2, and R3. Error bars represent PN.
48
0 0.5 1 1.5 2 2.5 3
−2
−1.5
−1
−0.5
0
0.5
1
1.5
2
(Accelerated + Reference)/2 (ml/g/min)
(Accelerated − Reference) (ml/g/min)
Mean = 0.06
Mean + 1.96 SD = 1.24
Mean - 1.96 SD = -1.12
Figure 3.4: Bland-Altman plot comparing regional MBF (6 segments per subject)
measured from the reference and accelerated ASL methods. No significant bias
was found between regional MBF measurements by the two methods.
Figure 3.6 shows PN and TN as a function of imaging window in 5 healthy
subjects. Error bars represent standard deviation across 30 data points (5 subject
x 6 segments each subject). As expected, TN is increased with shorter imaging
window. PN is significantly reduced in 150-100 ms acquisition window compared
to 300 ms acquisition window (P < 0.0005). PN in 100 ms acquisition window is
significantly higher compared to 150 ms acquisition window (P = 0.04). That may
beduetothermalnoiseamplificationfromparallelimagingrate3. Inthisstudy, we
observed that PN is still 3X larger than TN. Therefore, there are opportunities for
further improvement of sensitivity. The other confounding factors that contribute
to PN may be heart rate variation and the physiological fluctuations.
Table 3.2 shows global, septal, and per-segment MBF and PN measured from
the reference method (R = 1) and SENSE rate 2 and 3.
49
1 2 3 4 5
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
Subject Number
Global MBF (ml/g/min)

R1 (300ms)
R2 (150ms)
R3 (100ms)
Figure 3.5: Measurements of global MBF from the reference method and the accel-
erated method with SENSE factor of 2 and 3. Error bars indicate PN. No signif-
icant difference was found between global MBF measured from the three tech-
niques (p > 0.41). The mean and standard deviation of PN across all subjects
was 0.21± 0.09 ml/g/min from the reference method, 0.06± 0.01 ml/g/min from
SENSE rate 2, and 0.11± 0.04 ml/g/min from SENSE rate 3.
3.4 Discussion
This study demonstrates that PN in myocardial ASL can be significantly
reduced by shortening the acquisition window using SENSE, which supports our
hypothesisthatmotionduringimageacquisitionwasadominantsourceofPN.Fig-
ure 3.2 shows that motion artifacts from control images in the reference method are
significantly smaller after shortening the image acquisition window to roughly 150
mspercardiaccycle. Wesuspectthatmotionduringtheacquisitionwindowcauses
inconsistencies amongst reconstructed pairs of control and tagged images, which
leads to temporal fluctuations in measured MBF for the reference method. Short-
eningtheacquisitionwindowusingparallelimagingeffectivelydecreasedtheselike-
lihood of such inconsistencies, which in turn reduced PN by roughly 60%. The PN
reduction corresponds to a 158% increase in temporal SNR (tSNR). An increased
50
50 100 150 200 250 300 350
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Acquisition window (ms)
Physiological and Thermal Noise (ml/g/min)

Physiological Noise
Thermal Noise
Figure 3.6: Measured regional PN (square) and TN (triangle) from 5 subjects.
Error bars represent standard deviation across 30 data points (5 subject x 6 seg-
ments each subject). As expected, TN is increased with shorter imaging window.
PN is significantly reduced in 150-100 ms acquisition window compared to 300 ms
acquisition window (P < 0.0005). PN in 100 ms acquisition window is significantly
higher compared to 150 ms acquisition window (P = 0.04). That may be due to
thermalnoise amplificationfrom parallel imagingrate 3. In thisstudy, we observed
that PN is still 3X larger than TN. Therefore, there are opportunities for further
improvement of sensitivity. The other confounding factors that contribute to PN
may be heart rate variation and the physiological fluctuations.
51
52
Global Septal Regional
MBF PN MBF PN MBF PN
(ml/g/min) (ml/g/min) (ml/g/min) (ml/g/min) (ml/g/min) (ml/g/min)
R1 (300ms) 0.97± 0.27 0.21± 0.09 0.95± 0.31
∗
0.24± 0.08
∗
0.98± 0.45 0.30± 0.14
R2 (150ms) 1.03± 0.14
ns
0.06± 0.01
ss
1.02± 0.14
ns
0.12± 0.04
ss
1.05± 0.44
ns
0.15± 0.09
ss
R3 (100ms) 1.02± 0.15
ns
0.11± 0.04
ss
1.03± 0.17
ns
0.15± 0.04
ss
1.03± 0.44
ns
0.19± 0.09
ss
Table 3.2: Measured MBF and PN from R1, R2, and R3. Values are reported as mean± SD across all subjects in
ml/g/min unit. Values indicated by (*) are comparable to the results reported in the reference [32]; Values indicated
by (ns) are not significantly different from R1 (p > 0.41); Values indicated by (ss) are significantly different from R1
(p< 0.04)
tSNR directly translates to improved sensitivity to MBF. Furthermore, MBF mea-
sured using the accelerated method was comparable to those from the reference
method. There was no significant difference found between global MBF (p =
0.7297) and regional MBF (p = 0.5353) measurements from the two methods.
It may seem counterintuitive that the use of parallel imaging could increase
tSNR, when it is known that SNR decreases. Noise analysis from this study
revealed that measured TN increased from 0.0175± 0.0040 ml/g/min to 0.0241±
0.0045 ml/g/min after accelerating by rate 2. While PN decreased from 0.20±0.08
ml/g/min to 0.08± 0.05 ml/g/min after acceleration. There was an overall reduc-
tion in total noise because the increase in TN from the shortened acquisition win-
dow and g-factor losses was much smaller than the decrease in PN from more
consistent reconstructions of control and tagged image pairs. This suggests that
it may be possible to reduce noise further by pushing the acceleration until the
incremental decreases in PN are matched by equal increases in TN.
Although this study only included healthy subjects at rest, one would expect
the accelerated ASL method to also improve sensitivity of myocardial ASL in
patient cohorts at both at rest and stress. Zun et al. [33] reported that global PN
from 16 “normal” patients was 0.64 and 1.36 at rest and stress, respectively, which
is roughly 3 times higher than that measured from healthy subjects, suggesting an
even greater need for PN reduction.
Shortening the acquisition window using SENSE may have the additional ben-
efits of allowing imaging of subjects with high heart rates (>90 beats per minute)
and image during systole (<100 ms acquisition window). In subjects with high
heart rates, diastole is shorter and faster image acquisition may help reduce cardiac
motion artifacts. Likewise, systole is shorter than diastole and may also benefit
from the same type of accelerated image acquisition. In addition, the use of a
53
shorter acquisition window per slice may allow the spatial coverage to be extended
to 2 or more slices per heartbeat, following a single FAIR label.
One important drawback of SENSE is that when the FOV is smaller than the
signal-producing region, SENSE is unable to completely unwrap all aliasing and
can lead to residual artifacts at the center of the reconstructed images. Generalized
autocalibrating partially parallel acquisition (GRAPPA) [60] is a natural alterna-
tive and has been shown to be more robust when the FOV is smaller than the
object [154]. Rapid imaging methods based on partial Fourier and/or constrained
reconstruction may also be appropriate and remain to be explored.
Parallel imaging has been extensively used in first pass CMR perfusion imaging
to improve spatio-temporal resolution and spatial coverage [88]. The primary side
effect associated with its use is reduced SNR based on the shortened readout time
and g-factor losses. Compared to first-pass methods, cardiac ASL methods (based
on apparent-T
1
or signal subtraction) are limited by temporal SNR. This is a
fundamental difference. Many T
1
-based methods and signal subtraction methods
rely on long image acquisition windows, on the order of 30% of the R-R interval,
and it is conceivable that these methods will experience an increase in temporal
SNR by using parallel imaging, as was demonstrated in this study.
There are several inherent assumptions associated with this particular ASL
method. First, we assume a stable heart rate within each breath-hold. Heart
rate variability within a breath hold may result in control and tagged images
being acquired at slightly different cardiac phase. Second, we use baseline image
intensity as a surrogate for M
0
and assume the blood tissue partition coefficient
to be one. In brain ASL, M
0
is commonly estimated by scaling local tissue with
the blood tissue partition coefficient, where the local tissue signal is acquired from
a proton density weighted scan. Third, we assume perfect inversion efficiency for
54
all inversion pulses. Based on our own in vivo measurements the efficiency is
consistently above 94%. Note that the inversion slab thickness of 3 cm does not
introducequantificationerrorswhenusinga1R-Rlabelingdelay, becausethisdoes
not leave enough time for unlabeled blood within the LV blood pool to perfuse the
myocardium. The inversion thickness would, however, require compensation when
using longer labeling delays [148].
3.5 Conclusions
By using parallel imaging to shorten the acquisition window of human myocar-
dial ASL scans, we have demonstrated a significant reduction in physiological
noise, with no significant change in measured MBF. The reduction in PN provides
more temporally consistent measurements of MBF and improves the sensitivity of
myocardial ASL to MBF.
55
Chapter 4
Improved sensitivity of myocardial ASL using
double-gating
4.1 Introduction
Myocardialperfusionandperfusionreserveareimportantindicatorsofcoronary
artery disease (CAD) status. SPECT is the most widely used clinical test for
assessing myocardial perfusion and perfusion reserve, but has limitations related
to the use of ionizing radiation [10, 11]. In recent years, first-pass perfusion MRI
has demonstrated improved sensitivity and specificity [9]. The main limitation of
first-pass perfusion is the required use of Gadolinium-based contrast agents that
can be toxic to patients with renal dysfunction [155, 156].
Arterial spin labeled (ASL) MRI is a non-contrast technique that is capable
of quantifying tissue perfusion [19]. Unlike SPECT and first-pass MRI, ASL uses
blood itself as the tracer, and is therefore completely safe and repeatable. ASL is
widely used in the brain for assessment of neuro-pathological diseases [29, 24] but
its application to the heart is still an active area of research [34]. Myocardial ASL
has been adapted and developed for more than a decade on human and animal
models [137, 142, 157, 158, 143, 131, 32, 33, 140, 139]. Flow-sensitive alternating
56
inversion recovery (FAIR) [30, 31] is the most widely used approach; and has been
implemented in two ways: (i) T
1
apparent approach [137, 142, 157, 158, 143] that
utilizes curve fitting and (ii) subtraction approach [131, 32, 33, 140] that uses
Buxton general kinetic model [129] for quantification.
Both approaches have been shown to quantify myocardial perfusion and per-
fusion reserve. The latter is simpler and has been more widely adopted in recent
years, but it requires the same post labeling delay (PLD) in paired control and
labeled images. This restriction makes it sensitive to heart rate variation (HRV)
since either the labeling pulse or image acquisition can be cardiac-triggered in real-
time, but not both. The paired images are typically acquired 8-10 seconds apart,
within one breath-hold, to minimize respiratory motion. In cases of significant
HRV, the labeling pulses, control images, and labeled images will occur during
different cardiac phases and thus experience inconsistent cardiac motion. This has
been shown to be a dominant source of physiological noise (PN) (18). HRV may
arise from arrhythmias and changes in heart rate due to anxiety, subject motion,
deep breathing, or arousals from sleepiness during the examination. HRV can also
occur during free breathing [159], breathholds [160, 161, 162], physical stress [163],
and pharmaceutical stress with adenosine [164].
Poncelet et al first introduced double-gated myocardial ASL [137] which is
believed to be less sensitive to cardiac motion since it allows both labeling and
image acquisition to occur in the same cardiac phase (i.e. termed “double-gating”)
but, to date, this has not been well-validated experimentally. Poncelet et al suc-
cessfully demonstrated double-gating in swine and healthy human volunteers, but
their technique was prohibitively long with high noise levels even during rest flow
measurements.
57
In this study, we demonstrate that double-gated myocardial ASL is robust to
HRV compared to a single-gated approach in healthy human volunteers and heart
transplant recipients.
4.2 Methods
4.2.1 Experimental methods
The study was approved by our Institutional Review Board, and written
informed consent was obtained from all participants. Ten healthy adult subjects
(n=10, 3F/7M, age 23-34) and eight heart transplant recipients (n=8, 1F/7M, age
26-76) participated in the study. All experiments were performed on a 3 Tesla sys-
tem (Signa Excite HDxt, GE Healthcare) using an 8-chanel cardiac array. Myocar-
dial ASL perfusion imaging was performed using FAIR with balanced steady state
free precession (SSFP) image acquisition.
In each subject, a mid-ventricular short-axis slice was identified. Double-gated
and single-gated myocardial ASL scans were performed in a randomized order.
Scan time was approximately 3 min per scan. Each ASL scan comprised of 7
breath-holds. The first 5-second breath-hold was comprised of a baseline image
(without labeling pulse) and an inversion check (pulsed label applied immediately
before image acquisition). The next six 12-second breath-holds were each com-
prised of one control and one labeled image.
Pulse sequence diagram of single-gating and double-gating are shown in Fig-
ure 4.1. Single-gated myocardial ASL was implemented as previously described
[32, 140] where the labeling pulse is cardiac-triggered (dashed arrows) and the
post-labeling delay (PLD) is kept constant (solid two-sided arrows) for both con-
trol and labeled images.
58
Image Label
Single-gating
Flexible PLD Fixed PLD
Double-gating
ECG trigger
Mid-Diastole
60 bpm
65 bpm
55 bpm
Triger Delay
{
{
{
{
Figure 4.1: Sequence diagram of single-gated and double-gated myocardial ASL.
In single-gating, only labeling pulse is cardiac-triggered (dashed arrows) and the
post-labelingdelay(PLD)iskeptconstant(solidtwo-sidedarrows)forbothcontrol
and labeled images. As such, the imaging window will deviate from mid-diastole
in the presence of HRV (+5 bpm and -5 bpm in the 2nd and 3rd row, respectively).
In double-gating, both labeling and imaging are cardiac-trigged (dashed arrows),
and the PLD is allowed to be different (dashed two-sided arrows) for control and
tagged images. Hence the timing of image acquisition is adjusted in real-time to
be centered around mid-diastole.
As such, the imaging window will deviate from mid-diastole in the presence
of HRV (+5 bpm and -5 bpm in the 2nd and 3rd row, respectively). Double-
gated myocardial ASL was implemented in a similar manner as the single-gated
method except that both labeling and imaging were cardiac-triggered in real-time.
When using double-gating, both labeling and imaging are cardiac-trigged (dashed
arrows), and the PLD is allowed to be different (dashed two-sided arrows) for
control and tagged images. Hence the timing of image acquisition is adjusted in
real-time to be centered around mid-diastole. In this study, the trigger delay (TD)
was set to be 75% of the most recently computed RR interval [73] for both single-
gating and double-gating. In double-gating, precise inversion time was achieved
by using an “adaptive recovery time” method [102]. The adaptive recovery time
method was implemented by playing a series of 2-ms wait-pulses after the inversion
pulse until the next R-wave was detected.
59
Single-gated and double-gated myocardial ASL were performed with identical
sequence parameters that are T
E
= 1.4 (1.3-1.5) ms, T
R
= 3.2 (3.0-3.5) ms, flip
angle = 50
0
, slice thickness = 10 mm, field-of-view = 210 (160-260) mm, matrix
size = 96x96 with parallel imaging GRAPPA [60] rate 1.6, and 19−T
R
Kaiser-
Bessel ramp-up and ramp-down pulses. The Kaiser-Bessel ramp-up is the opti-
mal scheme for mitigating transient oscillations in balanced SSFP imaging [150].
TheKaiser-Besselramp-downoptimallypreserveslongitudinalmagnetizationafter
image acquisition. A ramp duration of 19−T
R
was chosen to match prior work
[140]. A fat-saturation pulse was applied immediately before the ramp-up pulses.
Non-selective hyperbolic secant adiabatic inversion pulse was used for labeling. A
30 mm slice-selective inversion slab was used for the control image. The heart rate
was recorded in all scans for evaluation of HRV.
4.2.2 Data analysis
Left ventricular myocardium was manually segmented for global and per-
segment (6 segments) analysis. Entire left ventricular myocardium region of inter-
est (ROI) was used for global analysis while AHA 6-segment model [104] of a
short axis slice was used for per-segment analysis. Single-gated myocardial blood
flow (MBF) was calculated using Buxton general kinetic model [129] as previously
described[32,140]. Double-gatedMBFwasquantifiedusingthesameequationbut
with interpolated signal difference from control and taggedT
1
curves as previously
described [137]. PN is a measure of the variability of measured MBF, and is mea-
sured in ml/g/min. PN is defined as the standard deviation of six repeated MBF
measurements as described in Zun et al [32]. Both single-gated and double-gated
PN were calculated identically. Temporal SNR (MBF/PN) was also calculated
from the two methods.
60
HRV was defined as the average absolute difference between the instantaneous
heart rate when control and labeled images were acquired. The data was then
divided into two sub-groups: low HRV (n=9, HRV < 4 bpm) and high HRV (n=9,
HRV≥ 4 bpm). MBF, PN, and temporal SNR measured from the two methods
were compared in the two sub-groups separately as well as jointly.
Paired Student’s t-test was used to assess statistical difference between mea-
suredMBF,PN,andtemporalSNRfromsingle-gatinganddouble-gating. P-values
< 0.05 were considered statistically significant. Results are reported as mean±
SD across subjects.
4.3 Results
There were no significant differences in measured heart rates between single-
gating (71± 12 bpm; range, 46-102 bpm) and double-gating (70± 13 bpm; range,
47-102 bpm), (P = 0.17). There were also no significant differences in measured
HRV between single-gating (3.2± 2.2 bpm; range 0.2-6.3) and double-gating (2.7
± 2.3 bpm; range 0.5-7.8 bpm) (P = 0.22).
Figure4.2showsMBFandPNmapsfromtworepresentativesubjectsmeasured
with single-gating and double-gating methods. It can be observed that PN is
lower with double-gating compared to single-gating in both subjects. Measured
MBF from the two methods are similar except for two segments (arrows) in the
first subject. This discrepancy may be explained by higher PN in the same two
segments in single-gating (arrowheads).
The differences in global MBF were not significantly different from the noise
level (P = 0.45), but the differences in per-segment MBF were statistically sig-
nificant from the noise level (P = 0.004). In this study, neither single-gating nor
61
0
1
2
3
4
0
0.1
0.2
0.3
0.4
0.5
MBF
PN
(HR,HRV) (75,6.33)
Single-gating Double-gating Single-gating Double-gating
ANT
ANL
ANS
INS
INF
INL
(76,3.83) (58,4.00) (59,7.83)
(ml/g/min)
Figure 4.2: MBF and PN maps from single-gated and double-gated myocar-
dial ASL in two representative subjects. Lower PN is observed in double-
gating compared to single-gating in both subjects. The MBF maps show
good agreement between the two methods except in two segments (arrows)
that might be explained by higher PN in the single-gating method (arrow-
heads). ANT=anterior; ANS=anteroseptal; INS=inferoseptal; INF=inferior;
INL=inferolateral; ANL=anterolateral.
double-gating is the ground-truth. Therefore, significant differences in MBF mea-
sured from the two methods may be expected, especially in high HRV subjects.
Figure 4.3 shows global PN measurements for double-gating and single-gating
as a function of HRV. Double-gating provided consistent PN across HRV while
single-gating PN increased with HRV.
Table 4.1 compares per-segment MBF, PN, and temporal SNR between single-
gating and double-gating in the two subgroups. There were significant reductions
in PN in low (P=0.04) and high (P<0.001) HRV groups, and improvements in
temporalSNRinlow(P=0.03)andhigh(P=0.004)HRVgroupswithoutsignificant
differences in measured MBF.
62
Figure 4.3: Physiological noise (PN) as a function of HRV for single-gating (circle)
and double-gating (square). Double-gating is robust to HRV (PN did not depend
on HRV). In contrast, single-gating PN increased with HRV.
63
64
Low HRV group (< 4 bpm) High HRV group (≥ 4 bmp)
MBF PN tSNR PN MBF tSNR
(ml/g/min) (ml/g/min) (MBF/PN) (ml/g/min) (ml/g/min) (MBF/PN)
Single-gating 1.40± 0.70 0.25± 0.18 8± 5 1.69± 0.87 0.40± 0.26 6± 4
Double-gating 1.48± 0.65 0.21± 0.12 10± 8 1.59± 0.70 0.22± 0.11 11± 15
P-Value NS (0.18) 0.04 0.03 NS (0.31) < 0.001 0.004
Table 4.1: Comparison of per-segment MBF, PN, and tSNR between double-gating and single-gating with respect to
HRV. Double-gating shows significant reduction in PN and significant increase in tSNR without significant difference
in measured MBF compared to single-gating in both subgroups, but it is more pronounced in the high HRV group.
Values are reported in mean± SD. NS = not significant.
Figure 4.4 shows linear regression and Bland-Altman analysis of measured
global MBF from single-gating and double-gating methods. Bland-Altman reveals
no significant bias and paired student’s t-test shows no significant difference (P >
0.69).
Figure 4.4: Comparisons of global MBF measured from single-gating and double-
gating using linear regression (left) and Bland-Altman analysis (right). Bland-
Altman plot reveals no significant bias between measured MBF from the two
methods (P = 0.69).
The mean± SD of global and per-segment MBF, PN, and temporal SNR from
the two methods for all subjects are listed in Table 4.2. Double-gating demon-
strates significant reductions in global and per-segment PN (P = 0.01 and P <
0.001, respectively) with increased global and per-segment temporal SNR (P =
0.02 and P < 0.001, respectively) compared to the reference single-gating method.
There is no statistically significant difference in measured global and per-segment
MBF between the two methods (P = 0.69 and P = 0.95, respectively).
It is worth noting that residual of fit was used in Poncelet et al (10), as a
surrogate for measurement variability. In our study, global double-gating residual
of fit and global double-gating PN were 0.10± 0.04 and 0.11± 0.03 ml/g/min, (P
= 0.28), respectively. Per-segment double-gating residual of fit and per-segment
65
66
Global (n = 18 subjects) Regional (n = 108 segments)
MBF PN tSNR PN MBF tSNR
(ml/g/min) (ml/g/min) (MBF/PN) (ml/g/min) (ml/g/min) (MBF/PN)
Single-gating 1.51± 0.44 0.20± 0.15 11± 8 1.54± 0.80 0.33± 0.23 7± 4
Double-gating 1.48± 0.45 0.11± 0.03 16± 8 1.54± 0.67 0.21± 0.12 11± 12
P-Value NS (0.69) 0.01 0.02 NS (0.95) < 0.001 < 0.001
Table 4.2: Comparison of global and per-segment myocardial ASL data quality between single-gating and double-gating
from all subjects. Double-gating provides significant PN reduction and tSNR increase without significant difference in
measured MBF compared to single-gating. Values are shown in Mean± SD. NS = not significant.
double-gating PN were 0.18± 0.10 and 0.21± 0.12 ml/g/min, (P < 0.0001),
respectively.
4.4 Discussion
This study demonstrates that double-gated myocardial ASL is robust to HRV
compared to current single-gated techniques, and provides significantly improved
temporal SNR. The current double-gating implementation overcomes two main
limitations of its predecessor: prohibitively long scan times, and high measurement
variability. This study showed that double-gated myocardial ASL is feasible in 3
minutes of scan time, with a lower noise level in MBF measurements in comparison
to the more widely used single-gating method [32, 33, 140].
Single-gating has been widely adopted in recent years due to the ease of imple-
mentation and quantification. This study demonstrated the feasibility of double-
gating with superior temporal SNR efficiency compared to single-gating without
significant difference in measured MBF. Superior temporal SNR directly translates
to superior sensitivity to measured MBF that, in turn, may be used to reduce total
scan time, increase spatial resolution, and/or increase spatial coverage. This find-
ing is expected since in both control and labeled images, labeling and imaging are
triggered to occur in the same cardiac phase. Consequently, double-gating is less
sensitive to HRV and cardiac motion (both in-plane and though-plane) compared
to single-gating.
Over all subjects, double-gating provided significant reduction in PN with
increased temporal SNR. The significant PN reduction and temporal SNR increase
were found in both low and high HRV groups, but were more pronounced in the
67
high HRV group (when HRV≤ 4 beats per minute). PN reduction may be clini-
cally important when myocardial perfusion reserve (MPR =MBF
stress
/MBF
rest
)
is used for evaluating severity of CAD based on a cutoff MPR value. A reduction
of PN of 0.1 ml/g/min would alter MPR by 10% (with an assumption that normal
MBF
rest
= 1.0 ml/g/min).
The cutoff HRV of 4 bpm was chosen based on the PN data splay for single-
gating ASL at HRV > 4 bpm, as seen in Figure 4.1. All subjects were equally
divided into two sub-groups (high HRV versus low HRV) to perform comparisons
between double-gating and single-gating in terms of temporal SNR. It would be
better to relate to different cardiac phases when grouping these subjects with
respect to HRV. However, this would require the knowledge of the quiescent dias-
tolic duration, which is patient specific and heart rate dependent. The mid-
diastolic quiescence has been shown to vary within the same subject even after
normalization to heart rate duration [165].
In our study, neither single-gating nor double-gating is the ground-truth.
Therefore, differences in MBF measured from the two methods may be expected,
especially in subjects with high HRV. Despite the fact that we are comparing two
different ASL methods, the differences in MBF measured from the two methods
in our study are within the range reported in the myocardial ASL literature for
test-retest variability from the same method. Our within subject coefficient of vari-
ation was 12.1%, which is smaller than that (21.8%) reported in Wang et al [131].
The Bland-Altman plot shows a 95% confidence interval of [-0.58, 0.64 ml/g/min],
which is similar to the inter-observer variability [-0.77, 0.47 ml/g/min] reported
in Northrup et al [143]. Our confidence interval is also in the range of inter- and
intra-observer variability reported in Capron et al [139].
68
HRV may arise from numerous sources including sinus arrhythmias, other
arrhythmias, andchangesinheartrateduetoanxiety, patientmotion, deepbreath-
ing, or arousals from sleepiness during the MRI exam. HRV is closely correlated
with image artifacts in CT coronary angiography [161, 162], with a reported HRV
range of 0-18.1 bpm in one study [162] and 10.9± 4 bpm in another [161]. HRV
was also reported in pharmacologic stress tests to be 18± 18 bpm during rest and
16± 22 bpm during adenosine infusion [164]. Therefore, double-gating is expected
to be useful in the clinical setting.
In double-gating, T
1
curve fitting was used as part of MBF quantification. In
this process, a precise inversion time was crucial for accurate T
1
curve fitting as
acknowledged by Zhang et al [158]. We implemented the “adaptive recovery time”
method [102] where, after an inversion pulse, a series of 2-ms wait-pulses are played
until the next detected R-wave to allow precise and consistent measurements of
inversion.
Our double-gated module is identical to that in Poncelet et al [137], but there
are several differences in our implementation including field strength and imaging
sequence. Our experiments were performed on a 3 Tesla scanner compared to
1.5 Tesla in Poncelet’s work. The higher field strength is beneficial for cardiac
ASL because of higher intrinsic SNR and longerT
1
relaxation times, both of which
improvethestrengthoftheASLsignal. Anotherdifferenceisthatweusedsnapshot
balanced SSFP for image acquisition compared to single-shot EPI in the prior
study. Single-shot EPI suffers from image distortion as well as signal loss due
to magnetic field inhomogeneity. Balanced SSFP, on the other hand, is free of
distortion and provides superior SNR efficiency. This sequence was not widely used
at the time of Poncelet’s work because it relies on fast gradients and shimming.
69
Both higher SNR and longerT
1
relaxation time directly translates to the improved
temporal stability and intra-scan variability of our method.
The study has several limitations. Only a small number of healthy subjects
and heart transplant recipients were recruited. Systematic evaluations of double-
gating on larger cohorts are warranted for future study. There was no ground-truth
for MBF in this study. First-pass MRI and SPECT may be used for comparison
but were not available in this cohort. Double-gating may also be validated in an
animal model where first-pass perfusion, micro-spheres, or PET may be used as
the reference standard. Only a single, mid-ventricular short axis slice was acquired
in this study. In this study, inversion was used for labeling pulse; However, the use
of saturation labeling may allow for complete data acquisition within two breath-
holds instead of seven breath-holds, potentially allowing for acquisition of three
sequential slices in the same scan time as single-slice with inversion labeling. The
feasibility and repeatability of saturation double-gated myocardial ASL remains to
be explored in future work.
4.5 Conclusions
This study demonstrates that double-gated myocardial ASL is robust to HRV
incomparisontosingle-gatedmyocardialASL.This, inturn, leadstosuperiortem-
poral SNR efficiency of double-gating compared to single-gating. This is expected
to be valuable for stress testing under physiologic or pharmacologic stress.
70
Chapter 5
Feasibility of non-contrast assessment of
microvascular integrity using ASL-CMR
5.1 Introduction
Microvascular obstruction (MVO) is a common complication after acute
myocardialinfarction(AMI)[166]. MVOisdescribedasa“no-reflow”phenomenon
[167, 168], in which myocardial blood perfusion is impaired at the capillary level
even after reperfusion. Recent studies have established that MVO is indepen-
dently associated with adverse ventricular remodeling and patient prognosis; hence
its detection and careful monitoring is crucial particularly in high-risk patients
[169, 170, 171]. Additionally, microvascular function after an AMI is often com-
promised where vasodilator response is impaired not only in the infarcted but also
in the remote myocardial territories [172]. Quantification of myocardial perfusion
can help assess the presence of MVO as well as evaluate microvascular function.
Myocardial perfusion has been quantified using several techniques such as micro-
spheres, computed tomography, PET, SPECT, and Gadolinium-based first-pass
CMR [7]. The microsphere technique is considered to be the gold standard for
assessment of myocardial tissue perfusion; however, it is invasive and requires
71
the animal to be sacrificed, which is expensive, limits repeatability, and prevents
longitudinal monitoring. Other imaging modalities are noninvasive but they have
limitations either involving ionizing radiation and/or the use of exogenous contrast
agents.
Arterial spin labeling CMR (ASL-CMR) [173] is a non-contrast CMR tech-
nique that is able to quantitatively assess myocardial blood flow (MBF) in small
animals[132,145,147,148,149], largeanimals[137,158]andhumans[32,131,140].
ASL-CMR is also capable of detecting clinically relevant increases in MBF with
vasodilation and has shown potential for identifying coronary artery disease in
patients [142, 33]. Most importantly, ASL-CMR is safe and can provide quantita-
tiveassessmentofMBFovertimeorevencontinuously[174]. ASL-CMRprovidesa
measurement of regional myocardial perfusion in units of ml-blood/ml-tissue/min,
identical to that provided by other imaging modalities.
In this work, we demonstrate the feasibility of ASL-CMR in a porcine model
under rest and stress conditions to assess vasodilator function and to demonstrate
the utility of ASL-CMR in the detection of perfusion deficit/MVO in a porcine
model of AMI. We hypothesize that a significant reduction in MBF is associated
with an MVO. Furthermore, rest and stress MBF measured from ASL-CMR would
allow assessment of vasodilator response not only in the infarcted but also in the
remote myocardial territories post-AMI.
72
5.2 Methods
5.2.1 Animal protocol
Our study utilized female Yorkshire pigs (N=23, 20-25 kg) obtained from
Caughell Farms (ON, Canada) and the animal protocol was approved by the Ani-
mal Care Committee of Sunnybrook Research Institute. Prior to all interventional
procedures and CMR imaging, animals were intubated and sedation was managed
using an anesthetic cocktail of atropine (0.05mg/kg) and ketamine (30mg/kg).
Respiration was controlled (20-25 breaths/min) using a mechanical ventilator and
isoflurane (1-5%) was administered to maintain the anesthetic plane. In a sub-
groupofanimals(N=17), theleftanteriordescendingartery(LAD)wascompletely
occluded for 90 minutes just beyond the second diagonal branch using a percuta-
neous balloon dilation catheter (Sprinter Legend Balloon Catheter, Medtronic).
After 90 minutes, the balloon was released and the vessel was allowed to reperfuse.
The interventional procedures were performed under X-ray fluoroscopy (Philips
Veradius, Philips Healthcare Systems) to guide balloon placement and inflation
and verify reperfusion. Animals were then recovered for subsequent CMR imag-
ing.
5.2.2 CMR imaging
All experiments were performed on a 3T scanner (MR750, GE Healthcare).
The scan protocol and imaging times are listed in Table 5.1. CMR imaging was
performed at baseline (healthy state), day 1-2, week 1-2, and week 4 post-AMI.
Cardiac function was assessed using a cine steady-state-free-precession (SSFP)
sequence (FIESTA, GE Healthcare) with the following parameters: 12-14 short-
axis slices, 3-5 long-axis slices, T
R
/T
E
= 4.0/1.7 ms, flip angle = 45, field-of-view
73
= 24x21.6 cm, acquisition matrix = 224x192, bandwidth = 125 kHz, 8 views-per-
segment and 20 cardiac phases.
Scan time CMR protocol
3 min Localization
10 min CINE (12-14 short-axis, 2-5 long-axis)
3 min ASL-CMR (rest)
3 min ASL-CMR (stress)
1 min First-pass CMR
5 min LGE CMR (8 min post Gad injection)
Table 5.1: CMR protocol
ASL-CMR was performed on mid-ventricular short axis slices that were identi-
fied based on 3-chamber and 4-chamber CINE scout images. Each ASL-CMR scan
was composed of 7 breath-holds and took approximately 3 min. A baseline image
(image without the labeled or control pulse) and a noise image were acquired in the
first 3-second breathhold. 6 pairs of control and label images were acquired in the
subsequent 6 12-second breathholds. Flow-sensitive alternating inversion recovery
(FAIR) was used for labeling and balanced SSFP was used for image acquisition.
The ASL-CMR sequence was described in detail in the reference [140], where the
sequence parameters are: T
R
/T
E
= 3.2/1.5 ms, flip angle = 50, slice thickness =
10 mm, field-of-view = 18-24 cm, acquisition matrix = 128x128, bandwidth = 62.5
kHz, SENSE parallel imaging rate 2 [175]. Labeling and imaging were triggered
to mid-diastole with post labeling delay (PLD) of 2 heartbeats.
ASL-CMR was repeated following an intravenous injection of Dipyridamole
that was used as the pharmacological vasodilator. Stress ASL-CMR was initiated
at 3 min after the intravenous injection of Dipyridamole (DIP, 0.56 mg/kg over 4
min).
74
During the intravenous injection of 8-12 mL of Gadolinium-DTPA (0.2
mmol/kg; Magnevist, Bayer Healthcare, Wayne, NJ), first pass imaging was per-
formed using a multiphase fast gradient-echo sequence to capture the first passage
of the contrast agent. The sequence parameters of first-pass CMR are following:
T
R
/T
E
= 2.8/1.3 ms, slice thickness = 7 mm, slice spacing = 3 mm, flip angle = 20,
acquisition matrix = 128x128. Late gadolinium enhancement (LGE) imaging was
initiated at 8 min post-injection using a T1-weighted inversion recovery gradient-
echo sequence with the following parameters: T
R
/T
E
= 4.1/1.9 ms, flip angle =
15, acquisition matrix = 224x192, 2RR intervals; inversion time was adjusted to
null the signal from normal myocardium (TI=280-320 ms).
5.2.3 Data collection
MBF at rest:
Twenty-three animals underwent functional CINE and ASL-CMR at rest. In
some animals, ASL-CMR was performed for 2-3 adjacent slices, resulting in 41
resting ASL-CMR datasets for analysis.
MBF at rest and stress:
A sub-group of thirteen animals underwent functional CINE and ASL-CMR at
rest and stress. In some animals, ASL-CMR was performed for 2-3 adjacent slices,
resulting in 25 rest-stress ASL-CMR datasets for analysis.
MBF post-AMI:
ASL-CMR was performed in an animal model of AMI that has been previously
validated to consistently present with an MVO [176, 177]. Animals were scanned
at four different time points: baseline healthy state (41 datasets) and day 1-2 (9
datasets), week 1-2 (9 datasets), and week 4 (5 datasets) post-AMI. First-pass
CMR and LGE were also performed in all animals post-AMI.
75
5.2.4 Data analysis
Left ventricular myocardium was manually segmented and the AHA 6-segment
model [104] was applied for regional analysis. Myocardial blood flow (MBF) was
quantified using Buxton’s general kinetic model [129] as previously described [140].
Physiological noise (PN) is a measure of intra scan variability and is measured in
ml/g/min unit. PN was defined as one standard deviation of 6 repeated measure-
ments as described in Zun et al. [32].
Segments with temporal signal-to-noise ratio (tSNR=MBF/PN) < 2 were
excluded from regional analysis in MBF at rest experiment. Similarly, segments
with rest tSNR or stress tSNR < 2 were excluded from regional analysis in MBF
at rest and stress experiments. No data exclusion was used for MBF post-AMI
experiment.
Based on the AMI model, the anteroseptal segment was considered the infarct
region. Three segments (inferior, inferolateral, and anterolateral) were considered
to be the remote region. MBF measured by ASL-CMR from infarct and remote
regions were compared across animal groups over time.
Student’s T-test was used to compare regional MBF at rest and stress from
differentsegmentsandfromdifferentanimalgroups. P-value<0.05wasconsidered
statistically significant. Values are reported as mean ś standard deviation (SD).
5.3 Results
5.3.1 MBF at rest
A total of 69 out of 246 (6 segments x 41 datasets) segments had tSNR < 2 and
were excluded from analysis. Among the excluded segments, 46% and 23% were
76
inferoseptal and inferior segments, respectively, whereas only 7-9% of rejected seg-
ments were from the other 4 segments. Potential reasons for this regional difference
in data quality are provided in the Discussion.
Signal-to-noise-ratio (SNR) in the image without a labeling pulse was 98± 31
(range 37 - 155) consistent with a previous study in humans where SNR was 90±
22 (range 53 - 110) [140]. At baseline (healthy state), regional MBF and PN were
1.08± 0.62 and 0.15± 0.10 (ml-blood/g-tissue/min), respectively.
5.3.2 MBF at rest and stress
A total of 52 and 18 out of 150 segments (6 segments x 25 datasets) had tSNR
< 2 at rest and stress, respectively. Segments with either rest or stress tSNR < 2
were excluded from analysis, resulting in a total of 53 excluded segments in rest
and stress data.
In the rest and stress experiment, regional MBFs were 1.08± 0.54 and 1.47
± 0.62 ml/g/min at rest and during vasodilation, respectively. That represents a
36% increase in MBF during vasodilation. Figure 5.1 shows rest and stress MBF
maps acquired from two representative animals. It was observed that inferoseptal
segments have very low MBF at rest (arrows) but elevated during vasodilation
(arrow heads), possibly due to longer arterial transit time in the right coronary
artery (RCA) branch (see Discussion).
Figure 5.2 shows a comparison between regional MBF at rest and stress mea-
sured from ASL-CMR. Regional MBF was significantly increased with vasodilation
from 1.08± 0.54 to 1.47± 0.62 ml/g/min. Increase in MBF with vasodilation was
found to be significant (P < 0.0001).
77
Rest MBF Stress MBF
ml/g/min
0.8
1.6
2.4
3.2
4
0
Figure 5.1: Rest and stress MBF maps from two representative healthy animals.
Low MBF was observed in inferior and inferoseptal segments at rest (arrows) but
was elevated during vasodilation (arrow heads). Global MBF± PN at rest and
stress are (top row) 0.87± 0.04 and 1.38± 0.02 ml/g/min and (bottom row) 0.78
± 0.16 and 1.39± 0.07 ml/g/min, respectively. Inferoseptal MBF± PN at rest
are (top) 0.03± 0.17 and (bottom) 0.13± 0.25 ml/g/min. These were elevated to
0.62± 0.09 and 1.31± 0.35 ml/g/min during vasodilation, respectively.
5.3.3 MBF post-AMI
Figure 5.3 shows representative examples of MBF maps acquired from ASL-
CMR at day-1, week-1 and week-4 post-AMI. ASL-CMR showed low MBF in the
infarct segment that is consistent with the perfusion deficit seen from first-pass
CMR and MVO seen from LGE.
78
1.0
2.0
3.0
4.0
Rest MBF Stress MBF
Myocardial Blood Flow (ml/g/min)
0.0
P < 0.0001
Figure 5.2: Box plot comparing regional rest and stress MBF measured from ASL-
CMR. Regional MBF was significantly increased with vasodilation from 1.08±
0.54 to 1.47± 0.62 ml/g/min (P < 0.0001).
Figure 5.4 shows resting MBF measurement by ASL-CMR in the infarct and
remote regions from animal groups over time. Resting MBF measured in the
infarct regions were significantly lower than in remote and healthy myocardium (P
< 0.04). The significant reduction in resting MBF is consistent with the perfusion
deficitseenfromfirst-passCMRandMVOseenfromLGE.Therewasnosignificant
differencebetweenmeasuredrestingMBFintheremotemyocardiumacrossanimal
groups and time (P > 0.26). It is worth noting that MVO typically resolves by
week 4 post-AMI. Hence low MBF in week 4 may be attributed to the absence of
vessels in the infarcted territories.
5.4 Discussion
The results of this study demonstrate the feasibility of ASL-CMR for the con-
trast free assessment of microvascular integrity. Specifically, ASL-CMR is capable
79
ASL-CMR First-pass LGE
ml/g/min
Day 1 post-AMI
Week 4 post-AMI

0.8
1.6
2.4
3.2
4
0
MVO
Perfusion
Deficit
Low
MBF
Week 1 post-AMI
Figure 5.3: Representative resting MBF maps from ASL-CMR in day-1, week-1
and week-4 post-AMI. Low MBF at rest measured in the infarct zone (arrows) is
consistent with a perfusion deficit seen from first-pass CMR and MVO seen from
LGE.
of quantitative assessment of regional myocardial blood flow at rest, under vasodi-
lation and also in the presence of MVO post-AMI. Regional MBF measured from
ASL-CMR was consistent with literature values both at rest and stress where ASL-
CMR can detect a mild increase of approximately 36% in MBF with vasodilation
in pigs. Significantly lower MBF measured from ASL-CMR in the infarct zone was
consistent with the perfusion deficit identified from first-pass CMR and MVO seen
from LGE.
Despite differences between human and pig, our results demonstrate that ASL-
CMR can be successfully applied in a pig model for quantitative assessment of
80
Remote Region Infarct Region
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
Myocardial Bblood Flow (ml/g/min)

Baseline
Day 1&2
Week 1&2
Week 4
NS (P > 0.26)
P < 0.04
*
*
*
Figure 5.4: Regional resting MBF measured in remote (inferior, inferolateral, and
anterolateral combined) and infarct (anteroseptal) region at baseline and post-
AMI. Error bars represent PN. There was a significant reduction in resting MBF
post-AMIcomparedtobaseline(P<0.04)thatisassociatedwithMVO.Therewas
no significant difference in resting MBF measured in the remote region between
post-AMI time points (P > 0.26).
regionalMBF.RegionalMBFmeasuredinthisstudy(1.08±0.62ml/g/min)using
ASL-CMR was consistent with previous studies in pigs where measured MBF was
1.30± 0.60 and 1.00± 0.40 ml/g/min using first-pass CMR and microspheres,
respectively [178]. Additionally, regional PN in this study (0.15± 0.10 ml/g/min)
using ASL-CMR was also comparable to a previous human study where measured
PN was 0.21± 0.11 ml/g/min [140].
Quantitative assessment of regional MBF is crucial for detection of focal perfu-
sion defect as well as monitoring temporal changes. It was learned from this study
that caution is needed before translating a human protocol to animals and vice
versa. In our human study, we typically use PLD of 1RR; however, low ASL signal
was observed in animals when PLD of 1RR was used. Therefore, a 2RR PLD was
used in this study. At rest, we noted a lower MBF measurement in inferoseptal
81
and inferior segments compared to other regions. We hypothesize that this could
be attributed to longer arterial transit time (ATT) in the RCA.
Technique Rest MBF Stress MBF
(ml/g/min) (ml/g/min)
Schmitt et al., [178] Microspheres 1.00± 0.40 NA
First-pass CMR 1.30± 0.60 NA
Mahnken et al., [179] CT perfusion 0.98± 0.19 1.34± 0.40
This study ASL-CMR 1.08± 0.54 1.47± 0.62
Table 5.2: Rest and stress MBF measured from ASL-CMR in comparison with
literature values.
This study shows that ASL-CMR is able to detect increase in regional MBF
with vasodilation from 1.08± 0.54 to 1.47± 0.62 ml/g/min (P < 0.0001). Approx-
imately 36% increase in MBF was detected that is consistent with another study
in pigs [179] as listed in the Table 5.2. This ability makes ASL-CMR a potential
tooltostudymicrovasculardysfunctionandvasodilatorresponsepost-AMIinboth
remote and infarcted territories. Availability of quantitative measure for regional
vasodilation response plays an important role in assessment of microvascular dys-
function as demonstrated by Uren at al., where microvascular function is not only
compromised in the infarct territories but also in the remote territories [172]. It is
worth noting that the MBF increase with vasodilation in pigs is significantly lower
than that in human, where the MBF increase is approximately 300% [153]. This
blunted vasodilator response may be explained by the fact that these animals were
under anesthesia of isoflurane, which was reported to have coronary vasodilation
effects [180, 181, 182, 183, 184]. For example, Gamperl et al., demonstrated that
isoflurane caused vasodilation in isolated swine coronary microvasculature that is
dose-dependent and endothelium-dependent [184].
This study demonstrates that ASL-CMR can detect MVO where low MBF
measured from ASL-CMR is consistent with perfusion deficit in first-pass CMR
82
and MVO in LGE. Consistent with the 90 min occlusion model [176] at day 1,
all animals demonstrated a perfusion deficit on first-pass CMR and a MVO on
LGE within the anterior infarcted territory. A significant reduction in MBF was
detected by ASL-CMR in the infarct region post-AMI compared to baseline that
is consistent with perfusion deficit in first-pass CMR and MVO in LGE as shown
in Figure 3 and 4. MVO is one of the common complications after reperfusion
[166] and previous studies showed that MVO is independently associated with
adverse ventricular remodeling and poor patient prognosis [169, 170, 171, 185].
Therefore, early detection and serial assessment of MVO plays an important role
in management post AMI patient population that may improve patient prognosis
and prevent recurrent AMI.
Quantitative assessment of regional MBF over time is important for monitoring
treatment efficacy. That in turn provides a quantitative measure that potentially
guides treatment plans and the development of new drugs and related therapies.
Several methods have been used for quantitative assessment of regional MBF that
include CT perfusion, SPECT, and Gadolinium first-pass CMR. These imaging
modalities have limitations either utilizing ionizing radiation and/or exogenous
contrast agents. Radiation used in modalities such as CT perfusion and SPECT
may limit frequent assessment of MBF. Exogenous Gadolinium contrast agents
used in first-pass CMR may be contraindicated in patients with renal insufficiency.
ASL-CMR, on the other hand, is safe, repeatable, and a direct measure of tissue
perfusion that makes it an ideal tool to study microvascular and coronary func-
tion. Assessment of MVO and microvascular dysfunction would be such potential
applications. Since MVO is defined as a “no-reflow” phenomenon, MBF would be
a direct measure of MVO and its severity. A previous study demonstrated that
vasodilator response that is an indicator of coronary and microvascular function
83
was impaired not only in the infarct but also in remote territories [172]. ASL-
CMR would be a safe and repeatable method to assess vasodilator response both
spatially and temporally.
Limitations
The main limitation of this study was the lack of validation of perfusion mea-
surements (both rest and stress states) against a quantitative gold standard. Alter-
native methods for comparison could be first-pass CMR perfusion, PET/SPECT
study, invasive FFR/CFR, or a histology-based (gold standard) microsphere tech-
nique. Nevertheless, we were able to utilize the MVO-model to verify the detection
of the perfusion deficit by ASL-CMR. This was a feasibility study that motivates
such future work.
This study only interrogated mid ventricular short-axis slices. Currently, ASL-
CMR takes 3 min per imaging slice and multiple slices are acquired in sequential
order. That is lengthy and may not be practical in patients because available
time for image acquisition during vasodilation is limited to approximately 3 min.
Future studies will investigate the use of simultaneous multiple slice [186] that
allows acquisition 3 to 4 simultaneous multiple slices in a single ASL-CMR run.
This study observed regional differences in data quality, specifically inadequate
tSNR in the inferior and inferolateral segments at rest. We noted below normal or
close to zero perfusion values in the inferoseptal and inferior segments (20/25 and
13/25, respectively) of some animals in the resting state, hence these segments
were excluded from the analysis. However, a lower number of inferoseptal and
inferior segments were excluded under the vasodilated or stressed state (15/25 and
3/25, respectively). This suggests that the low resting inferoseptal and inferior
values are not directly arising from a sequence limitation but are rather related to
coronary architecture, number of branches, vascular resistance and flow patterns.
84
Although differential filling of the coronary vessels is currently unclear in humans
[187], an initial qualitative analysis of coronary X-ray angiograms of pigs in our
study has revealed slower filling of the right coronary artery (RCA) compared to
the left circumflex (LCX) and left anterior descending (LAD) artery; the RCA is
known to provide blood supply to the inferior segments. We are investigating this
further by performing ASL-CMR with post-labeling delays of 1RR, 2RR, and 3RR.
Our preliminary data (not shown) indicates that increasing the post labeling delay
recovers ASL signal in the inferior segments, which suggests that transit times are
likely responsible for this regional variation. Next generation ASL-CMR methods
can be designed to account for regional variations in arterial transit time (ATT).
This study did not observe any statistically significant changes in MBF in
remotemyocardiumpost-AMI.Thismaybeduetothelowsamplesizeateachtime
point post-AMI. Secondly, to observe remote myocardial response, it is possible
that vasodilator response might need to be evaluated, which was not performed
in this study. A previous study has demonstrated T
2
-BOLD response alterations
in a porcine AMI model [188]; future studies could combine myocardial BOLD
response and rest-stress ASL measurements.
5.5 Conclusions
ASL-CMR is a non-contrast technique that is able to quantitatively assess
regional MBF at rest and under vasodilation, as well as detect changes in
regional MBF post-AMI. ASL-CMR could potentially be used to detect and mon-
itor microvascular injury/obstruction and microvascular function not only in the
infarcted region but also the salvageable and remote regions, which may be early
indicators of downstream adverse remodeling processes post-injury.
85
Chapter 6
Feasibility of non-contrast assessment of MBF
and MBV change using T
1
mapping
6.1 ASL-CMR using MOLLI
Modified Look-Locker Inversion Recovery (MOLLI) provides the highest pre-
cision and reproducibility for myocardial T
1
mapping, and extracellular volume
(ECV)mapping. Inthiswork, wedetermineitseffectivenessformeasuringmyocar-
dial blood flow (MBF), based on apparent-T
1
mapping under two conditions, slice-
selective inversion and non-selective inversion. We demonstrate that MOLLI pro-
vides measured MBF comparable to the reference FAIR-SSFP ASL method.
6.1.1 Introduction
Myocardial blood flow (MBF) is an important indicator of micro-vascular and
coronary artery dysfunction. Arterial spin labeled (ASL) CMR is a non-contrast
technique that can quantitatively measure MBF either using signal subtraction
[32, 131, 33, 140] or apparent T
1
approaches [158, 143]. MOLLI has been shown
to provide the highest precision and reproducibility among T
1
mapping methods
[103]. This study investigates the feasibility of using MOLLI 5(3s)3 [90] apparent
86
T
1
mapping for MBF quantification in humans. We compare MOLLI-based ASL
with conventional flow-sensitive alternating inversion recovery (FAIR) ASL.
6.1.2 Methods
Our Institutional Review Board approved the study protocol, and informed
consent was provided by all subjects. Six healthy adults were enrolled in the
study (5M/1F, age 22-32). Two subjects received a second scan on a separate day,
resulting in a total of 8 datasets. All experiments were performed on a 3T scanner
(Signa Excite HDxt, GE Healthcare). MOLLI 5(3s)3 [90] and FAIR sequences
were performed in all subjects. Scan time was 3 min for FAIR (7 breath-holds)
and1minforMOLLI(2breath-holds, 1withslice-selectiveinversion, andtheother
withnon-selectiveinversion). Imagingparameterswerethesameforbothmethods:
FOV=180-280 mm, slice thickness = 10 mm, matrix size = 96x96, GRAPPA factor
= 1.6 [60]. Flip angle was 50
0
and 35
0
for FAIR and MOLLI, respectively.
The myocardium was manually segmented for global and per-segment (6 seg-
ments) analysis. MBF in FAIR was analyzed in the same manner as described
previously [140]. MBF in MOLLI was quantified using the equation [132]:
F =
λ
T
1blood
·

T
1NS
T
1SS
− 1

(6.1)
where F is myocardial blood flow,λ andT
1blood
were chosen consistently with MBF
quantification in FAIR that are 1 ml/g and 1650 ms, respectively. T
1NS
and T
1SS
were obtained using nonlinear regression from MOLLI slice-selective inversion and
non-selective inversion series, respectively. Global and per-segment MBF of the
two methods were compared using paired Student’s T-Test. Results were reported
as mean± SD.
87
6.1.3 Results
0
500
1000
1500
2000
SS
NS
T1SS map T1NS map ml/g/min
0
1
2
3
4
ms MBF map
Figure 6.1: Image series from slice-selective inversion (top row) and non-selective
inversion (bottom row) and corresponding T
1
and MBF maps. Measured MBF
from this subject is 0.95 ml/g/min. SS=slice-selective; NS=non-selective.
Figure 6.1 shows T
1
maps acquired from MOLLI using slice-selective inversion
(left) and non-selective inversion (right) from a representative subject.
Global and per-segment T
1NS
and T
1SS
are summarized in the Table 6.1, con-
sistent with MOLLI literature range of 1005 to 1296 ms [189].
T
1NS
was significantly higher than T
1SS
(P < 0.001) that is consistent with
inflow of fresh blood to the imaging slice as shown in Figure 6.2.
There was no significant difference between global and per-segment MBF mea-
sured from the MOLLI compared to those measured in the reference FAIR method
(p>0.57) as shown in Figure 6.3. Scatter and Bland-Altman analysis show no sig-
nificant bias measured MBF from the two methods.
88
950 1000 1050 1100 1150 1200 1250
950
1000
1050
1100
1150
1200
1250
T1SS (ms)
T1NS (ms)
identity
T1SN = 0.98*T1SS + 60.5
R
2
= 0.95
p < 0.001
Figure 6.2: T
1SS
andT
1NS
(ms) measured from MOLLI.T
1NS
is significantly higher
thanT
1SS
(P<0.001)becauseofthecontributionofun-invertedinflowbloodwhen
slice-selective inversion is used.
0.5 1 1.5 2 2.5
0.5
1
1.5
2
2.5
MBF from FAIR (ml/g/min)
MBF from MOLLI (ml/g/min)

identity
Y = 0.95*X + 0.025
R
2
= 0.41
P < 0.001
1 1.5 2 2.5
−1
−0.5
0
0.5
1
(MOLLI+FAIR)/2 (ml/g/min)
MOLLI-FAIR (ml/g/min)
Mean = -0.05
Mean + 1.96 SD = 0.6
Mean + 1.96 SD = -0.7
Figure 6.3: Comparison of global MBF measured between MOLLI and FAIR using
scatter(left)andBland-Altmanplot. Thereisnosignificantdifferenceinmeasured
MBF between the two methods (P > 0.67).
89
T
1NS
T
1SS
MOLLI-ASL FAIR-ASL P-value
MBF MBF
(ms) (ms) (ml/g/min) (ml/g/min)
Global 1108± 56 1065± 56 1.47± 0.44 1.52± 0.29 0.67
Regional 1109± 76 1066± 72 1.48± 0.85 1.55± 0.58 0.57
Table 6.1: Measured T
1
and MBF from MOLLI. T
1
measured from MOLLI with
the non-selective inversion (T
1NS
) was the range reported from reference [189],
where the range was 1005 to 1296 ms . T
1NS
is significantly higher thanT
1SS
(P <
0.001) that is expected sinceT
1SS
is shorten due to inflowing of fresh blood. There
was no significant difference between global and per-segment MBF measured from
the two methods (P > 0.57). Values were reported as mean± SD.
6.1.4 Conclusions
This study demonstrates the feasibility of using MOLLI for assessment of MBF
on healthy subjects. The preliminary results show that MBF measured from
MOLLI is in good agreement with that measured in the reference FAIR method.
Multi-slice MOLLI has been routinely used in clinic. MOLLI ASL would simulta-
neously provide both MBF and T
1
maps. Complete physiological noise analysis is
a work in progress.
6.2 Vasodilator response in heart transplant
recipients using T
1
-based MBV mapping
6.2.1 Introduction
Myocardialspin-latticerelaxation time(T
1
)is sensitive towatercontent related
to fibrosis and to myocardial blood volume (MBV). Previous studies have demon-
strated that native T
1
is able to differentiate patients in several disease conditions
from normal controls. Assessment of fibrosis and blood volume plays an important
90
role in management of heart transplant recipients where fibrosis formation and
reduction in MBV often precede allograft rejection. We hypothesize that native
myocardialT
1
is higher in transplant recipients due to fibrosis, and that this leads
to reduced MBV change during vasodilation. This study applies myocardial T
1
mapping at rest and adenosine stress to assess native T
1
and T
1
change (ΔT
1
), as
potential surrogates of myocardial fibrosis and MBV change in heart transplant
recipients.
6.2.2 Methods
ThisstudywasapprovedbyourInstitutionalReviewBoardandallparticipants
providedwritteninformedconsent. MOLLI5(3s)3myocardialT
1
mappingwasper-
formed at rest and during adenosine stress in five heart transplant recipients (4M,
age 49 pm 16) and five healthy controls (4M, age 28± 4). Imaging parameters:
TR/TE = 3.2/1.5 ms, flip angle = 35
0
, slice thickness = 10 mm, field-of-view =
160-240cm, receiverbandwidth=62.5kHz, acquisitionmatrix=96x96, GRAPPA
factor 1.6x. Pixel-by-pixel T
1
maps were generated by Levenberg-Marquardt non-
linear curve fitting. Left ventricular myocardium was manually segmented and
regional analysis performed using the AHA 6-segment model. ΔT
1
was calculated
by the following formula [94]:
ΔT
1
= 100%·
T
1stress
−T
1rest
T
1rest
, (6.2)
whereT
1rest
andT
1stress
areT
1
maps measured at rest and during peak Adenosine
vasodilation. Values are reported as mean± SD. Paired Student’s T-Test was used
to compare regional myocardial T
1
and ΔT
1
from heart transplant recipients and
healthy subjects. P-value < 0.05 was considered to be statistically significant.
91
6.2.3 Results
Rest T
1
Stress T
1
0
500
1000
1500
2000
(ms)
T
1
change
(%)
2
4
6
8
10
Healthy
Control
Heart
Transplant
Figure6.4: RepresentativeT
1
and ΔT
1
mapsofahealthycontrolahearttransplant
recipient. Global (rest, stress) T
1
in the healthy subject and the patient are (1196
pm 67 ms, 1287 pm 72 ms) and (1249± 121 ms, 1295±100 ms). Higher T
1
and
spatial inhomogeneity are observed in patient compared to healthy control. That
may be indication of diffuse fibrosis and cardiomyocyte hypertrophy and that in
turn leads to smaller ΔT
1
in patients. ΔT
1
in the healthy subject and the patient
are 7.82± 1.95 % and 4.36± 2.61 %, respectively.
Figure 6.4 shows representative pixel-by-pixel T
1
maps at rest and stress from
a healthy subject and a heart transplant recipient, and the corresponding regional
ΔT
1
maps.
Figure 6.5 and Table 6.2 summarize regionalT
1
at rest and stress from all sub-
jects and the corresponding regional ΔT
1
. In healthy subjects, native myocardial
T
1
and ΔT
1
were 1133± 84 ms and 5.41± 3.12 %, respectively, consistent with
literature values. The main findings of this study are 1) native T
1
at rest was
significantly higher (p < 0.0001) in heart transplant recipients consistent with the
hypothesis of fibrosis and cardiomyocyte hypertrophy formation; 2) ΔT
1
(surro-
gate for MBV change) was significantly lower (P = 0.0222) in heart transplant
92
Healthy
Control
Heart
Transplant
1000
1050
1100
1150
1200
1250
1300
1350
1400
Myocardial T
1
(ms)

Rest T
1
Stress T
1
0
1
2
3
4
5
6
7
8
9
10
Myocardial T
1
change (%)
Healthy
Control
Heart
Transplant
*
**
**
**
**
(A) (B)
Figure 6.5: (A) Regional myocardialT
1
at rest and stress and (B) regional myocar-
dial ΔT
1
fromhealthycontrolandhearttransplantrecipient. (**)and(*)indicates
statistical difference with P < 0.0001 and P = 0.0222, respectively.
Healthy
Control
Heart
Transplant
Rest MBF
Stress MBF
Healthy
Control
Heart
Transplant
(A) (B)
0
1
2
3
4
5
6
MBF (ml/g/min)
0
1
2
3
4
5
Perfusion reserve
**
**
NS
NS
NS
Figure 6.6: (A) Regional myocardial MBF at rest and stress and (B) regional
myocardial MPR from healthy control and heart transplant recipient. (**) indi-
cates statistical difference with P < 0.0001 and NS = Not Significant.
93
recipients, which may be a consequence of fibrosis and cardiomyocyte hypertrophy
and/or impaired vasodilator response.
Rest T
1
Stress T
1
T
1
change
(ms) (ms) (%)
Healthy Control 1133± 84# 1095± 103* 5.41± 3.12#
Heart Transplant 1234± 42 1281± 42* 3.83± 3.11
P-value <0.0001 <0.0001 <0.0001
Table 6.2: Summary of per-segment myocardialT
1
at rest and stress andT
1
change
(ΔT
1
) measured from healthy controls and heart transplant recipients. (*) indi-
cates significant increase in myocardial T
1
with adenosine stress compared to rest
(P < 0.0001). (#) indicates the resting native T
1
and ΔT
1
measured from this
study are consistent with literature values, which are 1005-1295 ms, and 6.2± 0.5
%, respectively [94].
6.2.4 Conclusions
This study demonstrates that adenosine stress and rest myocardialT
1
mapping
is potentially able to assess regional MBV change in heart transplant patients
without the use of a contrast agent. That would potentially be a useful method
for spatial and temporal assessment of myocardial tissue function in patients with
cardiovascular disease especially those with renal inefficiency.
94
Chapter 7
Physiologically synchronized multi-module
pulsed arterial spin labeled (SymPASL) MRI
Physiologically synchronized multi-module pulsed arterial spin labeling (Sym-
PASL) involves pulsed labeling that is applied several times prior to pulsations in
the arterial blood supply. Simulations and in vivo measurements in human kidneys
demonstrate that SymPASL provides superior SNR and SNR efficiency compared
to conventional flow-sensitive alternating inversion recovery (FAIR) ASL with a
single labeling pulse. Simulations suggest that SymPASL provides comparable
SNR and SNR efficiency to pseudo-continuous ASL (PCASL), with lower specific
absorption rate (SAR).
7.1 Introduction
Arterial spin labeling (ASL) is a non-invasive and non-contrast MR method
that can quantify regional tissue perfusion. It is extensively used in the brain [29],
wherepseudo-continuouslabeling(PCASL)[116]isthemostwidelyusedapproach.
Application of ASL to the kidneys and the heart is an active area of research, and
the ideal labeling schemes may differ due to the complex geometry of feeding
95
vessels, pulsatile blood supply, quasi-periodic motion, and field inhomogeneity.
In the kidneys and the heart, pulsed labeling is most widely used, specifically the
flow-sensitive alternating inversion recovery (FAIR) technique [32, 33, 190]. In this
work, we demonstrate, through simulation and in vivo experiments, the superior
SNR efficiency of using multiple pulsed labeling modules appropriately timed with
the cardiac cycle. The proposed technique is termed physiologically synchronized
multi-module pulsed ASL (SymPASL).
7.2 Methods
7.2.1 Simulation
SymPASL Label Imaging
~150ms ~10ms
ECG signal
Abdominal Aortic Flow Waveform
Figure 7.1: SymPASL sequence timing with labeling pulses (red bars), and imag-
ing (blue box). Both labeling and imaging are ECG triggered in real-time (dashed
gray arrows), with trigger delays optimized depending on the target organ. Label-
ing pulses are played right before the rise of the abdominal aortic flow waveform
(sketched based on Reymond et al., [191]).
SymPASL timing is shown in Figure 7.1. Multiple labeling pulses (red) are
applied before image acquisition (blue). Labeling and imaging are triggered based
on the time-velocity profile of the arterial blood supply, and the target organ
motion pattern, respectively. Buxton’s general kinetic model [129] was used to
simulate ASL signal (S
ASL
) for PCASL, FAIR, and SymPASL. Simulation param-
eterswere: T
1blood
=1932ms[37], arterialtransittime(ATT)=1/3ofR-Rinterval,
96
instantaneous water exchange, bolus duration = 1 R-R. Heart rate = 64± 6 bpm
and T
1tissue
= 1584± 159ms were measured in vivo. An inversion label was used
for FAIR whereas a saturation label was used for SymPASL. A 1500ms labeling
duration and 1000ms post-labeling-delay (PLD) were used for PCASL simulation
[192].
7.2.2 Experiment
Six healthy subjects were studied (5M/1F, age 23-30 yrs). Experiments were
performed on a 3T scanner (Signa Excite HDxt, GE Healthcare). One FAIR scan
with 1 R-R PLD and six SymPASL scans, with 1, 2, 3, 4, 5, and 6 labeling pulses,
were performed in all subjects. Each scan was comprised of 6 breath-holds, each
comprised of one labeled and one control image. Imaging parameters for both
FAIR and SymPASL were: FOV=180-280 mm, slice thickness = 10 mm, flip angle
= 50 degrees, matrix size = 96x96, and GRAPPA factor =1.6. Heart rate was
recorded in all scans, renal cortexT
1tissue
was estimated using nonlinear regression
from global inversion recovery FAIR data. These values were used as subject
specific input parameters for simulation for comparison with in vivo measurement.
Labeling efficiency of both FAIR and SymPASL were measured in all subjects.
7.2.3 Analysis
A region of interest (ROI) containing 10-15% ( 180 mm
2
) of the renal cortex
within the slice was conservatively drawn to avoid partial volume effects. A refer-
ence ROI was drawn on the latissimus dorsi muscle (back muscle). The ASL signal
S
ASL
was calculated using:
S
ASL
=
C−L
B
, (7.1)
97
where C, L, and B are control, labeled, and baseline image. SNR efficiency
(SNR
eff
) was defined as:
SNR
eff
=
S
ASL
√
2·T
R
, (7.2)
where S
ASL
is the ASL signal and 2·T
R
is the total scan time to acquire a
pair of control and labeled image [193]. Physiological noise (PN) was estimated
to be the standard deviation of S
ASL
from 6 measurements [32]. Linear regression
was used to compared measured renal blood flow (RBF) quantified from FAIR and
5-label SymPASL.
7.3 Results
Figure 7.2 contains simulation and experimental data from a representative
subject. Consistent with simulation, S
ASL
increases with the number of labeling
pulses. PN are similar in all scans. S
ASL
and SNR efficiency measured in renal
cortex matched with simulation results while measured values in the latissimus
dorsi muscle are within the noise level.
Figure 7.3 shows S
ASL
and SNR efficiency measured from all six subjects in
comparison with simulations. Error bars represent the standard deviation across
subjects. Renal cortex S
ASL
and SNR efficiency are in good agreement with sim-
ulations. Additionally, simulated values from PCASL were added for comparison.
It is suggested that SymPASL can provide similar S
ASL
and SNR efficiency as
PCASL but with lower SAR.
Figure 7.4 shows correlation plot comparing RBF measured from FAIR and
SymPASL. A strong correlation (R = 0.93) was found between measured RBF
from the two methods. The discrepancy of 0.75 ml/g/min between measured RBF
98
0
0.05
0.1
0.15
0.2
FAIR SymPASL
1 label 2 labels 3 labels 4 labels 5 labels 6 labels
Cortex
ROI
Muscle
ROI
(A)
1 2 3 4 5 6 7 8
−0.04
−0.02
0
0.02
0.04
0.06
0.08
0.1
Number of labeling pulses
ASL Signal (a.u.)
FAIR - cortex
SymPASL - cortex
FAIR - muscle
SymPASL - muscle
SymPASL - cortex (sim)
FAIR - cortex (sim)
(B)
Figure 7.2: In vivo evaluation. (A) S
ASL
maps. (B) S
ASL
from ROIs on the renal
cortex and latissimus dorsi muscle of a representative subject. Error bars on panel
B represent physiological noise (PN), which is defined as standard deviation of 6
measurements. Measured renal cortex S
ASL
is in good agreement with simulation
while values on the muscle are within the noise level.
(A) (B)
1 2 3 4 5 6 7 8
−0.01
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
Number of labeling pulses
ASL Signal (a.u.)

1 2 3 4 5 6 7 8
−5
0
5
10
15
20
Number of labeling pulses
SNR efficiency (a.u.)

FAIR - cortex
SymPASL - cortex
FAIR - muscle
SymPASL - muscle
SymPASL - cortex (sim)
FAIR - cortex (sim)
PCASL - cortex (sim)
Figure 7.3: Measured (A) S
ASL
and (B) SNR efficiency from all six subjects in
comparison with simulations. Error bars are group standard deviations. Sym-
PASL provides superiorS
ASL
and SNR efficiency as compared to FAIR. Simulated
PCASL S
ASL
and SNR efficiency (1500 ms labeling duration and 1000 ms PLD)
are shown in blue, and that are comparable to SymPASL. As shown in SNR effi-
ciency plot, the optimal number of labels for SymPASL are 4-5. That is expected
because labeled blood is decayed away with a time constant of T
1blood
= 1932 ms.
Further adding labeling pulses does not significantly increase S
ASL
but the cost of
scan time.
99
Figure 7.4: Linear regression plot comparing RBF measured from FAIR and Sym-
PASL. RBF measured from the two methods are strongly correlated (R = 0.93).
Measured RBF are 2.98± 0.78 and 3.72± 0.73 ml/g/min from FAIR and 5-label
SymPASL, respectively. There was a bias of 0.75 ml/g/min between the two meth-
ods that may be explained by different labeling efficiency and sensitivity to arterial
transit time.
is due to different labeling efficiency and different sensitivity to ATT of the two
methods.
7.4 Conclusion
This study introduces a variant of pulsed labeling, termed SymPASL, that
takes advantage of the pulsatile blood supply. SymPASL provides superior SNR
efficiency compared to FAIR, and is likely to be advantageous for 3D ASL of the
kidneys and heart. Simulations and in vivo measurements in the human kidneys
confirm the advantages of SymPASL over FAIR.
100
Chapter 8
Conclusions
Myocardial arterial spin labeling (ASL) is a promising technology for perfu-
sion stress testing yet to be developed for clinical use. Unlike existing tests,
myocardial ASL does not require the use of exogenous contrast agents or ion-
izing radiation. Absence of contrast and radioactive agents makes myocardial ASL
attractive to coronary artery disease (CAD) screening in patients with kidney dys-
function, includingmorethan20millionchronickidneydisease(CKD)and610,000
end-stage renal disease (ESRD) patients in the United States. However, several
challenges remain, including low sensitivity, coarse spatial resolution, and limited
spatial coverage.
For quantitative cardiovascular magnetic resonance (CMR), managing cardiac
motion is essential. As subtraction technique with effective ASL signal of only
about 1-4% of the acquired image signal, managing cardiac motion is even more
critical in myocardial ASL. Any sources of variation can corrupt the tiny ASL
signal. Chapter 3 demonstrated that shortening the image acquisition window
significantly improves sensitivity of myocardial ASL. In this work, we found that
shortening the acquisition window from 300 ms to 150 ms using parallel imaging
101
significantlyreducesmeasurementvariabilitywhichimprovessensitivityofmyocar-
dial ASL.
Single-gating is more widely adopted in myocardial ASL compared to double-
gating because of its simpler implementation and quantification. However, single-
gating assumes constant heart rate, which is not often the case in practice. Double-
gating allows both the labeling and the image acquisition to be cardiac triggered in
real-time. Thatmakesdouble-gatinglesssensitivetoheartratevariation(HRV).In
chapter 4, I implemented double-gating myocardial at 3.0 Tesla and quantitatively
compared its performance to single-gating. Double-gating consistently provides
higher sensitivity compared to single-gating. Furthermore, this work demonstrated
that double-gating is robust to HRV while single-gating suffers from low sensitivity
due to HRV.
Robustness and sensitivity of myocardial ASL are improved with parallel imag-
ing and double-gating. That enables exploration of potential clinical applications
using myocardial ASL. Microvascular obstruction (MVO), i.e. “no-reflow phe-
nomenon”, is a common complication after re-perfusion, a treatment of an acute
myocardial infarction (AMI). MVO is typically diagnosed using Gadolinium based
CMR such as first-pass CMR and late Gadolinium contrast enhancement (LGE).
However, these techniques require the use of Gadolinium contrast agents that are
contraindicated in patients with kidney dysfunction. In chapter 5, I demonstrated
that myocardial ASL is able to detect MVO at rest, consistent with MVO seen in
LGE and perfusion deficit seen in first-pass CMR. This work also demonstrated
that ASL is able to measure regional stress response. Ability to assess regional
stress response would potentially be useful for assessment of microvascular dys-
function since studies have shown that microvascular function is impaired not only
in the infarct but also in the remote territories.
102
Myocardial blood flow (MBF) and myocardial blood volume (MBV) are two
important physiological indicators of coronary and microvascular function. Chap-
ter 6 demonstrated feasibility of quantitative assessment of MBF usingT
1
mapping
in healthy subjects and the potential application of T
1
mapping in assessment of
MBV change in heart transplant recipients. T
1
mapping may allow efficient MBF
mapping in two breath-holds instead of 7 breath-holds as in current myocardial
ASL methods, permitting 3-4 slices coverage in 3-4 minutes of adenosine vasodi-
lation stress. The results also suggest that T
1
may be able to assess myocardial
fibrosis and microvascular function in heart transplant recipients.
Pseudo-continuous ASL (PCASL) is the preferred labeling scheme in the brain,
however, the optimal labeling scheme in the heart may be different due to chal-
lenges associated with the heart including cardiac and respiratory motion, com-
plex blood flow path, complex coronary geometry, pulsatile blood flow, and B
1
and B
0
field inhomogeneity. These challenges were considered in the proposal of
physiologically synchronized multi-module pulsed ASL (SymPASL) in chapter 7.
SymPASL utilizes several labeling pulses that are synchronized to pulsatile blood
flow of coronary arteries, while the image acquisition is synchronized to the car-
diac motion. SymPASL was validated using simulations and in vivo experiments
in human kidneys as a proof-of-concept. Consistent with simulation, experimen-
tal results suggested that SymPASL could provide superior ASL signal and SNR
efficiency compared to pulsed ASL with 1 heartbeat post-labeling-delay and com-
parableASLsignalandSNRefficiencytoPCASLwith1.5-secondlabelingduration
and 1.0-second post-labeling-delay.
This dissertation informs the following future work:
• Scan efficiency of flow-sensitive alternating inversion recovery (FAIR) would
be vastly improved with saturation instead of inversion. I would recommend
103
investigation of double-gating FAIR with saturation label. Saturation label
is independent of spin history therefore ASL image can be acquired in every
2 heartbeats. That would significantly increase the number of averages dur-
ing three minutes scan time. The increase in scan efficiency could be used
to increase spatial coverage to 3-4 slices in 3-4 minutes of peak adenosine
vasodilation stress.
• Myocardial T
1
mapping with appropriate flow-sensitized preparation would
allow quantitative assessment of MBF, MBV change, and MBF change with
vasodilation. I would recommend investigation of T
1
mapping for quantita-
tive assessment of MBF, MBV change, and MBF change each of which only
requires two T
1
maps. Myocardial T
1
mapping for a single-slice takes one
breath-hold only, this would enable acquisition of 3-5 slices of MBF or MBV
change or MBF change in 3-4 minutes of peak adenosine vasodilation stress.
• SymPASL can potentially provide higher ASL signal and SNR efficiency and
less sensitive to arterial transit time (ATT) compared to traditional PASL.
It also has similar ASL signal and SNR efficiency compared to PCASL with
potentially lower SAR. SymPASL is expected to be compatible to volumetric
imaging therefore I recommend to explore potential application of SymPASL
in the heart and in the kidney in combination with volumetric imaging such
as sequential multi-slice, simultaneous multi-slice, or 3D imaging.
104
Bibliography
[1] D. Mozaffarian, E. J. Benjamin, A. S. Go, D. K. Arnett, M. J. Blaha,
M. Cushman, S. R. Das, S. de Ferranti, J.-P. Després, H. J. Fullerton, et al.,
“Heart disease and stroke statistics - 2016 update,” Circulation, vol. 133,
no. 4, pp. e38–e360, 2016.
[2] NIH, “What Is Coronary Heart Disease?.” https://www.nhlbi.nih.gov/
health/health-topics/topics/cad, 2017. [Online; accessed 23-Mar-2017].
[3] H. A. Hadi, C. S. Carr, and J. A. Suwaidi, “Endothelial dysfunction: car-
diovascular risk factors, therapy, and outcome,” Vascular health and risk
management, vol. 1, no. 3, p. 183, 2005.
[4] F. W. Sellke, E. R. Mohler III, A. Lerman, R. J. Widmer,
and F. Crea, “Coronary artery endothelial dysfunction: Basic
concepts - uptodate.” https://www.uptodate.com/contents/
coronary-artery-endothelial-dysfunction-basic-concepts, 2016.
[Online; accessed 25-Feb-2017].
[5] L. J. Shaw, R. Bugiardini, and C. N. B. Merz, “Women and ischemic heart
disease: evolvingknowledge,” Journal of the American College of Cardiology,
vol. 54, no. 17, pp. 1561–1575, 2009.
[6] P. S. Douglas, “Screening for coronary heart disease - uptodate.” https://
www.uptodate.com/contents/screening-for-coronary-heart-disease,
2015. [Online; accessed 25-Feb-2017].
[7] M. Salerno and G. A. Beller, “Noninvasive assessment of myocardial perfu-
sion,” Circulation: Cardiovascular Imaging, vol. 2, no. 5, pp. 412–424, 2009.
[8] L. J. Shaw, F. G. Hage, D. S. Berman, R. Hachamovitch, and A. Iskan-
drian, “Prognosis in the era of comparative effectiveness research: Where is
nuclear cardiology now and where should it be?,” Journal of nuclear cardi-
ology, vol. 19, no. 5, pp. 1026–1043, 2012.
[9] J. P. Greenwood, N. Maredia, J. F. Younger, J. M. Brown, J. Nixon, C. C.
Everett, P. Bijsterveld, J. P. Ridgway, A. Radjenovic, C. J. Dickinson, et al.,
105
“Cardiovascular magnetic resonance and single-photon emission computed
tomography for diagnosis of coronary heart disease (ce-marc): a prospective
trial,” The Lancet, vol. 379, no. 9814, pp. 453–460, 2012.
[10] M. D. Cerqueira, K. C. Allman, E. P. Ficaro, C. L. Hansen, K. J. Nichols,
R. C. Thompson, W. A. Van Decker, and M. Yakovlevitch, “Recommen-
dations for reducing radiation exposure in myocardial perfusion imaging,”
Journal of nuclear cardiology, vol. 17, no. 4, pp. 709–718, 2010.
[11] R. Fazel, V. Dilsizian, A. J. Einstein, E. P. Ficaro, M. Henzlova, and L. J.
Shaw, “Strategies for defining an optimal risk-benefit ratio for stress myocar-
dial perfusion spect,” Journal of Nuclear Cardiology, vol. 18, no. 3, pp. 385–
392, 2011.
[12] E. C. Lin, “Radiation risk from medical imaging,” Mayo Clinic Proceedings,
vol. 85, no. 12, pp. 1142–1146, 2010.
[13] P. H. Kuo, E. Kanal, A. K. Abu-Alfa, and S. E. Cowper, “Gadolinium-based
mr contrast agents and nephrogenic systemic fibrosis 1,” Radiology, vol. 242,
no. 3, pp. 647–649, 2007.
[14] R.J.McDonald, J.S.McDonald, D.F.Kallmes, M.E.Jentoft, D.L.Murray,
K. R. Thielen, E. E. Williamson, and L. J. Eckel, “Intracranial gadolinium
deposition after contrast-enhanced mr imaging,” Radiology, vol. 275, no. 3,
pp. 772–782, 2015.
[15] T. Kanda, T. Fukusato, M. Matsuda, K. Toyoda, H. Oba, J. Kotoku,
T. Haruyama, K. Kitajima, and S. Furui, “Gadolinium-based contrast agent
accumulates in the brain even in subjects without severe renal dysfunction:
evaluation of autopsy brain specimens with inductively coupled plasma mass
spectroscopy,” Radiology, vol. 276, no. 1, pp. 228–232, 2015.
[16] A. A. Malayeri, K. M. Brooks, L. H. Bryant, R. Evers, P. Kumar, D. S.
Reich, and D. A. Bluemke, “National institutes of health perspective on
reports of gadolinium deposition in the brain,” J Am Coll Radiol, vol. 13,
no. 3, pp. 237–241, 2016.
[17] T. Kanda, Y. Nakai, H. Oba, K. Toyoda, K. Kitajima, and S. Furui,
“Gadolinium deposition in the brain,” Magnetic Resonance Imaging, vol. 34,
no. 10, pp. 1346–1350, 2016.
[18] U. S. R. D. System, “2016 usrds annual data report: Epidemiology of
kidney disease in the united states..” https://www.usrds.org/2016/view/
Default.aspx, 2016. [Online; accessed 23-Mar-2017].
106
[19] J. A. Detre, J. S. Leigh, D. S. Williams, and A. P. Koretsky, “Perfusion
imaging,” Magnetic Resonance in Medicine, vol. 23, no. 1, pp. 37–45, 1992.
[20] D. S. Williams, J. A. Detre, J. S. Leigh, and A. P. Koretsky, “Magnetic
resonance imaging of perfusion using spin inversion of arterial water,” Pro-
ceedings of the National Academy of Sciences, vol. 89, no. 1, pp. 212–216,
1992.
[21] J.-C. Ferré, E. Bannier, H. Raoult, G. Mineur, B. Carsin-Nicol, and J.-Y.
Gauvrit, “Arterial spin labeling (asl) perfusion: techniques and clinical use,”
Diagnostic and interventional imaging, vol. 94, no. 12, pp. 1211–1223, 2013.
[22] F. Calamante, D. L. Thomas, G. S. Pell, J. Wiersma, and R. Turner, “Mea-
suring cerebral blood flow using magnetic resonance imaging techniques,”
Journal of cerebral blood flow & metabolism, vol. 19, no. 7, pp. 701–735,
1999.
[23] J. A. Detre, H. Rao, D. J. Wang, Y. F. Chen, and Z. Wang, “Applications
of arterial spin labeled mri in the brain,” Journal of Magnetic Resonance
Imaging, vol. 35, no. 5, pp. 1026–1037, 2012.
[24] N. A. Telischak, J. A. Detre, and G. Zaharchuk, “Arterial spin labeling mri:
clinical applications in the brain,” Journal of Magnetic Resonance Imaging,
vol. 41, no. 5, pp. 1165–1180, 2015.
[25] J. M. Watts, C. T. Whitlow, and J. A. Maldjian, “Clinical applications of
arterial spin labeling,” NMR in Biomedicine, vol. 26, no. 8, pp. 892–900,
2013.
[26] J. A. Detre, J. Wang, Z. Wang, and H. Rao, “Arterial spin-labeled perfusion
mri in basic and clinical neuroscience,” Current opinion in neurology, vol. 22,
no. 4, pp. 348–355, 2009.
[27] B. Griffith and R. Jain, “Perfusion imaging in neuro-oncology: basic tech-
niques and clinical applications,” Magnetic Resonance Imaging Clinics of
North America, vol. 24, no. 4, pp. 765–779, 2016.
[28] E. C. Wong, “Vessel-encoded arterial spin-labeling using pseudocontinuous
tagging,” Magnetic Resonance in Medicine, vol. 58, no. 6, pp. 1086–1091,
2007.
[29] D. C. Alsop, J. A. Detre, X. Golay, M. Günther, J. Hendrikse, L. Hernandez-
Garcia, H. Lu, B. J. MacIntosh, L. M. Parkes, M. Smits, et al., “Rec-
ommended implementation of arterial spin-labeled perfusion mri for clini-
cal applications: A consensus of the ismrm perfusion study group and the
107
european consortium for asl in dementia,” Magnetic resonance in medicine,
vol. 73, no. 1, pp. 102–116, 2015.
[30] S.-G. Kim, “Quantification of relative cerebral blood flow change by flow-
sensitive alternating inversion recovery (fair) technique: Application to func-
tional mapping,” Magnetic Resonance in Medicine, vol. 34, no. 3, pp. 293–
301, 1995.
[31] K. K. Kwong, D. A. Chesler, R. M. Weisskoff, K. M. Donahue, T. L. Davis,
L. Ostergaard, T. A. Campbell, and B. R. Rosen, “Mr perfusion studies with
t1-weighted echo planar imaging,” Magnetic Resonance in Medicine, vol. 34,
no. 6, pp. 878–887, 1995.
[32] Z. Zun, E. C. Wong, and K. S. Nayak, “Assessment of myocardial blood
flow (mbf) in humans using arterial spin labeling (asl): feasibility and noise
analysis,” Magnetic resonance in medicine, vol. 62, no. 4, pp. 975–983, 2009.
[33] Z. Zun, P. Varadarajan, R. G. Pai, E. C. Wong, and K. S. Nayak, “Arterial
spin labeled cmr detects clinically relevant increase in myocardial blood flow
with vasodilation,” JACC: Cardiovascular Imaging, vol. 4, no. 12, pp. 1253–
1261, 2011.
[34] F. H. Epstein and C. H. Meyer, “Myocardial perfusion using arterial spin
labeling cmr: promise and challenges,” JACC: Cardiovascular Imaging,
vol. 4, no. 12, pp. 1262–1264, 2011.
[35] P.A.Bottomley, T.H.Foster, R.E.Argersinger, andL.M.Pfeifer, “Areview
of normal tissue hydrogen nmr relaxation times and relaxation mechanisms
from 1–100 mhz: dependence on tissue type, nmr frequency, temperature,
species, excision, and age,” Medical physics, vol. 11, no. 4, pp. 425–448, 1984.
[36] P. A. Bottomley, C. Hardy, R. Argersinger, and G. Allen-Moore, “A review
of 1h nuclear magnetic resonance relaxation in pathology: Are t1 and t2
diagnostic?,” Medical physics, vol. 14, no. 1, pp. 1–37, 1987.
[37] G. J. Stanisz, E. E. Odrobina, J. Pun, M. Escaravage, S. J. Graham, M. J.
Bronskill, andR.M.Henkelman, “T1, t2relaxationandmagnetizationtrans-
fer in tissue at 3t,” Magnetic resonance in medicine, vol. 54, no. 3, pp. 507–
512, 2005.
[38] H. Carr, “Steady-state free precession in nuclear magnetic resonance,” Phys-
ical Review, vol. 112, no. 5, p. 1693, 1958.
[39] O.BieriandK.Scheffler, “Fundamentalsofbalancedsteadystatefreepreces-
sion mri,” Journal of Magnetic Resonance Imaging, vol. 38, no. 1, pp. 2–11,
2013.
108
[40] B. Hargreaves, “Rapid gradient-echo imaging,” Journal of Magnetic Reso-
nance Imaging, vol. 36, no. 6, pp. 1300–1313, 2012.
[41] K. Scheffler, “On the transient phase of balanced ssfp sequences,” Magnetic
resonance in medicine, vol. 49, no. 4, pp. 781–783, 2003.
[42] K. Scheffler and S. Lehnhardt, “Principles and applications of balanced ssfp
techniques,” European radiology, vol. 13, no. 11, pp. 2409–2418, 2003.
[43] B. A. Hargreaves, S. S. Vasanawala, J. M. Pauly, and D. G. Nishimura,
“Characterization and reduction of the transient response in steady-state mr
imaging,” Magnetic resonance in medicine, vol. 46, no. 1, pp. 149–158, 2001.
[44] P. Le Roux, “Simplified model and stabilization of ssfp sequences,” Journal
of Magnetic Resonance, vol. 163, no. 1, pp. 23–37, 2003.
[45] V. S. Deshpande, Y.-C. Chung, Q. Zhang, S. M. Shea, and D. Li, “Reduction
of transient signal oscillations in true-fisp using a linear flip angle series
magnetization preparation,” Magnetic resonance in medicine, vol. 49, no. 1,
pp. 151–157, 2003.
[46] Y. Zur, S. Stokar, and P. Bendel, “An analysis of fast imaging sequences
with steady-state transverse magnetization refocusing,” Magnetic resonance
in medicine, vol. 6, no. 2, pp. 175–193, 1988.
[47] R. Freeman and H. Hill, “Phase and intensity anomalies in fourier transform
nmr,” Journal of Magnetic Resonance (1969), vol. 4, no. 3, pp. 366–383,
1971.
[48] M. Markl, M. Alley, C. Elkins, and N. Pelc, “Flow effects in balanced steady
statefreeprecessionimaging,” Magnetic resonance in medicine,vol.50,no.5,
pp. 892–903, 2003.
[49] M. Markl and N. J. Pelc, “On flow effects in balanced steady-state free
precession imaging: Pictorial description, parameter dependence, and clin-
ical implications,” Journal of Magnetic Resonance Imaging, vol. 20, no. 4,
pp. 697–705, 2004.
[50] M. Schär, S. Kozerke, S. E. Fischer, and P. Boesiger, “Cardiac ssfp imaging
at 3 tesla,” Magnetic resonance in medicine, vol. 51, no. 4, pp. 799–806,
2004.
[51] J. Wansapura, R. Fleck, E. Crotty, and W. Gottliebson, “Frequency scouting
for cardiac imaging with ssfp at 3 tesla,” Pediatric radiology, vol. 36, no. 10,
pp. 1082–1085, 2006.
109
[52] J. Lee, M. Lustig, D.-h. Kim, and J. M. Pauly, “Improved shim method
based on the minimization of the maximum off-resonance frequency for bal-
anced steady-state free precession (bssfp),” Magnetic resonance in medicine,
vol. 61, no. 6, pp. 1500–1506, 2009.
[53] K. S. Nayak, H.-L. Lee, B. A. Hargreaves, and B. S. Hu, “Wideband
ssfp: alternating repetition time balanced steady state free precession with
increased band spacing,” Magnetic resonance in medicine, vol. 58, no. 5,
pp. 931–938, 2007.
[54] S. S. Vasanawala, J. M. Pauly, and D. G. Nishimura, “Linear combination
steady-state free precession mri,” Magnetic resonance in medicine, vol. 43,
no. 1, pp. 82–90, 2000.
[55] N. K. Bangerter, B. A. Hargreaves, S. S. Vasanawala, J. M. Pauly, G. E.
Gold, andD.G.Nishimura, “Analysisofmultiple-acquisitionssfp,” Magnetic
resonance in medicine, vol. 51, no. 5, pp. 1038–1047, 2004.
[56] Q.-S. Xiang and M. N. Hoff, “Banding artifact removal for bssfp imaging
with an elliptical signal model,” Magnetic resonance in medicine, vol. 71,
no. 3, pp. 927–933, 2014.
[57] T. A. Gallagher, A. J. Nemeth, and L. Hacein-Bey, “An introduction to the
fourier transform: relationship to mri,” American journal of roentgenology,
vol. 190, no. 5, pp. 1396–1405, 2008.
[58] A. Deshmane, V. Gulani, M. A. Griswold, and N. Seiberlich, “Parallel mr
imaging,” Journal of Magnetic Resonance Imaging, vol. 36, no. 1, pp. 55–72,
2012.
[59] K. P. Pruessmann, M. Weiger, M. B. Scheidegger, P. Boesiger, et al., “Sense:
sensitivity encoding for fast mri,” Magnetic resonance in medicine, vol. 42,
no. 5, pp. 952–962, 1999.
[60] M. A. Griswold, P. M. Jakob, R. M. Heidemann, M. Nittka, V. Jellus,
J.Wang, B.Kiefer, andA.Haase, “Generalizedautocalibratingpartiallypar-
allel acquisitions (grappa),” Magnetic resonance in medicine, vol. 47, no. 6,
pp. 1202–1210, 2002.
[61] F. A. Breuer, S. A. Kannengiesser, M. Blaimer, N. Seiberlich, P. M. Jakob,
and M. A. Griswold, “General formulation for quantitative g-factor calcula-
tion in grappa reconstructions,” Magnetic Resonance in Medicine, vol. 62,
no. 3, pp. 739–746, 2009.
110
[62] Wikipedia, “Coronary circulation — Wikipedia, The Free Encyclopedia.”
https://upload.wikimedia.org/wikipedia/commons/9/93/Blausen_
0256_CoronaryArteries_02.png, 2017. [Online; accessed 23-Mar-2017].
[63] R. L. Ehman, M. McNamara, M. Pallack, H. Hricak, and C. Higgins, “Mag-
netic resonance imaging with respiratory gating: techniques and advan-
tages,” American journal of Roentgenology, vol. 143, no. 6, pp. 1175–1182,
1984.
[64] M. B. Hofman, C. B. Paschal, D. Li, E. M. Haacke, A. C. van Rossum, and
M. Sprenger, “Mri of coronary arteries: 2d breath-hold vs 3d respiratory-
gated acquisition,” Journal of computer assisted tomography, vol. 19, no. 1,
pp. 56–62, 1995.
[65] Q. Yuan, L. Axel, E. H. Hernandez, L. Dougherty, J. J. Pilla, C. H. Scott,
V. A. Ferrari, and A. S. Blom, “Cardiac-respiratory gating method for mag-
netic resonance imaging of the heart,” Magnetic resonance in medicine,
vol. 43, no. 2, pp. 314–318, 2000.
[66] Y. Wang, P. J. Rossman, R. C. Grimm, S. J. Riederer, and R. L. Ehman,
“Navigator-echo-based real-time respiratory gating and triggering for reduc-
tion of respiration effects in three-dimensional coronary mr angiography.,”
Radiology, vol. 198, no. 1, pp. 55–60, 1996.
[67] D. Bailes, D. Gilderdale, G. Bydder, A. Collins, and D. Firmin, “Respiratory
ordered phase encoding (rope): a method for reducing respiratory motion
artefacts in mr imaging.,” Journal of computer assisted tomography, vol. 9,
no. 4, pp. 835–838, 1985.
[68] G. Glover and J. Pauly, “Projection reconstruction techniques for reduction
of motion effects in mri,” Magnetic resonance in medicine, vol. 28, no. 2,
pp. 275–289, 1992.
[69] A. C. Larson, P. Kellman, A. Arai, G. A. Hirsch, E. McVeigh, D. Li, and
O. P. Simonetti, “Preliminary investigation of respiratory self-gating for free-
breathing segmented cine mri,” Magnetic resonance in medicine, vol. 53,
no. 1, pp. 159–168, 2005.
[70] L. Feng, L. Axel, H. Chandarana, K. T. Block, D. K. Sodickson, and
R. Otazo, “Xd-grasp: Golden-angle radial mri with reconstruction of extra
motion-state dimensions using compressed sensing,” Magnetic resonance in
medicine, vol. 75, no. 2, pp. 775–788, 2016.
111
[71] J. Allen and A. Murray, “Age-related changes in peripheral pulse timing
characteristics at the ears, fingers and toes,” Journal of human hypertension,
vol. 16, no. 10, pp. 711–717, 2002.
[72] K.R.Johnson, S.J.Patel, A.Whigham, A.Hakim, R.I.Pettigrew, andJ.N.
Oshinski, “Three-dimensional, time-resolved motion of the coronary arter-
ies,” Journal of Cardiovascular Magnetic Resonance, vol. 6, no. 3, pp. 663–
673, 2004.
[73] M. Vembar, M. Garcia, D. Heuscher, R. Haberl, D. Matthews, G. Böhme,
and N. Greenberg, “A dynamic approach to identifying desired physiological
phasesforcardiacimagingusingmultislicespiralct,” Medical physics, vol.30,
no. 7, pp. 1683–1693, 2003.
[74] A. M. Gharib, D. A. Herzka, A. O. Ustun, M. Y. Desai, J. Locklin, R. I.
Pettigrew, and M. Stuber, “Coronary mr angiography at 3t during diastole
andsystole,” Journal of Magnetic Resonance Imaging, vol.26, no.4, pp.921–
926, 2007.
[75] Wikipedia, “Wiggers diagram — Wikipedia, The Free Encyclopedia.”
https://upload.wikimedia.org/wikipedia/commons/f/f4/Wiggers_
Diagram.svg, 2017. [Online; accessed 23-Mar-2017].
[76] A. C. Larson, R. D. White, G. Laub, E. R. McVeigh, D. Li, and O. P.
Simonetti, “Self-gated cardiac cine mri,” Magnetic Resonance in Medicine,
vol. 51, no. 1, pp. 93–102, 2004.
[77] M. Buehrer, J. Curcic, P. Boesiger, and S. Kozerke, “Prospective self-gating
for simultaneous compensation of cardiac and respiratory motion,” Magnetic
resonance in medicine, vol. 60, no. 3, pp. 683–690, 2008.
[78] R. Rokey, R. E. Wendt, and D. L. Johnston, “Monitoring of acutely iii
patients during nuclear magnetic resonance imaging: Use of a time-varying
filter electrocardiographic gating device to reduce gradient artifacts,” Mag-
netic resonance in medicine, vol. 6, no. 2, pp. 240–245, 1988.
[79] M. Polson, A. Barker, and S. Gardiner, “The effect of rapid rise-time mag-
netic fields on the ecg of the rat,” Clinical Physics and Physiological Mea-
surement, vol. 3, no. 3, p. 231, 1982.
[80] A. N. Shetty, “Suppression of radiofrequency interference in cardiac gated
mri: a simple design,” Magnetic resonance in medicine, vol. 8, no. 1, pp. 84–
88, 1988.
112
[81] K. K. Wong, Z. Sun, J. Tu, S. G. Worthley, J. Mazumdar, and D. Abbott,
“Medical image diagnostics based on computer-aided flow analysis using
magnetic resonance images,” Computerized Medical Imaging and Graphics,
vol. 36, no. 7, pp. 527–541, 2012.
[82] A. K. Attili, A. Schuster, E. Nagel, J. H. Reiber, and R. J. van der Geest,
“Quantification in cardiac mri: advances in image acquisition and process-
ing,” The international journal of cardiovascular imaging, vol. 26, no. 1,
pp. 27–40, 2010.
[83] J. P. Earls, V. B. Ho, T. K. Foo, E. Castillo, and S. D. Flamm, “Cardiac mri:
recent progress and continued challenges,” Journal of magnetic resonance
imaging, vol. 16, no. 2, pp. 111–127, 2002.
[84] M. P. Marra, J. A. Lima, and S. Iliceto, “Mri in acute myocardial infarction,”
European heart journal, p. ehq409, 2010.
[85] R. Y. Kwong and H. Korlakunta, “Diagnostic and prognostic value of car-
diac magnetic resonance imaging in assessing myocardial viability,” Topics
in Magnetic Resonance Imaging, vol. 19, no. 1, pp. 15–24, 2008.
[86] A. Wagner, H. Mahrholdt, T. A. Holly, M. D. Elliott, M. Regenfus,
M.Parker, F.J.Klocke, R.O.Bonow, R.J.Kim, andR.M.Judd, “Contrast-
enhanced mri and routine single photon emission computed tomogra-
phy (spect) perfusion imaging for detection of subendocardial myocardial
infarcts: an imaging study,” The Lancet, vol. 361, no. 9355, pp. 374–379,
2003.
[87] A.R. Patel, F. H.Epstein, andC. M.Kramer, “Evaluationof themicrocircu-
lation: advances in cardiac magnetic resonance perfusion imaging,” Journal
of nuclear cardiology, vol. 15, no. 5, pp. 698–708, 2008.
[88] P. Kellman and A. E. Arai, “Imaging sequences for first pass perfusionÑa
review,” Journal of Cardiovascular Magnetic Resonance, vol. 9, no. 3,
pp. 525–537, 2007.
[89] M. Jerosch-Herold, R. T. Seethamraju, C. M. Swingen, N. M. Wilke, and
A. E. Stillman, “Analysis of myocardial perfusion mri,” Journal of Magnetic
Resonance Imaging, vol. 19, no. 6, pp. 758–770, 2004.
[90] P. Kellman and M. S. Hansen, “T1-mapping in the heart: accuracy and
precision,” Journal of cardiovascular magnetic resonance, vol. 16, no. 1, p. 2,
2014.
113
[91] V. O. Puntmann, E. Peker, Y. Chandrashekhar, and E. Nagel, “T1 mapping
in characterizing myocardial disease,” Circulation Research, vol. 119, no. 2,
pp. 277–299, 2016.
[92] V. O. Puntmann, T. Voigt, Z. Chen, M. Mayr, R. Karim, K. Rhode, A. Pas-
tor, G. Carr-White, R. Razavi, T. Schaeffter, et al., “Native t1 mapping in
differentiation of normal myocardium from diffuse disease in hypertrophic
and dilated cardiomyopathy,” JACC: Cardiovascular imaging, vol. 6, no. 4,
pp. 475–484, 2013.
[93] J. A. Lima, “The promise of myocardial fibrosis assessment by t1 mapping,”
2013.
[94] A. Liu, R. S. Wijesurendra, J. M. Francis, M. D. Robson, S. Neubauer, S. K.
Piechnik, and V. M. Ferreira, “Adenosine stress and rest t1 mapping can
differentiate between ischemic, infarcted, remote, and normal myocardium
without the need for gadolinium contrast agents,” JACC: Cardiovascular
Imaging, vol. 9, no. 1, pp. 27–36, 2016.
[95] E. Nagel and V. O. Puntmann, “Is myocardial native t1 the one answer for
all?,” JACC: Cardiovascular Imaging, vol. 9, no. 1, pp. 37–39, 2016.
[96] D. C. Look and D. R. Locker, “Time saving in measurement of nmr and epr
relaxation times,” Review of Scientific Instruments, vol. 41, no. 2, pp. 250–
251, 1970.
[97] R. Kaptein, K. Dijkstra, and C. Tarr, “A single-scan fourier transform
method for measuring spin-lattice relaxation times,” Journal of Magnetic
Resonance (1969), vol. 24, no. 2, pp. 295–300, 1976.
[98] D. R. Messroghli, A. Radjenovic, S. Kozerke, D. M. Higgins, M. U. Sivanan-
than, and J. P. Ridgway, “Modified look-locker inversion recovery (molli) for
high-resolution t1 mapping of the heart,” Magnetic resonance in medicine,
vol. 52, no. 1, pp. 141–146, 2004.
[99] S. K. Piechnik, V. M. Ferreira, E. Dall’Armellina, L. E. Cochlin, A. Greiser,
S. Neubauer, and M. D. Robson, “Shortened modified look-locker inversion
recovery (shmolli) for clinical myocardial t1-mapping at 1.5 and 3 t within
a 9 heartbeat breathhold,” Journal of cardiovascular magnetic resonance,
vol. 12, no. 1, p. 69, 2010.
[100] K. Chow, J. A. Flewitt, J. D. Green, J. J. Pagano, M. G. Friedrich, and
R. B. Thompson, “Saturation recovery single-shot acquisition (sasha) for
myocardial t1 mapping,” Magnetic resonance in medicine, vol. 71, no. 6,
pp. 2082–2095, 2014.
114
[101] S. Weingärtner, M. Akçakaya, T. Basha, K. V. Kissinger, B. Goddu, S. Berg,
W. J. Manning, and R. Nezafat, “Combined saturation/inversion recovery
sequences for improved evaluation of scar and diffuse fibrosis in patients
with arrhythmia or heart rate variability,” Magnetic resonance in medicine,
vol. 71, no. 3, pp. 1024–1034, 2014.
[102] G. S. Slavin and J. A. Stainsby, “True t1 mapping with smart1map (satura-
tion method using adaptive recovery times for cardiac t1 mapping): a com-
parison with molli,” Journal of Cardiovascular Magnetic Resonance, vol. 15,
no. Suppl 1, p. P3, 2013.
[103] S.Roujol, S.Weingärtner, M.Foppa, K.Chow, K.Kawaji, L.H.Ngo, P.Kell-
man, W.J.Manning, R.B.Thompson, andR.Nezafat, “Accuracy, precision,
andreproducibilityoffourt1mappingsequences: ahead-to-headcomparison
of molli, shmolli, sasha, and sapphire,” Radiology, vol. 272, no. 3, pp. 683–
689, 2014.
[104] M. D. Cerqueira, N. J. Weissman, V. Dilsizian, A. K. Jacobs, S. Kaul, W. K.
Laskey, D. J. Pennell, J. A. Rumberger, T. Ryan, and M. S. Verani, “Stan-
dardized myocardial segmentation and nomenclature for tomographic imag-
ing of the heart - a statement for healthcare professionals from the cardiac
imaging committee of the council on clinical cardiology of the american heart
association,” Circulation, vol. 105, pp. 539–42, 2002.
[105] “Medviso.” http://medviso.com, 2017. [Online; accessed 23-Mar-2017].
[106] “Cmrsegtools.” https://www.creatis.insa-lyon.fr/CMRSegTools/,
2017. [Online; accessed 23-Mar-2017].
[107] “Medis.” https://www.medis.nl, 2017. [Online; accessed 23-Mar-2017].
[108] “Circle cardiovascular imaging.” https://www.circlecvi.com, 2017.
[Online; accessed 23-Mar-2017].
[109] Wikipedia, “Thoracic cavity — Wikipedia, The Free Encyclope-
dia.” https://commons.wikimedia.org/wiki/File:Blausen_0467_
HeartLocation.png, 2017. [Online; accessed 23-Mar-2017].
[110] Wikipedia, “Heart short axis view — Wikipedia, The Free Encyclopedia.”
https://commons.wikimedia.org/wiki/File:Heart_normal_short_
axis_section.jpg, 2017. [Online; accessed 23-Mar-2017].
[111] J.-U. Voigt, G. Pedrizzetti, P. Lysyansky, T. H. Marwick, H. Houle, R. Bau-
mann, S. Pedri, Y. Ito, Y. Abe, S. Metz, et al., “Definitions for a common
standard for 2d speckle tracking echocardiography: consensus document of
115
the eacvi/ase/industry task force to standardize deformation imaging,” Eur
Heart J Cardiovasc Imaging, p. jeu184, 2014.
[112] H. Lu, C. Clingman, X. Golay, and P. van Zijl, “Determining the longitudinal
relaxation time (t1) of blood at 3.0 tesla,” Magnetic resonance in medicine,
vol. 52, no. 3, pp. 679–682, 2004.
[113] X. Zhang, E. Petersen, E. Ghariq, J. Vis, A. Webb, W. Teeuwisse, J. Hen-
drikse, and M. Osch, “In vivo blood t1 measurements at 1.5 t, 3 t, and 7 t,”
Magnetic resonance in medicine, vol. 70, no. 4, pp. 1082–1086, 2013.
[114] J. Hendrikse, E. T. Petersen, and X. Golay, “Vascular disorders: insights
from arterial spin labeling,” Neuroimaging clinics of North America, vol. 22,
no. 2, pp. 259–269, 2012.
[115] S. Haller, G. Zaharchuk, D. L. Thomas, K.-O. Lovblad, F. Barkhof, and
X. Golay, “Arterial spin labeling perfusion of the brain: emerging clinical
applications,” Radiology, vol. 281, no. 2, pp. 337–356, 2016.
[116] W.Dai, D.Garcia, C.deBazelaire, andD.C.Alsop, “Continuousflow-driven
inversion for arterial spin labeling using pulsed radio frequency and gradient
fields,” Magnetic resonance in medicine, vol. 60, no. 6, pp. 1488–1497, 2008.
[117] E. C. Wong, R. B. Buxton, and L. R. Frank, “Quantitative imaging of per-
fusion using a single subtraction (quipss and quipss ii),” Magnetic resonance
in medicine, vol. 39, no. 5, pp. 702–708, 1998.
[118] E. C. Wong, “Quantifying cbf with pulsed asl: technical and pulse sequence
factors,” Journal of Magnetic Resonance Imaging, vol.22, no.6, pp.727–731,
2005.
[119] E. C. Wong, R. B. Buxton, and L. R. Frank, “A theoretical and experimental
comparison of continuous and pulsed arterial spin labeling techniques for
quantitative perfusion imaging,” Magnetic Resonance in Medicine, vol. 40,
no. 3, pp. 348–355, 1998.
[120] R. R. Edelman, B. Siewert, D. Darby, V. Thangaraj, A. Nobre, M. Mesulam,
and S. Warach, “Qualitative mapping of cerebral blood flow and functional
localization with echo-planar mr imaging and signal targeting with alternat-
ing radio frequency.,” Radiology, vol. 192, no. 2, pp. 513–520, 1994.
[121] R. R. Edelman and Q. Chen, “Epistar mri: multislice mapping of cerebral
blood flow,” Magnetic resonance in medicine, vol. 40, no. 6, pp. 800–805,
1998.
116
[122] E. C. Wong, M. Cronin, W.-C. Wu, B. Inglis, L. R. Frank, and T. T. Liu,
“Velocity-selective arterial spin labeling,” Magnetic Resonance in Medicine,
vol. 55, no. 6, pp. 1334–1341, 2006.
[123] W. T. Dixon, L. N. Du, D. D. Faul, M. Gado, and S. Rossnick, “Projection
angiograms of blood labeled by adiabatic fast passage,” Magnetic resonance
in medicine, vol. 3, no. 3, pp. 454–462, 1986.
[124] W.-C. Wu, M. Fernández-Seara, J. A. Detre, F. W. Wehrli, and J. Wang,
“A theoretical and experimental investigation of the tagging efficiency of
pseudocontinuous arterial spin labeling,” Magnetic Resonance in Medicine,
vol. 58, no. 5, pp. 1020–1027, 2007.
[125] H. Jahanian, D. C. Noll, and L. Hernandez-Garcia, “B0 field inhomogeneity
considerations in pseudo-continuous arterial spin labeling (pcasl): effects on
tagging efficiency and correction strategy,” NMR in Biomedicine, vol. 24,
no. 10, pp. 1202–1209, 2011.
[126] S. Aslan, F. Xu, P. L. Wang, J. Uh, U. S. Yezhuvath, M. van Osch, and
H. Lu, “Estimation of labeling efficiency in pseudocontinuous arterial spin
labeling,” Magnetic resonance in medicine, vol. 63, no. 3, pp. 765–771, 2010.
[127] Y. Chen, D. J. Wang, and J. A. Detre, “Test–retest reliability of arterial spin
labeling with common labeling strategies,” Journal of Magnetic Resonance
Imaging, vol. 33, no. 4, pp. 940–949, 2011.
[128] Y. Wang, K. Moin, O. Akinboboye, and N. Reichek, “Myocardial first pass
perfusion: Steady-state free precession versus spoiled gradient echo and seg-
mentedechoplanarimaging,” Magnetic resonance in medicine, vol.54, no.5,
pp. 1123–1129, 2005.
[129] R. B. Buxton, L. R. Frank, E. C. Wong, B. Siewert, S. Warach, and R. R.
Edelman, “A general kinetic model for quantitative perfusion imaging with
arterial spin labeling,” Magnetic resonance in medicine, vol. 40, no. 3,
pp. 383–396, 1998.
[130] D. C. Alsop and J. Detre, “Reduced transit-time sensitivity in noninvasive
magnetic resonance imaging of human cerebral blood flow,” Journal of Cere-
bral Blood Flow & Metabolism, vol. 16, no. 6, pp. 1236–1249, 1996.
[131] D. J. Wang, X. Bi, B. B. Avants, T. Meng, S. Zuehlsdorff, and J. A. Detre,
“Estimation of perfusion and arterial transit time in myocardium using free-
breathing myocardial arterial spin labeling with navigator-echo,” Magnetic
resonance in medicine, vol. 64, no. 5, pp. 1289–1295, 2010.
117
[132] V. Belle, E. Kahler, C. Waller, E. Rommel, S. Voll, K.-H. Hiller, W. R.
Bauer, and A. Haase, “In vivo quantitative mapping of cardiac perfusion
in rats using a noninvasive mr spin-labeling method,” Journal of Magnetic
Resonance Imaging, vol. 8, no. 6, pp. 1240–1245, 1998.
[133] S. D. Wolff and R. S. Balaban, “Magnetization transfer contrast (mtc) and
tissue water proton relaxation in vivo,” Magnetic resonance in medicine,
vol. 10, no. 1, pp. 135–144, 1989.
[134] R. Henkelman, G. Stanisz, and S. Graham, “Magnetization transfer in mri:
a review,” NMR in Biomedicine, vol. 14, no. 2, pp. 57–64, 2001.
[135] R. I. Grossman, J. M. Gomori, K. N. Ramer, F. J. Lexa, and M. D. Schnall,
“Magnetizationtransfer: theoryandclinicalapplicationsinneuroradiology.,”
Radiographics, vol. 14, no. 2, pp. 279–290, 1994.
[136] S. D. Wolff and R. S. Balaban, “Magnetization transfer imaging: practical
aspects and clinical applications.,” Radiology, vol. 192, no. 3, pp. 593–599,
1994.
[137] B. P. Poncelet, T. M. Koelling, C. J. Schmidt, K. K. Kwong, T. G. Reese,
P. Ledden, H. L. Kantor, T. J. Brady, and R. M. Weisskoff, “Measurement
of human myocardial perfusion by double-gated flow alternating inversion
recovery epi,” Magnetic resonance in medicine, vol. 41, no. 3, pp. 510–519,
1999.
[138] C. M. Wacker, M. Bock, A. W. Hartlep, G. Beck, G. van Kaick, G. Ertl,
W. R. Bauer, and L. R. Schad, “Changes in myocardial oxygenation and
perfusion under pharmacological stress with dipyridamole: assessment using
t* 2 and t1 measurements,” Magnetic resonance in medicine, vol. 41, no. 4,
pp. 686–695, 1999.
[139] T. Capron, T. Troalen, B. Robert, A. Jacquier, M. Bernard, and F. Kober,
“Myocardial perfusion assessment in humans using steady-pulsed arterial
spin labeling,” Magnetic resonance in medicine, vol. 74, no. 4, pp. 990–998,
2015.
[140] H. P. Do, T. R. Jao, and K. S. Nayak, “Myocardial arterial spin labeling
perfusion imaging with improved sensitivity,” J Cardiovasc Magn Reson,
vol. 16, no. 1, p. 15, 2014.
[141] H. P. Do, A. J. Yoon, M. W. Fong, F. Saremi, M. L. Barr, and K. S. Nayak,
“Double-gated myocardial asl perfusion imaging is robust to heart rate vari-
ation,” Magnetic resonance in medicine, 2016.
118
[142] C. M. Wacker, F. Fidler, C. Dueren, S. Hirn, P. M. Jakob, G. Ertl, A. Haase,
and W. R. Bauer, “Quantitative assessment of myocardial perfusion with a
spin-labeling technique: Preliminary results in patients with coronary artery
disease,” Journal of Magnetic Resonance Imaging, vol.18, no.5, pp.555–560,
2003.
[143] B. E. Northrup, K. S. McCommis, H. Zhang, S. Ray, P. K. Woodard, R. J.
Gropler, and J. Zheng, “Resting myocardial perfusion quantification with
cmr arterial spin labeling at 1.5 t and 3.0 t,” Journal of Cardiovascular
Magnetic Resonance, vol. 10, no. 1, pp. 1–9, 2008.
[144] G. A. Keith, C. T. Rodgers, M. A. Chappell, and M. D. Robson, “A look-
locker acquisition scheme for quantitative myocardial perfusion imaging with
fair arterial spin labeling in humans at 3 tesla,” Magnetic Resonance in
Medicine, 2016.
[145] C. Waller, E. Kahler, K.-H. Hiller, K. Hu, M. Nahrendorf, S. Voll, A. Haase,
G. Ertl, and W. R. Bauer, “Myocardial perfusion and intracapillary blood
volume in rats at rest and with coronary dilatation: Mr imaging in vivo with
use of a spin-labeling technique 1,” Radiology, vol. 215, no. 1, pp. 189–197,
2000.
[146] J. U. Streif, M. Nahrendorf, K.-H. Hiller, C. Waller, F. Wiesmann, E. Rom-
mel, A. Haase, and W. R. Bauer, “In vivo assessment of absolute perfusion
and intracapillary blood volume in the murine myocardium by spin label-
ing magnetic resonance imaging,” Magnetic resonance in medicine, vol. 53,
no. 3, pp. 584–592, 2005.
[147] M. H. Vandsburger, R. L. Janiczek, Y. Xu, B. A. French, C. H. Meyer, C. M.
Kramer, and F. H. Epstein, “Improved arterial spin labeling after myocardial
infarction in mice using cardiac and respiratory gated look-locker imaging
with fuzzy c-means clustering,” Magnetic resonance in medicine, vol. 63,
no. 3, pp. 648–657, 2010.
[148] A. E. Campbell-Washburn, A. N. Price, J. A. Wells, D. L. Thomas, R. J.
Ordidge, and M. F. Lythgoe, “Cardiac arterial spin labeling using segmented
ecg-gated look-locker fair: Variability and repeatability in preclinical stud-
ies,” Magnetic resonance in medicine, vol. 69, no. 1, pp. 238–247, 2013.
[149] F. Kober, I. Iltis, M. Izquierdo, M. Desrois, D. Ibarrola, P. J. Cozzone,
and M. Bernard, “High-resolution myocardial perfusion mapping in small
animals in vivo by spin-labeling gradient-echo imaging,” Magnetic resonance
in medicine, vol. 51, no. 1, pp. 62–67, 2004.
119
[150] P. Le Roux, “Simplified model and stabilization of ssfp sequences,” Journal
of Magnetic Resonance, vol. 163, no. 1, pp. 23–37, 2003.
[151] J. M. Otton, J. Phan, M. Feneley, C.-y. Yu, N. Sammel, and J. McCrohon,
“Defining the mid-diastolic imaging period for cardiac ct–lessons from tissue
doppler echocardiography,” BMC medical imaging, vol. 13, no. 1, p. 1, 2013.
[152] T. Jao, Z. Zun, P. Varadarajan, R. G. Pai, and K. S. Nayak, “Mapping of
myocardial asl perfusion and perfusion reserve data,” in Proceedings of the
19th Annual Meeting of ISMRM, Montreal, Canana, 2011.
[153] P. Chareonthaitawee, P. A. Kaufmann, O. Rimoldi, and P. G. Camici,
“Heterogeneity of resting and hyperemic myocardial blood flow in healthy
humans,” Cardiovascular research, vol. 50, no. 1, pp. 151–161, 2001.
[154] M. A. Griswold, S. Kannengiesser, R. M. Heidemann, J. Wang, and P. M.
Jakob, “Field-of-view limitations in parallel imaging,” Magnetic resonance
in medicine, vol. 52, no. 5, pp. 1118–1126, 2004.
[155] A. Z. Khawaja, D. B. Cassidy, J. Al Shakarchi, D. G. McGrogan, N. G.
Inston, and R. G. Jones, “Revisiting the risks of mri with gadolinium based
contrast agentsÑreview of literature and guidelines,” Insights into imaging,
vol. 6, no. 5, pp. 553–558, 2015.
[156] M. A. Perazella, “Nephrogenic systemic fibrosis, kidney disease, and gadolin-
ium: is there a link?,” Clinical Journal of the American Society of Nephrol-
ogy, vol. 2, no. 2, pp. 200–202, 2007.
[157] F. Fidler, C. M. Wacker, C. Dueren, M. Weigel, P. M. Jakob, W. R. Bauer,
and A. Haase, “Myocardial perfusion measurements by spin-labeling under
different vasodynamic states,” Journal of Cardiovascular Magnetic Reso-
nance, vol. 6, no. 2, pp. 509–516, 2004.
[158] H. Zhang, S. M. Shea, V. Park, D. Li, P. K. Woodard, R. J. Gropler, and
J. Zheng, “Accurate myocardial t1 measurements: toward quantification of
myocardial blood flow with arterial spin labeling,” Magnetic resonance in
medicine, vol. 53, no. 5, pp. 1135–1142, 2005.
[159] M.Clynes,“Computeranalysisofreflexcontrolandorganization: respiratory
sinus arrhythmia,” Science, vol. 131, no. 3396, pp. 300–302, 1960.
[160] A. J. Raper, D. Richardson, H. Kontos, and J. Patterson, “Circulatory
responses to breath holding in man.,” Journal of applied physiology, vol. 22,
no. 2, pp. 201–206, 1967.
120
[161] S. Leschka, S. Wildermuth, T. Boehm, L. Desbiolles, L. Husmann, A. Plass,
P. Koepfli, T. Schepis, B. Marincek, P. A. Kaufmann, et al., “Noninvasive
coronary angiography with 64-section ct: Effect of average heart rate and
heart rate variability on image quality 1,” Radiology, vol. 241, no. 2, pp. 378–
385, 2006.
[162] G. Sun, M. Li, X. Jiang, L. Li, Z. Peng, G. Li, and L. Xu, “320-detector
row ct coronary angiography: effects of heart rate and heart rate variability
on image quality, diagnostic accuracy and radiation exposure,” The British
journal of radiology, 2014.
[163] X. Jouven, P. J. Schwartz, S. Escolano, C. Straczek, M. Tafflet, M. Desnos,
J. P. Empana, and P. Ducimetière, “Excessive heart rate increase during
mild mental stress in preparation for exercise predicts sudden death in the
general population,” European heart journal, vol. 30, no. 14, pp. 1703–1710,
2009.
[164] D. R. Okada, B. B. Ghoshhajra, R. Blankstein, J. A. Rocha-Filho, L. D.
Shturman, I. S. Rogers, H. G. Bezerra, A. Sarwar, H. Gewirtz, U. Hoffmann,
et al., “Direct comparison of rest and adenosine stress myocardial perfusion
ct with rest and stress spect,” Journal of nuclear cardiology, vol. 17, no. 1,
pp. 27–37, 2010.
[165] L. Ravichandran, C. A. Wick, and S. Tridandapani, “Detection of quies-
centphasesinechocardiographydatausingnon-linearfilteringandboundary
detection,” in Engineering in Medicine and Biology Society (EMBC), 2012
Annual International Conference of the IEEE, pp. 1562–1565, IEEE, 2012.
[166] K. C. Wu, “Cmr of microvascular obstruction and hemorrhage in myocardial
infarction,” Journal of Cardiovascular Magnetic Resonance, vol. 14, no. 1,
p. 68, 2012.
[167] S. H. Rezkalla and R. A. Kloner, “No-reflow phenomenon,” Circulation,
vol. 105, no. 5, pp. 656–662, 2002.
[168] G. Niccoli, F. Burzotta, L. Galiuto, and F. Crea, “Myocardial no-reflow
in humans,” Journal of the American College of Cardiology, vol. 54, no. 4,
pp. 281–292, 2009.
[169] H. Ito, T. Tomooka, N. Sakai, H. Yu, Y. Higashino, K. Fujii, T. Masuyama,
A.Kitabatake, andT.Minamino, “Lackofmyocardialperfusionimmediately
after successful thrombolysis. a predictor of poor recovery of left ventricu-
lar function in anterior myocardial infarction.,” Circulation, vol. 85, no. 5,
pp. 1699–1705, 1992.
121
[170] K. C. Wu, E. A. Zerhouni, R. M. Judd, C. H. Lugo-Olivieri, L. A. Barouch,
S. P. Schulman, R. S. Blumenthal, and J. A. Lima, “Prognostic significance
of microvascular obstruction by magnetic resonance imaging in patients with
acute myocardial infarction,” Circulation, vol. 97, no. 8, pp. 765–772, 1998.
[171] V. Hombach, O. Grebe, N. Merkle, S. Waldenmaier, M. Höher, M. Kochs,
J. Wöhrle, and H. A. Kestler, “Sequelae of acute myocardial infarction
regarding cardiac structure and function and their prognostic significance as
assessed by magnetic resonance imaging,” European heart journal, vol. 26,
no. 6, pp. 549–557, 2005.
[172] N. G. Uren, T. Crake, D. C. Lefroy, R. de Silva, G. J. Davies, and A. Maseri,
“Reducedcoronaryvasodilatorfunctionininfarctedandnormalmyocardium
after myocardial infarction,” New England Journal of Medicine, vol. 331,
no. 4, pp. 222–227, 1994.
[173] F. Kober, T. Jao, T. Troalen, and K. S. Nayak, “Myocardial arterial spin
labeling,” Journal of Cardiovascular Magnetic Resonance, vol. 18, no. 1,
p. 22, 2016.
[174] H. P. Do, A. Javed, T. R. Jao, H.-w. Kim, A. J. Yoon, and K. S. Nayak,
“Arterialspinlabelingcmrperfusionimagingiscapableofcontinuouslymon-
itoring myocardial blood flow during stress,” Journal of Cardiovascular Mag-
netic Resonance, vol. 17, no. 1, p. P145, 2015.
[175] K.P.Pruessmann, M.Weiger, P.Börnert, andP.Boesiger, “Advancesinsen-
sitivity encoding with arbitrary k-space trajectories,” Magnetic Resonance in
Medicine, vol. 46, no. 4, pp. 638–651, 2001.
[176] N. R. Ghugre, M. Pop, J. Barry, K. A. Connelly, and G. A. Wright, “Quan-
titative magnetic resonance imaging can distinguish remodeling mechanisms
after acute myocardial infarction based on the severity of ischemic insult,”
Magnetic resonance in medicine, vol. 70, no. 4, pp. 1095–1105, 2013.
[177] N. R. Ghugre, V. Ramanan, M. Pop, Y. Yang, J. Barry, B. Qiang, K. A.
Connelly, A. J. Dick, and G. A. Wright, “Quantitative tracking of edema,
hemorrhage, and microvascular obstruction in subacute myocardial infarc-
tion in a porcine model by mri,” Magnetic Resonance in Medicine, vol. 66,
no. 4, pp. 1129–1141, 2011.
[178] M. Schmitt, G. Horstick, S. E. Petersen, A. Karg, N. Hoffmann, T. Gum-
brich, N. Abegunewardene, and W. G. Schreiber, “Quantification of resting
myocardial blood flow in a pig model of acute ischemia based on first-pass
mri,” Magnetic resonance in medicine, vol. 53, no. 5, pp. 1223–1227, 2005.
122
[179] A. H. Mahnken, E. Klotz, H. Pietsch, B. Schmidt, T. Allmendinger,
U. Haberland, W. A. Kalender, and T. Flohr, “Quantitative whole heart
stress perfusion ct imaging as noninvasive assessment of hemodynamics in
coronary artery stenosis: preliminary animal experience,” Investigative radi-
ology, vol. 45, no. 6, pp. 298–305, 2010.
[180] D. A. Schwinn, R. W. McIntyre, and J. Reves, “Isoflurane-induced vasodila-
tion: Role of the [alpha]-adrenergic nervous system.,” Anesthesia & Analge-
sia, vol. 71, no. 5, pp. 451–459, 1990.
[181] D. R. Larach and H. G. Schuler, “Direct vasodilation by sevoflurane, isoflu-
rane, and halothane alters coronary flow reserve in the isolated rat heart.,”
Anesthesiology, vol. 75, no. 2, pp. 268–278, 1991.
[182] G. J. Crystal, E. A. Czinn, J. M. Silver, and R. M. Salem, “Coronary vasodi-
lation by isoflurane abrupt versus gradual administration,” The Journal of
the American Society of Anesthesiologists, vol. 82, no. 2, pp. 542–549, 1995.
[183] G. J. Crystal and M. R. Salem, “Isoflurane causes vasodilation in the coro-
nary circulation,” The Journal of the American Society of Anesthesiologists,
vol. 98, no. 4, pp. 1030–1030, 2003.
[184] A. K. Gamperl, T. W. Hein, L. Kuo, and B. A. Cason, “Isoflurane-induced
dilation of porcine coronary microvessels is endothelium dependent and
inhibited by glibenclamide,” The Journal of the American Society of Anes-
thesiologists, vol. 96, no. 6, pp. 1465–1471, 2002.
[185] R. Jaffe, T. Charron, G. Puley, A. Dick, and B. H. Strauss, “Microvascu-
lar obstruction and the no-reflow phenomenon after percutaneous coronary
intervention,” Circulation, vol. 117, no. 24, pp. 3152–3156, 2008.
[186] F. A. Breuer, M. Blaimer, R. M. Heidemann, M. F. Mueller, M. A. Gris-
wold, and P. M. Jakob, “Controlled aliasing in parallel imaging results in
higher acceleration (caipirinha) for multi-slice imaging,” Magnetic resonance
in medicine, vol. 53, no. 3, pp. 684–691, 2005.
[187] C. M. Gibson, C. P. Cannon, W. L. Daley, J. T. Dodge, B. Alexander, S. J.
Marble, C. H. McCabe, L. Raymond, T. Fortin, W. K. Poole, et al., “Timi
frame count,” Circulation, vol. 93, no. 5, pp. 879–888, 1996.
[188] N. R. Ghugre, V. Ramanan, M. Pop, Y. Yang, J. Barry, B. Qiang, K. A.
Connelly, A. J. Dick, and G. A. Wright, “Myocardial bold imaging at 3 t
using quantitative t2: application in a myocardial infarct model,” Magnetic
resonance in medicine, vol. 66, no. 6, pp. 1739–1747, 2011.
123
[189] F. von Knobelsdorff-Brenkenhoff, M. Prothmann, M. A. Dieringer, R. Wass-
muth, A.Greiser, C.Schwenke, T.Niendorf, andJ.Schulz-Menger, “Myocar-
dial t 1 and t 2 mapping at 3 t: reference values, influencing factors and
implications,” Journal of Cardiovascular Magnetic Resonance, vol. 15, no. 1,
p. 53, 2013.
[190] M. Fenchel, P. Martirosian, J. Langanke, J. Giersch, S. Miller, N. I. Stauder,
U. Kramer, C. D. Claussen, and F. Schick, “Perfusion mr imaging with fair
true fisp spin labeling in patients with and without renal artery stenosis:
Initial experience 1,” Radiology, vol. 238, no. 3, pp. 1013–1021, 2006.
[191] P. Reymond, F. Merenda, F. Perren, D. Rüfenacht, and N. Stergiopulos,
“Validation of a one-dimensional model of the systemic arterial tree,” Amer-
ican Journal of Physiology-Heart and Circulatory Physiology, vol. 297, no. 1,
pp. H208–H222, 2009.
[192] P. M. Robson, A. J. Madhuranthakam, W. Dai, I. Pedrosa, N. M. Rofsky,
and D. C. Alsop, “Strategies for reducing respiratory motion artifacts in
renal perfusion imaging with arterial spin labeling,” Magnetic resonance in
medicine, vol. 61, no. 6, pp. 1374–1387, 2009.
[193] J. Guo and E. C. Wong, “Increased snr efficiency in velocity selective arte-
rial spin labeling using multiple velocity selective saturation modules (mm-
vsasl),” Magnetic resonance in medicine, vol. 74, no. 3, pp. 694–705, 2015.
124

Abstract (if available)

Abstract Coronary artery disease (CAD) affects more than 15.5 million Americans and causes approximately 310,000 deaths per year. One of the most common diagnostic tests is perfusion stress testing, primarily performed using single photon emission computed tomography (SPECT) or first-pass cardiovascular magnetic resonance (CMR). These methods require the use of ionizing radiation or exogenous contrast agents that carry associated risks to patients, especially those who require frequent assessment or have kidney dysfunction. Myocardial arterial spin labeling (ASL) is a promising MRI-based perfusion imaging method that can quantitatively measure myocardial tissue perfusion without the use of ionizing radiation or exogenous contrast agents. Feasibility of CAD detection using ASL has been previously demonstrated, however, several challenges remain, including low sensitivity, coarse spatial resolution, and limited spatial coverage. The contributions of this dissertation are (1) improving sensitivity, (2) exploring clinical applications, and (3) developing a new and advantageous labeling method for myocardial ASL. ❧ Low sensitivity is one of the major limitations of current myocardial ASL methods. I found that cardiac motion is one of the dominant sources of measurement variability and that shortening the acquisition window from 300 ms to 150 ms using parallel imaging significantly reduces measurement variability, i.e. improves sensitivity. I also implemented double-gating which further improves sensitivity of myocardial ASL method. Furthermore, I demonstrated that double-gating is robust to heart rate variation. ❧ I demonstrated a potential application of myocardial ASL in assessment of microvascular obstruction (MVO) in a pig model of acute myocardial infarction (AMI). This study also demonstrated that myocardial ASL is able to assess clinically relevant increase in blood flow during vasodilation. That potentially allows myocardial ASL to assess microvascular dysfunction as studies have shown that microvascular function is impaired not only in the infarcted but also in the remote territories. ❧ Myocardial T₁ mapping with appropriate flow-sensitive preparation is able to assess myocardial blood flow, myocardial blood flow change, and myocardial blood volume change under vasodilation. I demonstrated the feasibility of using myocardial T₁ mapping for assessment of MBF in human subjects. I also explored potential application of T₁-based myocardial blood volume change mapping in heart transplant recipients using myocardial T₁ mapping. ❧ I developed a new and more efficient labeling method named physiologically synchronized multi-module pulsed ASL (SymPASL), which may enable myocardial ASL with volumetric coverage. SymPASL utilizes several labeling pulses that are synchronized with blood flow and image acquisitions are synchronized with motion of the target organ. From simulation and in vivo experiment in human kidneys, SymPASL provides superior ASL signal and signal-to-noise ratio (SNR) efficiency compared to the pulsed ASL method with 1 heart beat post-labeling-delay and has similar ASL signal and SNR efficiency to pseudo-continuous ASL (PCASL) with 1.5-second labeling duration and 1.0-second post-labeling-delay.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Improving sensitivity and spatial coverage of myocardial arterial spin labeling

PDF

Improving the sensitivity and spatial coverage of cardiac arterial spin labeling for assessment of coronary artery disease

PDF

Assessment of myocardial blood flow in humans using arterial spin labeled MRI

PDF

Radiofrequency pulse performance for myocardial ASL

PDF

Measurement of human brain perfusion with arterial spin labeling magnetic resonance imaging at ultra-high field

PDF

Dynamic cardiovascular magnetic resonance imaging for improved assessment of ischemic heart disease

PDF

Improving arterial spin labeling in clinical application with deep learning

PDF

Characterization of lenticulostriate arteries using high-resolution black blood MRI as an early imaging biomarker for vascular cognitive impairment and dementia

PDF

Mapping water exchange rate across the blood-brain barrier

PDF

Estimating liver iron non-invasively with high-field MRI

PDF

Seeing sleep: real-time MRI methods for the evaluation of sleep apnea

PDF

Configurable imaging platform for super-harmonic contrast-enhanced ultrasound imaging

Asset Metadata

Creator Do, Hung Phi (author)

Core Title Improved myocardial arterial spin labeled perfusion imaging

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Physics

Publication Date 05/11/2019

Defense Date 03/21/2017

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag acute myocardial infarction,AMI,ASL,CAD,coronary artery disease,improved sensitivity,magnetic resonance imaging,MRI,myocardial arterial spin labeled,myocardial blood flow,myocardial blood volume,myocardial perfusion imaging,myocardial T₁ mapping,OAI-PMH Harvest,perfusion stress testing,physiological noise

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Nakano, Aiichiro (committee chair), Nayak, Krishna S. (committee chair), El-Naggar, Moh (committee member), Haas, Stephan (committee member), Haldar, Justin (committee member), Wong, Eric, (University of California at San Diego) (committee member)

Creator Email hungdo@usc.edu,hungdop@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-372782

Unique identifier UC11255891

Identifier etd-DoHungPhi-5310.pdf (filename),usctheses-c40-372782 (legacy record id)

Legacy Identifier etd-DoHungPhi-5310.pdf

Dmrecord 372782

Document Type Dissertation

Rights Do, Hung Phi

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA