assessment of peripheral microcirculation using magnetic resonance imaging … · resonance imaging...
TRANSCRIPT
Assessment of Peripheral Microcirculation using Magnetic Resonance Imaging
by
Hou-Jen Howard Chen
A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy
Graduate Department of Medical Biophysics University of Toronto
© Copyright by Hou-Jen Chen 2018
ii
Assessment of Peripheral Microcirculation using Magnetic
Resonance Imaging
Hou-Jen Chen
Doctor of Philosophy
Graduate Department of Medical Biophysics University of Toronto
2018
Abstract
This thesis describes the development of a new measurement and analysis tool to assess the
microcirculation in limb ischemia. Peripheral arterial disease (PAD) is diagnosed by reduced bulk
blood flow and blood pressures in the legs. Patients with severe limb ischemia have limited
walking ability and may experience rest pain or foot ulceration. PAD can be treated with
revascularization for their large-vessel diseases. However, some patients respond poorly to
revascularization. It is hypothesized that the microvascular disease component of limb ischemia
may contribute to the differential outcome of revascularization. Therefore, this project is motivated
to probe peripheral microvascular perfusion and to investigate the related vascular physiology and
clinical indications of limb ischemia. A magnetic resonance imaging-based technique, called
arterial spin labeling (ASL), was developed to characterize the dynamic perfusion response to brief
periods of flow interruption. In a group of healthy subjects, the technique captured the changes of
calf muscle reactive hyperemia with respect to the different durations of flow interruption,
reflecting the microvascular function in regulating perfusion for different amounts of metabolic
stresses. In addition, the technique also captured the differences between patients with PAD and
healthy subjects in the dynamic characteristics of reactive hyperemic responses. Next, a
quantitative physiological model, incorporating the influences of arterial stenoses and
microvascular dysfunction, was established to facilitate the physiological interpretation of ASL
reactive hyperemia. Model-derived perfusion indices were compared between subject groups.
Patients were found to have higher arterial resistance and lower microvascular sensitivity to
hypoxia than the healthy subject group. Finally, the perfusion measures were compared with
symptom severity and functional outcome 6-month following successful endovascular
iii
revascularization in a cohort of patients. Measures of arterial resistance were related to the
manifested severity, whereas the microvascular sensitivity index appeared to differentiate
prognosis following revascularization. Overall, this thesis improved the ASL method to assess
peripheral muscle perfusion and demonstrated the clinical relevance of perfusion measures to
symptom severity and revascularization outcome of PAD for the first time. A future study with a
larger patient population and longer follow-up period will be required to confirm the current
findings.
iv
Acknowledgments
My doctoral work was performed under the supervision of Dr. Graham Wright. I would like to
thank him for providing me the opportunity and guiding me through this long journey. His wise
words, dedication to work, and tireless effort in mentoring students are an inspiration. I am also
grateful for his persistent friendliness and positivity.
I would like to thank my supervisory committee members, Drs. David Goertz and Bojana
Stefanovic, for their insights and encouragements. I believe my project and myself have benefited
greatly from their advice.
There were lab members and peers who helped me in my project. I thank Venkat and Dr. Garry
Liu for helping me on the MRI sequences. I thank Robert, Greg, Li, Adrienne, Philippa, Trisha,
and Siavash for keeping things interesting around the lab. Also, I am grateful for their participation
in my volunteer study, along with Hirad, Justin, and Omodele, to make things happen. I would
also like to thank Trisha and Mary for recruiting patients in my clinical study.
I would like to thank my parents and my wife Liz for their love and support during these years.
Most of all, I thank Liz for keeping my life outside of work filled with friendship, entertainment,
and adventure.
v
List of Contributions
Journal papers
1. Chen HJ, Wright GA, “A Physiological Model for Interpretation of Arterial Spin Labeling
Reactive Hyperemia of Calf Muscles.” PLoS ONE, 2017.
2. Chen HJ, Roy TL, Wright GA, “Perfusion Measures for Symptom Severity and Differential
Outcome of Revascularization in Limb Ischemia — Preliminary Results With Arterial Spin
Labeling Reactive Hyperemia.” Journal of Magnetic Resonance Imaging, 2017.
3. Chen HJ, Wright GA, “Investigating discrepancies in arterial spin labeling to improve
characterization of calf muscle perfusion.” NMR in Biomedicine, 2017 (In review).
Conference presentations
1. Chen HJ, Roy TL, Wright GA, “Calf reactive hyperemia indicates severity and predicts
functional outcome of limb ischemia: a pilot study using arterial spin labeling and model-
based analysis.” Society for Magnetic Resonance Angiography 2017, Stellenbosch, South
Africa
2. Chen HJ, Wright GA, “A physiological model for arterial spin labeling reactive hyperemia
in calf muscles: assessing microvascular dysfunction in the presence of arterial stenoses.”
Society for Magnetic Resonance Angiography 2016, Chicago, USA
3. Chen HJ, Wright GA, “Physiological model for the interpretation of ASL-measured
reactive hyperemia in leg muscles.” ISMRM Workshop on Non-Contrast Cardiovascular
MRI 2015, Long Beach, CA, USA
4. Chen HJ, Wright GA, “Protocol optimization and physiological specificity of ASL-
measured reactive hyperemia in skeletal muscle.” The Society of Cardiovascular Magnetic
Resonance 2014, New Orleans, USA
5. Chen HJ, Wright GA, “Investigation of physiological parameters for pulsed arterial spin
labeling in calf muscles.” ISMRM 2012, Melbourne, Australia.
vi
Table of Contents
Acknowledgments .................................................................................................................... iv
List of Contributions ..................................................................................................................v
Table of Contents ..................................................................................................................... vi
List of Tables ........................................................................................................................... ix
List of Figures ............................................................................................................................x
List of Abbreviations .............................................................................................................. xii
Chapter 1 ....................................................................................................................................1
Introduction ...........................................................................................................................1
1.1 Peripheral Arterial Disease ............................................................................................1
1.1.1 Diagnosis............................................................................................................2
1.1.2 Peripheral revascularization and outcome .........................................................4
1.1.3 Motivation to assess the microcirculation ..........................................................5
1.2 Microvascular disease and assessment ..........................................................................5
1.2.1 The form of microvascular disease ....................................................................6
1.2.2 Structure and function of the microcirculation ..................................................6
1.2.3 Assessment of microvascular dysfunction .........................................................9
1.2.4 Noninvasive measurement of peripheral microcirculation ..............................11
1.3 MRI for peripheral microvascular assessment .............................................................12
1.3.1 MRI physics .....................................................................................................12
1.3.2 Arterial spin labeling ........................................................................................13
1.3.3 DCE and BOLD imaging .................................................................................14
1.4 Thesis overview ...........................................................................................................15
Chapter 2 ..................................................................................................................................17
ASL reactive hyperemia in calf muscles .............................................................................17
2.1 Background ..................................................................................................................17
vii
2.1.1 Technical considerations in ASL .....................................................................18
2.1.2 Physiological influences on perfusion characteristics .....................................20
2.2 General methods ..........................................................................................................21
2.3 Technical experiments of peripheral ASL ...................................................................23
2.3.1 Background artifact ..........................................................................................23
2.3.2 Tibial blood velocity ........................................................................................26
2.3.3 Simulation of pseudo-continuous labeling .......................................................27
2.4 Physiological influences on perfusion characteristics .................................................29
2.5 Discussion ....................................................................................................................34
2.5.1 Technical experiments of peripheral ASL .......................................................34
2.5.2 Characteristics of ASL reactive hyperemia .....................................................36
2.6 Conclusion ...................................................................................................................38
Chapter 3 ..................................................................................................................................39
Model-based interpretation of reactive hyperemia .............................................................39
3.1 Background ..................................................................................................................39
3.2 The model ....................................................................................................................40
3.2.1 The lumped parameters ....................................................................................42
3.2.2 Microvascular regulation .................................................................................44
3.2.3 Model behavior ................................................................................................47
3.3 Methods........................................................................................................................49
3.4 Results ..........................................................................................................................51
3.5 Discussion ....................................................................................................................54
3.5.1 Establishment of the model ..............................................................................54
3.5.2 Major findings ..................................................................................................56
3.6 Conclusion ...................................................................................................................59
Appendix .............................................................................................................................60
viii
A. The flow equations of the system ....................................................................60
B. Modeling the regulatory influences .................................................................61
C. Definition of microvascular regulatory parameters .........................................63
Chapter 4 ..................................................................................................................................65
Perfusion in different status of limb ischemia ....................................................................65
4.1 Introduction ..................................................................................................................65
4.2 Methods........................................................................................................................67
4.3 Results ..........................................................................................................................70
4.4 Discussion ....................................................................................................................78
4.5 Conclusion ...................................................................................................................80
Chapter 5 ..................................................................................................................................83
Summary and future work ...................................................................................................83
5.1 Summary ......................................................................................................................83
5.2 Experimental improvement ..........................................................................................85
5.3 Future directions: variations of ASL for different applications ...................................85
5.3.1 Vascular territory mapping ..............................................................................86
5.3.2 Multislice peripheral perfusion ........................................................................86
5.4 Future directions: clinical studies ................................................................................87
5.4.1 Validation of microvascular disease with invasive measurement ...................87
5.4.2 The influence of diabetes on PAD ...................................................................88
5.4.3 Evaluation of therapeutic effect .......................................................................88
5.5 Final words...................................................................................................................90
References ................................................................................................................................91
ix
List of Tables
Table 2.1: Reactive hyperemia in soleus muscle measured in previous ASL studies. ................. 18
Table 2.2: Comparing subtraction schemes. ................................................................................. 26
Table 2.3: Tibial blood velocity throughout the ischemia-reperfusion paradigm (n=4). .............. 26
Table 2.4: Characteristics of reactive hyperemia induced by different durations of ischemia. .... 32
Table 2.5: Intrasession repeatability of 2-min ischemia. .............................................................. 32
Table 3.1: Constant parameters of the model. .............................................................................. 42
Table 4.1: Demographics of the participants. ............................................................................... 70
Table 4.2: Comparison of measurement indices between outcome groups .................................. 77
x
List of Figures
Figure 1.1: The diagnostic measurements in the vascular lab report. ............................................. 3
Figure 1.2: Pressure distribution and arteriole structure. ................................................................ 7
Figure 1.3: Biomechanical components of vessel wall tension. ..................................................... 8
Figure 1.4: Skeletal muscle hyperemia. ........................................................................................ 10
Figure 2.1: Illustration of ASL reactive hyperemia induced by 2 minutes of flow interruption. . 17
Figure 2.2: The configuration and difference signal in continuous and pulsed labeling. ............. 19
Figure 2.3: Slice prescription and determination of region-of-interest in pCASL experiments. .. 23
Figure 2.4: Sanity check of pCASL in a healthy subject. ............................................................. 25
Figure 2.5: Bloch simulation of velocity-driven adiabatic inversion. .......................................... 28
Figure 2.6: Reactive hyperemia induced by various durations of ischemia ................................. 31
Figure 2.7: Comparison of reactive hyperemia between healthy subjects and patients. .............. 33
Figure 3.1: The flow circuit of calf circulation. ............................................................................ 41
Figure 3.2: Ischemic responses of the tissue and arterioles. ......................................................... 48
Figure 3.3: The effects of ischemic duration, arterial stenosis, and microvascular dysfunction on
reactive hyperemia. ....................................................................................................................... 49
Figure 3.4: Fitting the responses of a healthy subject with the model. ......................................... 52
Figure 3.5: Fitting of patients’ responses to 2-min cuffing. ......................................................... 53
Figure 3.6: Characteristics of reactive hyperemia induced by 2-min cuffing. .............................. 54
Figure 4.1: An illustration of physiological influences on the shape of reactive hyperemia ........ 66
xi
Figure 4.2: An illustration of the acquisition and model-based characterization of ASL reactive
hyperemia. ..................................................................................................................................... 69
Figure 4.3: A diagram of patient participation. ............................................................................. 71
Figure 4.4: Bilateral comparison. .................................................................................................. 73
Figure 4.5: Perfusion characteristics and averaged waveforms for different severities. .............. 74
Figure 4.6: Individual reactive hyperemia before and after revascularization. ............................ 75
Figure 4.7: A scatter plot of perfusion indices in the two outcome groups. ................................. 76
xii
List of Abbreviations
ABI Ankle-brachial index
AFP Adiabatic fast passage
AMP Adenosine monophosphate
ANOVA Analysis of variance
ASL Arterial spin labeling
ATP Adenosine triphosphate
BOLD Blood oxygenation level-dependent
CAD Coronary artery disease
CFR Coronary flow reserve
CLI Critical limb ischemia
CV Coefficient of variation
DCE Dynamic contrast-enhanced
EPI Echo-planer imaging
FAIR Flow-sensitive alternating inversion recovery
FFR Fractional flow reserve
FMD Flow-mediated dilatation
HFV Hyperemic flow volume
ICC Intraclass correlation coefficient
IMR Index of microvascular resistance
LDF Laser Doppler flowmetry
MRI Magnetic resonance imaging
Mz Longitudinal magnetization
NO Nitric oxide
PAD Peripheral arterial disease
pCASL Pseudo-continuous ASL
xiii
PET Positron emission tomography
PLD Post-labeling delay in ASL
PTA Percutaneous transluminal angioplasty
Q2TIPS QUIPSS II with Thin-slice TI1 Periodic Saturation
QUIPSS Quantitative Imaging of Perfusion Using a Single Subtraction
RBC Red blood cells
RF Radio frequency
ROI Region of interest
SD Standard deviation
SNR Signal-to-noise ratio
SSFP Steady-state free precession
T1 Longitudinal relaxation time, related to spin-lattice relaxation
T1B Blood T1
T2 Transverse relaxation time, related to spin-spin relaxation
T2B Blood T2
T2* Apparent transverse relaxation time due to the effect of local field inhomogeneity
TcPO2 Transcutaneous oxygen tension
TE Echo time
TI1 Inversion time in the Q2TIPS scheme
TR Repetition time
TTP Time-to-peak of reactive hyperemic response
TTR Time-to-recovery of reactive hyperemic response
VENC Velocity encoding in phase contrast MRI
VOP Venous occlusion plethysmography
1
Chapter 1
Introduction Limb ischemia refers to the situation where the oxygen demand of the leg muscles is not met by
the supply due to reduced blood flow. Limb ischemia is usually caused by progressive narrowing
of the major arteries of the leg, called peripheral arterial disease (PAD). Patients with mild to
moderate disease usually present with leg pain during or after exercise, a symptom called
intermittent claudication. Severe PAD can cause symptoms of critical limb ischemia (CLI),
including rest pain, ulcers, and gangrene. CLI can lead to leg amputation if not treated successfully.
Limb salvage remains challenging in patients with CLI. Although the occlusive large-vessel
disease component can be assessed and treated clinically, little is known about the microcirculatory
changes in limb ischemia.
This thesis deals with the assessment of the microcirculation in patients with known PAD.
Magnetic resonance imaging (MRI) allows direct measurement of dynamic perfusion in the leg
muscles; therefore, this thesis presents a new MRI-based technique to characterize perfusion in
patients with different disease status, providing a better understanding of the role of the
microcirculation in the overall pathophysiology of limb ischemia. Chapter 1 discusses the context
in which perfusion MRI can be applied and motivates the bodies of work in Chapter 2, 3 and 4.
Chapter 5 discusses the proposed future directions for peripheral microvascular assessment with
MRI.
1.1 Peripheral Arterial Disease
It is estimated that PAD exists in 3 to 10% of the entire population, increasing to 15 to 20% in
those over 70 years [1]. In Canada, PAD affects approximately 800,000 people [2]. In these
patients, arterial narrowing (called stenosis) and occlusions are the main focus for diagnosis and
treatment.
2
1.1.1 Diagnosis
The standard method to detect PAD is to measure the ankle-brachial index (ABI) and blood
velocity of the leg arteries. The key measurements in the vascular laboratory exam report of a
patient is shown in Figure 1.1. ABI is the ratio of brachial artery blood pressure to the tibial artery
blood pressure at the ankle, reflecting the pressure gradient of bulk arterial flow. An ABI less than
0.9 is defined as PAD and less than 0.5 is deemed severe. Normal individuals should have an ABI
ranging between 0.9 and 1.2.�Blood pressure measurement involves vessel compression with an
air-inflated cuff placed around the limb, followed by blood flow detection with ultrasound during
a gradual deflation of the air cuff. Incompressible arteries can occur in patients with significant
arterial calcification, often seen in patients with comorbidity of diabetes, resulting in ABI above
1.2.
3
Figure 1.1: The diagnostic measurements in the vascular lab report. The image was adapted from a subject in the
patient cohort.
To identify the location and assess the impact of the arterial blockages, the peak systolic blood
velocity and velocity waveform along the arteries of the leg are measured with duplex ultrasound.
The blood velocity in the leg arteries is pulsatile and triphasic. Taking the femoral artery for
example, the peak velocity is 40 cm/s and the mean velocity over a pulse is 10 cm/s [3]. Triphasic
flow means that for each cardiac cycle there is forward flow with a clear systolic peak, followed
by a reverse (retrograde) phase and then a much smaller forward flow. Abnormal blood velocity
patterns include high blood velocity through stenotic segments, turbulent flow, and monophasic
Preliminary Scan
Exercise related limb pain
New Outpatient
J. Hardaker, RVTM. Bennett, MD
R. Maggisano, MD, FRCS(C), FACS
Aug 31, 2015 2:27 PMReason for Study:
Primary Indication:
Admission Status:
Technologist:Interpreting Physician:
Referring Physicians:
Study Date:
Gender:Date of Birth: May 05, 1934
Female
Patient Name: Whitty, Bernie
Hypertension MedsHyperlipidemia MedsAnticoagulation Warfarin; A-fib.
HISTORYMI ~ 2001CVA - June /15 affecting right hand and speechHx Pain in right calf with walking ~ 5 min.; rests 4 min.
RESULTSVelocity(cm/s)
Size(cm)
Stenosis
Aorta:
Velocity(cm/s)
<50%Biphasic67<50%Biphasic112<50%Biphasic116<50%Biphasic106
<50%Biphasic78
<50%Triphasic100
StenosisFlow
Left
108 Biphasic <50%74 Biphasic <50%
<50%Biphasic116
<50%Biphasic87
Prox PFA:Prox SFAMid SFADist SFA
Popliteal:
Post.Tibial:Peroneal:Ant. Tibial:Dors. Pedis:
Dist EIA:CFA:
Pop AK:
Pop BK:
Brachial:Ant. Tibial:Post.Tibial:
ABI
126160156
Pressure(mmHg)
1.131.11
Prox PFA:
Right
Prox SFAMid SFADist SFA
Popliteal:
Post.Tibial:Peroneal:Ant. Tibial:
Velocity(cm/s)
Flow Stenosis
<50%Biphasic92<50%Biphasic45<50%Biphasic37
50-99%Stenotic546
50-99%Stenotic237
<50%Biphasic44
<50%Biphasic48Dors. Pedis:
Dist EIA: 86 Triphasic <50%CFA: 83 Biphasic <50%
Pop AK: <50%Monophasic18
Pop BK: <50%Biphasic46
Brachial:Ant. Tibial:Post.Tibial:
ABI
1418090
Pressure(mmHg)
0.570.64
FINDINGSRight: Left:Systolic velocities and heterogeneous plaque suggest a 50-99% stenosis in the distal superficial femoral and mid popliteal arteries. Heterogeneous plaque suggests <50% stenosis throughout the lower extremity. Ankle-brachial indices (ABIs) are in the moderate range.
Heterogeneous plaque suggests <50% stenosis throughout the lower extremity. Ankle-brachial indices (ABIs) are in the normal range.
Sunnybrook Health Sciences CentreVascular Laboratory Room E245
Patient Name:Patient ID:Study Date:
Whitty, Bernie81906308/31/2015Lower Extremity Arterial Duplex
ActiveReports Evaluation. Copyright 2002-2004 (c) Data Dynamics, Ltd. All Rights Reserved. Page 1 of 2
2075 Bayview Avenue, Room E245Toronto, ON M4N 3M5Tel: (416) 480-6851 Fax: (416) 480-6892
4
flow. Essentially, arterial segments with abnormally high blood velocity and loss of a triphasic
pattern are where occlusive lesions exist.
In addition to blood pressure and blood velocity measurements, anatomical and morphological
details such as the degree of stenosis, length of occlusion, and quality of run-off vessels are
acquired via imaging. Imaging techniques include X-ray angiography, computed tomography
angiography, and magnetic resonance angiography.
1.1.2 Peripheral revascularization and outcome
The conventional treatment for severe claudication and CLI is revascularization via bypass surgery
or percutaneous endovascular interventions. Bypass surgery uses synthetic or venous grafts to
redirect blood flow to the ischemic area. In a percutaneous intervention, a balloon catheter is
advanced into the obstructed artery and inflated to expand the narrowing while the vascular
anatomy is visualized using X-ray imaging. Sometimes vascular stents may be deployed to
maintain the opening. Percutaneous revascularization is preferred over bypass treatment because
it is much less invasive and provides faster recovery than bypass surgery. However, technical
failure may occur when the lesions are too long or too stiff to cross with the interventional
guidewire. Those who are not suitable for percutaneous treatment or fail in the attempt may
undergo bypass surgery. Patients with no revascularization option will often undergo amputation.
The outcome of PAD varies enormously among different patient populations. For patients with
claudication, less than 4% require major amputation over a 5-year period [1]. For patients with
CLI, the amputation-free survival is around 65% in 1 year and 25-40% in 5 years after
revascularization [4]. Moreover, limb salvage does not necessarily mean freedom from ischemic
pain. There is roughly only a 50% chance to stay pain resolved for patients with CLI one year after
revascularization [1]. The size, morphology, and location of lesions have a large impact on the
technical success rate of revascularization, but the overall outcome also depends on the residual
disease burden and occlusion recurrence rate. It is common to see patients with multi-level disease
and a mixture of focal and diffuse plaques along the peripheral artery tree. Diffuse plaques or distal
lesions in the arteries of the calf and foot are less likely to be treated with percutaneous intervention
than are focal, proximal lesions. Meanwhile, restenosis following angioplasty is quite common,
5
leading to repeated intervention [5], [6]. There are also patients with isolated tibial disease, who
tend to have even worse outcomes than those with multi-level disease [7]. Overall, the complexity
of the pathophysiology in PAD goes beyond large-vessel disease and revascularization. A more
comprehensive understanding of limb ischemia, including the role of microcirculation, is needed.
1.1.3 Motivation to assess the microcirculation�
The influence of microcirculation in PAD has been recognized in several aspects, including
symptom presentation, treatment efficacy, and treatment durability. First, the actual cause of
ischemic pain is insufficient blood flow at the tissue level, or perfusion, which involves
microvascular regulation in addition to arterial inflow. The correlation between symptom severity
and disease burden detectable in the aforementioned macrovascular assessment is weak,
suggesting the presence of an additional variable related to the microcirculation. Secondly, in
predicting healing of a foot ulcer after revascularization, sustained oxygenation of the skin
measured by transcutaneous oxygen tension (TcPO2) performs better than the indication of arterial
flow restoration by ABI [8], [9]. The fact that TcPO2 takes skin microcirculation into account may
explain its better predictive power. Lastly, the rate of bypass graft thrombosis and restenosis
following angioplasty and stenting is associated with low arterial blood velocity and shear stress
[10], which are determined by the microvascular resistance downstream. The burden at the
microvascular level is very likely to be variable within the patient population, resulting in
differential responses to revascularization and limited usefulness of macrovascular assessment for
predicting prognosis. Therefore, we set out to develop a tool to assess the peripheral
microcirculation, which may provide a better understanding of the pathophysiology involved in
limb ischemia.
1.2 Microvascular disease and assessment
Unlike macrovascular disease, microvascular disease in limb ischemia is not well understood. In
this section, we will explore the form of microvascular disease in other organs, describe
microvascular physiology, and review a few assessment techniques.
6
1.2.1 The form of microvascular disease
The more recognized microvascular diseases are the microvascular complications of diabetes and
hypertension, including retinopathy, nephropathy, and neuropathy. These pathologies are
characterized by leaky small vessels, wall thickening of small arteries, and thickening of the
capillary basement membrane [11]. Although not counted among the classic microvascular
complications, reduced leg muscle capillary density in type 2 diabetes has also been reported [12].
Overall, these morphological findings are direct evidence of microvascular disease. But more
importantly, morphological changes may correspond to a reduction in microvascular blood flow if
functional assessment is performed.
Perhaps a more likely form of the microvascular disease in skeletal muscle is a functional
impairment similar to the one assessed with coronary flow reserve (CFR). CFR is the ratio of
stressed to resting myocardial blood flow, reflecting microvascular vasodilation in response to
increased metabolic demand. Impaired CFR in patients without apparent coronary artery disease
(CAD) is an indication of coronary microvascular dysfunction and is associated with adverse
cardiovascular outcomes [13], [14]. Likewise, because leg muscles rely on a large flow reserve for
their wide range of metabolic rates between rest and full activity, reduced leg perfusion reserve
may indicate the presence of microvascular dysfunction and contribute to limb ischemia ,
additionally /in addition to/ arterial occlusive disease. Therefore, it is assumed in this thesis that
microvascular dysfunction as a form of microvascular disease exists in limb ischemia, and the
overall goal is to develop an assessment scheme that is sensitive to microvascular dysfunction. The
following section will review the structure and function of the microcirculation, from which a
proper assessment approach will be derived.
1.2.2 Structure and function of the microcirculation
A microvascular network is composed of arterioles, capillaries, and venules. The arterioles and
venules conduct blood toward and away from the capillaries, respectively. The capillaries are
channels of 5-10 µm in diameter with a single layer of endothelial cells forming the thin wall,
allowing gas, nutrients, and metabolic wastes to pass through. The delicate capillaries are protected
7
from arterial pulsatility and change of systemic pressure by the arterioles. The arterioles are
referred to as resistance vessels because, in the body overall, they account for the most significant
portion of flow resistance (see Figure 1.2a) when compared to other vascular segments. The
relationship between blood flow and resistance in laminar flow is described by Ohm’s law and the
Hagen–Poiseuille equation:
∆PQ
= R=8µLπr4 , ( 1.1 )
where ∆P denotes the pressure drop, Q blood flow, R resistance, µ viscosity, L vessel length, and
r vessel radius. A change of arteriole radius has a large impact on blood flow. Taking muscle
activation during exercise as an example, doubling the average arteriolar radius of a microvascular
network would result in an increase of perfusion to16 times, provided that the arterial input
pressure to the network remains the same. Hence, the arterioles play a crucial role in perfusion
regulation to the tissue.
Figure 1.2: Pressure distribution and arteriole structure. (a) The arterioles account for the largest portion of flow
resistance in the circulation system. (b) The three layers of the arteriolar wall. Images were adapted from Betts’paper
[15] under Creative Commons license.
The wall of an arteriole has three layers starting from the lumen and moving out: tunica intima,
tunica media, and tunica adventitia (see Figure 1.2b). The vascular smooth muscle of the media
can relax or contract, which changes the vessel wall tension, leading to vasodilation or
vasoconstriction. The relationship between the wall tension T and the luminal radius r is described
by Laplace's law:
Tunica adventitia
Tunica media
Tunica intima
a bAorta and arteries Arterioles Capillaries Venules and veins
20
40
60
80
100
120
Blo
od p
ress
ure
(mm
Hg) Systolic pressure
Mean arterial pressure
Diastolic pressure
8
T=(Pin-Pout)r , ( 1.2 )
where Pin and Pout are the pressure inside and outside of the vessel, respectively. On the other hand,
biomechanical studies have shown that the wall tension is a function of circumferential length and
can be decomposed into an active muscular component Tm and a passive elastic component Te (see
Figure 1.3) [16].
The wall tension T is the sum of the passive component and the active component scaled by the
activation level of the vascular smooth muscle A between 0 and 1:
T=Te+A�Tm . ( 1.3 )
The intima does not contribute significantly to the biomechanical properties of the vessel.
However, the endothelium mediates the tone of vascular smooth muscle and plays a crucial role
in flow regulation. Flow regulation is achieved essentially by the vascular smooth muscle changing
the activation level A in response to a variety of physiological stimuli.
The precise radius of an arteriole at any given moment is determined by the flow regulatory signals
originated from the vessel wall itself as well as the surrounding tissues. A few regulatory
mechanisms are introduced here to help understand the physiology behind the techniques of
microvascular assessment. Myogenic response describes the reactive contraction of the vascular
Figure 1.3: Vessel wall tension. Left: The solid curves show passive tension Te, active tension Tm, and total tension
with maximal activation Te+Tm. Dashed curves show active tension and total tension with half activation with A=0.5.
Dots show typical observed behavior during increasing activation. Right: Data points of activation and tension, fitted
with a sigmoid function. The plots are from Carlson’s paper [16].
9
smooth muscle when arterioles are stretched in situations such as when luminal pressure is changed
by body posture [17]. An increase in blood flow shear stress exerted on the endothelium of the
intima will stimulate the production of nitric oxide (NO) molecules, which diffuse into the media
and cause shear stress-induced vasodilation. Myogenic response and shear stress-induced
vasodilation are wall-derived intrinsic properties of the vessel. On the other hand, metabolite-
derived mechanisms usually involve vasoactive substances released in the blood or extravascular
tissues. The endothelial cell membrane has a variety of receptors for vasoactive substances,
including acetylcholine, ATP, adenosine, histamine, and so on [18]. A transient drop of oxygen
saturation in red blood cells will cause the release of ATP molecules, which bind to the receptors
on the endothelium and induce vasodilation [19]. In the case of exercise or prolonged hypoxia,
accumulation of adenosine in the interstitial space of skeletal muscles becomes the main reason
for sustained vasodilation [20]. These mechanisms will be incorporated into a model of
microcirculation behavior in Chapter 3.
Microvascular dysfunction is reflected in reduced vasodilatory capacity, indicating impairment in
the regulatory mechanisms described above. Assessment of microvascular dysfunction requires a
test of vascular reactivity by perturbing one of these mechanisms. The following section will
discuss perturbations and measurements.
1.2.3 Assessment of microvascular dysfunction
Perturbations
Exercise and ischemia-reperfusion paradigms are the two commonly used perturbations for
assessing skeletal muscle flow reserve. These methods stress the leg muscle and accumulate
metabolites to induce hyperemia (see Figure 1.4). An exercise such as walking or plantar flexion
will induce active hyperemia, the level of which reflects the maximum functional capacity of
vasodilation in the calf muscles. However, an exercise can only perturb the microvascular network
of the engaged muscle groups. The dependence of workload on subjects’ voluntary effort may
cause variability in perfusion measurements unless a quantitative control scheme is applied. In
addition to these limitations, performing controlled leg exercise in an MR scanner can post
challenges in both instrumentation and imaging. On the other hand, the ischemia-reperfusion
10
paradigm uses a blood pressure cuff to temporarily block arterial flow to induce reactive hyperemia
following flow restoration. Unlike an exercise paradigm that establishes a steady-state active
hyperemia which decays slowly, reactive hyperemia in skeletal muscles tends to rise and attenuate
quickly (in tens of seconds) creating challenges for measurement and analysis. The dynamic
response involves complex vascular physiology which has not been clearly understood.
In either paradigms, a standard protocol to probe microvascular dysfunction does not exist. I
choose to use reactive hyperemia for reasons of reduced instrumentation requirements and greater
ease of controlling the amount of stress applied to the subjects, hence focusing efforts on
measurement, analysis, and physiological interpretation of the acquired data.
Figure 1.4: Skeletal muscle hyperemia. (a) Active hyperemia induced by exercise. (b) Reactive hyperemia induced
by ischemia.
Flow-mediated dilatation
The ischemia-reperfusion paradigm has been widely adopted in flow-mediated dilatation (FMD).
FMD uses ultrasound to measure the diameter change of the brachial artery of the arm in response
to ischemia. The procedure typically involves a 5-minute cuffing of the upper arm, causing
ischemia-induced vasodilation of the lower arm. Following cuff release, the resulting reactive
hyperemia further causes brachial artery dilation through the shear stress-mediated NO pathway,
a hallmark of endothelial function. Reduced FMD is attributed to systemic endothelial dysfunction,
indicating deteriorated overall vascular health and higher risk for cardiovascular disease [21].
While FMD is a secondary response to reactive hyperemia, the primary reactive hyperemia itself
represents the microvascular function of the arm microvasculature [22]. However, arm-specific
11
microvascular assessment is of limited interest, perhaps because the arm circulation is rarely a
target of occlusive disease or perfusion deficiency. What can be learned from FMD, however, is
that 5 minutes of ischemia seems to be a favorable protocol for generating a large response with
stable reproducibility [23].
Cardiac stress test
In the cardiac stress test for coronary flow reserve (CFR), coronary vasodilation is induced by
intravenous infusion of pharmacological agents, such as adenosine, dipyridamole, or dobutamine
[24]. Pharmacological perturbation can establish a steady-state hyperemia in a dose-controlled
manner, which is advantageous for measurement of flow/perfusion reserve. Plenty of drugs can
induce hyperemia effectively when infused directly to the leg arteries, which requires canalization
of the femoral artery. However, no drug can be infused intravenously to induce significant and leg-
specific hyperemia. Therefore, this project does not use drug perturbation in the study subjects.
CFR can be measured invasively with a coronary pressure wire or noninvasively with myocardial
perfusion imaging using positron emission tomography (PET) or MRI. An early study using
invasive CFR has shown that early deterioration of microvascular function is associated with
coronary restenosis in patients treated with balloon angioplasty [25]. Recent PET studies
demonstrated that CFR is associated with cardiovascular outcomes independently of angiographic
coronary artery disease (CAD) [13], [14]. The goal of this thesis shares the same concept as these
CFR studies and should hold promise. However, it is recognized that both poor coronary inflow
and coronary microvascular dysfunction can reduce CFR. Only in non-CAD patients or after
revascularization can reduced CFR be attributed to microvascular dysfunction. Therefore, when
applying a limb stress test to PAD, the influence of arterial occlusive disease may dominate the
response. In other words, the sensitivity of a limb stress test to microvascular dysfunction is
unclear.
1.2.4 Noninvasive measurement of peripheral microcirculation
There are a few techniques to measure peripheral microcirculation noninvasively. Laser Doppler
flowmetry (LDF) and transcutaneous oximetry (TcPO2) utilize optics to measure the blood flow
12
and oxygen tension of the skin, respectively [26]. Both techniques are cheap and easy to use, but
they only probe the skin. A conventional method to estimate the perfusion in leg muscles is the
venous occlusion plethysmography (VOP), which measures the rate of limb volume change when
the venous outflow is temporarily blocked, attributing the volume increase to arterial inflow [27].
VOP has a poor temporal resolution because it requires periodic inflation and deflation of a
pressure cuff. Imaging techniques, including radioactive tracer nuclear imaging [28], [29],
contrast-enhanced and noncontrast MRI, and microbubble ultrasound [30], [31], are favorable
modalities to look at skeletal muscle dynamic perfusion specifically. This thesis focuses on an
MRI-based approach, but relevant findings of peripheral microvascular assessment using nuclear
imaging and microbubble ultrasound will also be referenced.
In summary, Section 1.2 explained that microvascular assessment involves a scheme to probe
microvascular dysfunction in leg muscles. Reactive hyperemia using an ischemia-reperfusion
paradigm is preferred over other perturbations, but challenges in measurement, analysis, and
interpretation are expected. The requirement to specifically measure skeletal muscle perfusion
leads to the use of imaging techniques, particularly MRI.
1.3 MRI for peripheral microvascular assessment
This section will briefly introduce the principle of MRI and the technique of arterial spin labeling
(ASL), followed by a summary of the findings of other peripheral ASL studies. Also, two
microvascular assessment techniques alternative to ASL, i.e., blood oxygenation-level dependent
(BOLD) and dynamic contrast-enhanced (DCE) imaging, will be reviewed in a sub-section.
1.3.1 MRI physics
Nuclear spins of protons in the human body can be seen as microscopic magnets, which have
random orientations. When placed in the strong magnetic field created by an MR scanner, these
tiny magnets become more aligned to the field, resulting in a non-zero net sum macroscopically
called magnetization. When radio-frequency (RF) excitation pulses are applied, the magnetization
will rotate by a flip angle about the main magnetic field in the longitudinal direction. The
13
longitudinal component of the magnetization will recover to the unexcited strength with a time
constant of T1, following a process called T1 relaxation. On the other hand, the transverse
component will decay with a time constant of T2, following a process called T2 relaxation. Imaging
involves excitation and relaxation of the magnetization, followed by a readout process that
measures the signal of the transverse magnetization at each location of the field of view in a
spatially encoded way. The whole imaging process is programmed into a pulse sequence.
Optimization of a pulse sequence is usually to improve the signal-to-noise ratio (SNR) and to
create desired contrast between different tissues as efficiently as possible.
1.3.2 Arterial spin labeling
Arterial spin labeling (ASL) is an MRI-based technique that measures perfusion without using any
contrast agent, applied first in the brain [32], [33] but later also in leg muscles during exercise [34].
The idea of ASL is to measure the difference signal between two acquisitions, one with
magnetically labeled blood and the other with unperturbed blood. In the labeled acquisition, the
magnetization of the upstream blood is inverted by labeling pulses. After a period of transit time
for the labeled blood to wash into the downstream tissue, the tissue is imaged. In the control
acquisition, the same imaging is repeated except that the upstream blood is not inverted.
Subtracting the two images will cancel the signal of static tissue, resulting in a difference
proportional to the amount of blood moving into the microvascular bed in a given time, which
corresponds to perfusion. In other words, ASL measures absolute perfusion and does not require
exogenous contrast agents. Moreover, ASL measurement can be repeated to resolve the dynamic
perfusion changes in response to perturbation, which makes ASL signals more straightforward
than DCE and BOLD to be related to microvascular function.
Recently, a few research groups have used different ASL sequences to characterize reactive
hyperemia in healthy subjects and patients with PAD [35]-[37]. The studies consistently showed
that PAD patients had blunted responses, exhibiting lower and delayed response peaks when
compared to those of the healthy subjects. Augmented reactive hyperemia after revascularization
was also reported [38]. However, the amplitude and temporal characteristics of the responses were
very different across the studies. Since ASL is an absolute measurement of perfusion, the same
population should have similar response characteristics. Therefore, the usability of ASL remained
14
to be improved. As will be explained in Chapter 2, the potential sources of discrepancies may be
associated with the adequacy of different ASL sequences for tibial perfusion with ischemic
perturbations.
The major disadvantage of ASL is that the perfusion signal-to-noise ratio (SNR) is very low, due
to the biological nature that the blood volume is small. In resting skeletal muscle, roughly 3% of
the volume is blood in the microvascular network, and 97% of the volume consists of muscle, fat,
interstitial fluid, and connective tissue [39]. Therefore, other MRI techniques with higher SNR
have been used to study peripheral microcirculation. The next sub-section will describe these
techniques and their uses in PAD.
1.3.3 DCE and BOLD imaging
In dynamic contrast-enhanced (DCE) MRI, images are acquired continuously as the bolus of
gadolinium-based contrast agents is injected and flows into the tissue. These contrast agents
shorten T1 and T2 relaxation, resulting in stronger magnetization signal and better tissue
visualization. Most contrast agents will leak into the tissue space from the vessels, and later be
cleared out of the body by the kidneys. When the contrast agent enters the calf perfusion territory,
the uptake rate depends on the arterial inflow, microvascular perfusion, and capillary permeability
to the agent. A standard way to calculate perfusion is to perform deconvolution of the time-signal
intensity with a simultaneously measured arterial input function. Isbell et al. proposed a
semiquantitative approach, which looks at the ratio of the slopes of the arterial blood to muscle
time-intensity curves, and applied it to assess peak-exercise perfusion in PAD [40]. Another way
to take out the dependence of perfusion on arterial inflow is to inject contrast agent when arterial
inflow is blocked, a method reported by Thompson et al [41]. The arterial input function
immediately after flow interruption was shown to be a step function, identical in all subjects. This
way, the tissue time-intensity curve can be directly compared without further calculation. No
patient study using this method was found. Overall, DCE-MRI offers strong image signal, but the
calculated perfusion is an estimation for the first-pass interval. The method is unavailable for
patients with significantly impaired renal function since standard MR contrast agents are contra-
indicated in this population.
15
The technique of blood oxygenation-level dependent (BOLD) imaging was first developed for
MRI of brain function. It relies on the principle that a change of oxygen saturation in blood
hemoglobin will affect local magnetic susceptibility, resulting in a change of the transverse
relaxation time constant, denoted as T2* [42]. This phenomenon has been widely utilized to detect
functional activity in the brain, but the same principle can be applied to image the activity of the
myocardium or skeletal muscle [43], [44]. Using the ischemia-reperfusion paradigm, Ledermann
et al. demonstrated that the calf muscle reactive hyperemic T2* time curve rose more slowly and
had a lower peak in patients with PAD [26], [45]. Recently, Bajwa et al. showed that the rate of
signal decay during flow interruption and the rate of signal increase during reperfusion of the T2*
time curve are more powerful than the peak or time-to-peak to stratify populations between young
healthy subjects, age-matched control subjects, and patients with CLI [46]. In summary, the BOLD
technique can generate stronger SNR than ASL and does not require contrast agents. However, it
is a relative measurement and not specific to perfusion. The dynamic T2* change is a combined
result of changes in blood flow, blood volume and oxygen extraction and consumption, which can
be an advantage or disadvantage depending on the purpose.
In summary, Section 1.3 explained that ASL measures absolute perfusion and is the more
straightforward technique for deriving peripheral microvascular function from perfusion
dynamics, but challenges associated with low SNR are expected. DCE-MRI and BOLD techniques
are proven feasible alternatives to detect perfusion differences between subject populations under
some circumstances.
1.4 Thesis overview
The goal of this project is to develop an assessment scheme that is sensitive to microvascular
dysfunction in limb ischemia. Specifically, the project aimed to optimize ASL perfusion
measurement, interpret the data in the context of pathophysiology, and then compare the findings
of changes in perfusion with clinical results in patients with PAD.
Although there are many approaches to estimate perfusion, a standard measurement of dynamic
perfusion in leg muscles has not been established. With ASL, the discrepancies in the measured
characteristics of reactive hyperemia across studies were rarely discussed and remained
16
unexplained. Moreover, the ability of ASL to show the microvascular function in regulating flow
for different metabolic demand had not been tested. Therefore, in Chapter 2, the pulsed ASL
technique was optimized and used to characterize reactive hyperemia in healthy subjects, who
underwent varying durations of cuffing, and in patients with PAD.
The pathophysiology of reduced reactive hyperemia responses in PAD was not clear. Developing
a theory to facilitate the interpretation of the changes in reactive hyperemia was necessary. In
Chapter 3, the data acquired in the previous chapter was utilized to establish a physiological model
of reactive hyperemia, incorporating oxygen transport, tissue metabolism, and vascular regulation
mechanisms involved in the ischemia-reperfusion paradigm. Based on the model, the influences
of arterial stenosis and microvascular dysfunction on reactive hyperemia were simulated. Studying
these effects led to a novel way to characterize reactive hyperemia using model-based
physiological measures.
Chapter 4 demonstrates the relevance of the developed ASL measures, including microvascular
measures, to the symptom severity, response to revascularization, and short-term outcome in
patients with PAD. In contrast, previous ASL studies focused on the detection of perfusion
differences between populations with known conditions of leg arteries.
Finally, Chapter 5 summarizes the project and discusses future research directions.
17
Chapter 2
ASL reactive hyperemia in calf muscles This chapter is adapted from the manuscript “Investigating discrepancies in arterial spin labeling
to improve characterization of calf muscle perfusion” authored by Hou-Jen Chen and Graham
Wright and submitted to NMR in Biomedicine in 2017. This chapter describes a study that was
conducted to improve the consistency in ASL measurement of calf reactive hyperemia, and also
to investigate the physiological effect of varying ischemic duration on the perfusion responses of
healthy subjects and the effect of large-vessel disease on the responses of patients with PAD.
2.1 Background
Reactive hyperemia is a rapid increase of perfusion due to vasodilation induced by metabolic
stresses during flow interruption (see Figure 2.1). The response can reach to more than 10 times
of the baseline level, and is commonly characterized by the peak and TTP (time to peak). In
patients with PAD, the reactive hyperemia may present a reduced peak and delayed TTP. The
blunted responses can be influenced by impaired vascular reactivity associated with microvascular
dysfunction, as well as reduced perfusion pressure associated with occlusive large-vessel disease.
Figure 2.1: Illustration of ASL reactive hyperemia induced by 2 minutes of flow interruption.
Recently, a few research groups have used ASL to characterize cuff-induced reactive hyperemia
in healthy subjects and patients with PAD [35]-[38], as summarized in Table 2.1. Although these
60 120 180 2400
10
20
30
40
50
Perfu
sion
(mL/
100g
/min
)
Time (s)
Rest Ischemia Reperfusion
Risingphase
Recovery phase
PeakTTP
18
studies detected differences between different subject groups, the amplitude and temporal
characteristics are not consistent across studies. Since ASL is an absolute measurement of
perfusion, the same population should have similar response characteristics. Therefore, the first
aim is to investigate the potential sources of discrepancies in previous ASL-based reactive
hyperemia studies.
Table 2.1: Reactive hyperemia in soleus muscle measured in previous ASL studies.
Peak perfusion (mL/100g/min) Time-to-peak (s)
Previous studies Healthy PAD Healthy PAD
Continuous ASL [35] 91 ± 39 92 ± 45 * 47 ± 12 64 ± 21 *
Pulsed ASL [37] 74.9 ± 33 28.9 ± 13.9 25.0 ± 13.0 72 ± 49
Pulsed ASL [36] 179 ± 60 81 ± 59 14 23
Pre-PTA Post-PTA Pre-PTA Post-PTA
Pseudo-continuous ASL [38] 74 ± 52 129 ± 80 59 ± 29 41 ± 14
PTA is percutaneous transluminal angioplasty. * denotes combining data of all patient categories in the study [35]. The soleus muscle is compared because it is the region all the listed studies had quantified.
2.1.1 Technical considerations in ASL
In ASL, the magnetization of arterial blood is inverted as it flows from an upstream labeling region
into an imaging region where images are acquired after a post-labeling delay (PLD). The PLD
accommodates the transit time of the labeled blood to travel from the labeling region to the imaging
plane. This labeled scan is alternated with a control scan in which the arterial magnetization is not
inverted. Subtracting the labeled from the control images should cancel out the static tissue signal,
leaving only the blood signal with the difference created by blood labeling, which is proportional
to perfusion [47].
The (pseudo-) continuous labeling and pulsed labeling are reviewed here in order to establish an
optimized ASL solution. In continuous labeling, a labeling pulse irradiates the spins of the blood
that are flowing through the labeling region continuously for a few seconds, generating a velocity-
driven inversion of blood magnetization (Figure 2.2). The labeled blood travels into the
19
microvasculature a few centimeters downstream where the image is acquired. A pseudo-
continuous approach replaces the long labeling pulse with a train of short pulses, otherwise being
identical to the continuous labeling. On the other hand, in pulsed labeling, an RF pulse irradiates
a volume of blood upstream of the image plane within 10-20 milliseconds, a process independent
of blood velocity. If the distance for the labeled blood to travel to the tissue is the same, which
implies identical transit time, continuous labeling can provide a stronger perfusion signal than
pulsed labeling because a portion of the blood is labeled at later time point and undergoes less T1
decay. In practice, however, the geometry of the vascular tree and the blood velocity can affect the
transit time, which affects the choice for optimal configuration of ASL. While the leg vessels have
a simple geometry that goes in the superior-inferior direction, the blood velocity is a major concern
that will be studied in this chapter.
Figure 2.2: Illustration of the configuration and difference signal in continuous and pulsed labeling. (a) A
labeling plane in continuous labeling, prescribed perpendicular to the popliteal artery, is separated from the imaging
plane by a transit distance. The distance is assumed to require 2 s of transit time, corresponding to a proper PLD of 2
20
s. (b) When the image is acquired after 1.5 s of labeling and 2 s of PLD, the difference signal is contributed by the
labeled blood with 2 to 3.5 s of relaxation at T1=1.6 s at 3 Tesla. (c) The volume of labeling region in pulsed labeling
and imaging plane are separated by a narrow gap, which is assumed to require a transit time of 0.7 s based on my
experiments. (d) When the image is acquired at 1.4 s of PLD, the difference signal is contributed by 0.7 s of labeled
blood with 1.4 s of T1 relaxation.
With a stress test of peripheral muscles, there are particular technical considerations. First, the
signal of static tissue varies over the ischemia-reperfusion interval due to changes of oxygenation
and cannot be canceled out by direct pairwise subtraction of labeled scans taken at different times
from the control. Second, the wide dynamic range of blood velocities between resting and
hyperemia implies variable time for the transit time and refreshment of the labeling region. Lastly,
in (pseudo-) continuous schemes the labeling efficiency may be affected by the unique blood
velocity profile in the tibial region, which is triphasic and relatively slow [3]. To examine if the
previous ASL sequence settings were adequate in the considered context, first, two commonly
used subtraction schemes for removal of background tissue signal were compared in pseudo-
continuous ASL (pCASL). Second, tibial blood velocity was measured with phase-contrast
imaging across ischemia-reperfusion to estimate a reasonable repetition rate of ASL. Third,
pseudo-continuous labeling of popliteal triphasic flow was studied using a Bloch simulation. These
experiments would clarify the factors affecting ASL signal behavior. Based on the findings, more
robust ASL parameter settings were then used to characterize the perfusion responses in PAD
patients and in healthy subjects who underwent varying durations of cuffing.
2.1.2 Physiological influences on perfusion characteristics
The second aim is to investigate the physiological influences, including the duration of ischemia
and presence of PAD, on the characteristics of reactive hyperemia. Although the protocol of 5-
minute ischemia has been commonly used to generate a strong and prolonged hyperemic response
for conventional flow measurement, associated patient discomfort may lead to premature
termination of the test [38], particularly in those who already suffer from critical limb ischemia.
Studying the effect of shorter ischemic durations may lead to a more tolerable stress protocol.
More importantly, it is known that ischemic duration is an important determinant of the
mechanisms responsible for the subsequent vasodilator responses [22]. Therefore, a study of
ischemic duration may reveal potential application of ASL for assessment of microvascular
21
dysfunction associated with particular vasodilatory. In addition, the feasibility of characterizing
ASL reactive hyperemia induced by 2-min ischemia in patients with PAD was tested, and the
results were compared to the responses of the healthy subjects.
2.2 General methods
This section describes the subjects and common experimental settings. The detailed method of
each experiment will be described in the corresponding sections.
The informed consent of the participants was obtained following a protocol approved by the
Institutional Research Ethics Board. The adequacy of pCASL in calf perfusion was tested through
a series of technical experiments described in Section 2.3, which only involved healthy subjects
and 3-minutes of ischemia. The healthy group consisted of seven subjects (< 30 years old) without
known cardiovascular disease. Five subjects participated in the study of pCASL response, and four
participated in the measurement of velocity response.
In the study of physiological influences in Section 2.4, flow-sensitive alternating inversion
recovery (FAIR, a version of pulsed ASL) was used to characterize responses induced by varying
ischemic durations in the healthy group and responses in a group of patients. The patient group
included nine subjects (71.8 ± 10.4 years old, ABI = 0.62 ± 0.19, two males) with diagnosed PAD
based on their symptoms and vascular lab findings. Seven patients had claudication, and the
remaining two had rest pain. Patients with vascular stents in the thigh and other MRI
contraindication were excluded.
Imaging was performed at 3 Tesla (MR 750 by General Electric, Milwaukee, Wisconsin) with an
8-channel cardiac receive array placed in the calf region. Subjects were in a foot-first supine
position and rested for 5 minutes before imaging started. Flow interruption was achieved by thigh
compression with a 12 cm long air cuff, which was inflated to 220 mmHg within 10 s. The cuff
was deflated within 3 s by opening the air valve after ischemic durations were reached. The healthy
group underwent multiple sessions of imaging on different days, each session involving repeated
ischemia separated by at least 15 minutes of rest. The images were reconstructed and processed in
MATLAB (MathWorks, Natick, MA).
22
Statistical analysis was performed in Prism (GraphPad Software, San Diego, CA) and Stata
(College Station, TX). Data are presented as means ± SD. All data were assumed normally
distributed as the Shapiro–Wilk test did not indicate otherwise. Significance was considered at p
<0.05.
23
2.3 Technical experiments of peripheral ASL
2.3.1 Background artifact
Methods
A multi-slice version of the pCASL sequence described by Dai et al was used to record perfusion
responses to 3 minutes of ischemia [48]. The purpose of imaging multiple slices was to examine
if the labeled blood could create location-dependent perfusion signals with respect to the velocity
of blood flow. As shown in Figure 2.3, the labeling plane of pCASL was prescribed to be
perpendicular to the popliteal artery below the knee and 6 cm proximal to the widest part of the
mid-calf.
Figure 2.3: (a) The labeling plane (red solid line), measurement of labeled arterial blood (purple dashed line), and the
upper, middle, and lower slice of perfusion images (white dashed lines) in the pCASL experiments were indicated. (b)
The region-of-interest (ROI) on the perfusion images enclosed all muscles of the whole mid-calf cross section. The
voxels affected by the arterial blood were detected by thresholding the temporal standard deviation of the ASL
difference images (c) and excluded from the ROI.
The pCASL acquisition started 30 s before flow interruption and continued to a time 2.5 min after
deflation of the cuff. The parameters were set to mimic the previous continuous ASL study [35],
with labeling duration = 2 s, PLD = 1.9 s, single-shot gradient-echo echo-planer imaging (EPI)
with flip angle = 90°, field of view = 16×16 cm2, in-plane matrix size = 64×64, TE/TR = 17/4000
ms, and slice thickness = 1 cm, prescribed at 4, 6, and 8 cm distal to the labeling plane (Figure
2.3a). High-resolution reference images corresponding to the location of pCASL images were
acquired [35].
24
The region of interest (ROI) was manually drawn based on the reference images to cover the entire
muscle region (Figure 2.3b). The arterial voxels were identified by their high temporal standard
deviation of ASL images (Figure 2.3c) and excluded from the ROI. In some cases, cuff inflation
caused a vertical shift of more than one voxel, which was corrected by shifting the images during
ischemia to align with the rest of the images. Cases with displacement smaller than one voxel were
not corrected. The data points experiencing cuff inflation and deflation were plotted as zeros and
excluded from quantification.
Pairwise subtraction and time-matched subtraction were compared in the pCASL dataset. These
subtraction methods were used in different studies previously [35], [49]; reproducing the
subtraction methods for the same recorded data may explain the discrepancies in the literature. In
the pairwise subtraction, the labeled images were subtracted from the control images acquired at
the later time points. The difference images were normalized by their corresponding control
images. To reproduce the quantification process of the previous study [35], every two of the
normalized difference signal data points were averaged, resulting in temporal resolution of 16 s.
In time-matched subtraction, adjacent control data were averaged to yield a control series
temporally matched with the labeled series before subtraction was performed.
Result
Flow interruption induced considerable dynamic signal changes (Figure 2.4a), including a marked
decrease in the early phase of ischemia and a rapid increase at the beginning of the reperfusion,
followed by a slow attenuation. The control and labeled repetitions were sampled alternatingly
along the time course of varying signal intensity. With direct pairwise subtraction (Figure 2.4b),
the time courses of the difference signal at the three locations of the calf were similar. In other
words, the labeled blood did not seem to perfuse any particular location downstream. In addition,
strong negative signals occurred in the early phase of ischemia. With time-matched subtraction
(Figure 2.4c), the difference signals were much smaller overall, and only the slice 6 cm distal to
the labeling plane demonstrated a rapid increase resembling reactive hyperemia. Attempt to
quantify perfusion from pairwise and time-matched subtraction methods led to distinct perfusion
time courses, with the difference most noticeable in the magnitude and time of the peak (Figure
2.4d). These results indicated that a large component of the difference signal in pairwise
subtraction was contributed by the time derivative of the static tissue signal instead of the labeled
blood.
25
Figure 2.4: Sanity check of pCASL in a healthy subject. (a) The time courses of the ROI signal intensity, (b) the
difference signal by averaged pairwise subtraction, (c) the difference signal by time-matched subtraction, and (d) the
middle slice perfusion responses calculated from the difference signals of the two subtraction schemes. The conversion
from pCASL signal to perfusion was performed following Wu’s paper.
These signal characteristics were consistently observed in all subjects and the two subtraction
schemes differed significantly (Table 2.2; two-sided paired Student’s t-tests.). Averaged pairwise
subtraction resulted in a response peak roughly 3 times larger and a slightly delayed TTP when
compared with time-matched subtraction, as well as a small negative difference in the later phase
of recovery. These results indicated that a strategy to minimize the contamination, such as using
time-matched subtraction, could lead to distinct results in the quantification of perfusion.
26
Table 2.2: Comparing subtraction schemes (n=5).
Response characteristics Averaged pairwise Time-matched p value
Peak signal (%) 3.34 ± 0.46 0.68 ± 0.16 0.0002
TTP (s) 16.8 ± 5.82 11.4 ± 6.31 0.0309
1 min post-ischemia signal (%) -0.19 ± 0.10 0.05 ± 0.02 0.0059
2.3.2 Tibial blood velocity
Methods
The tibial velocity profiles were recorded using single-slice cardiac-gated phase-contrast imaging.
The image slice was prescribed at 2 cm proximal to the widest part of the mid-calf to estimate the
inflow velocity. Two subjects had the anterior and posterior tibial artery at the location, whereas
the other two subjects also had the peroneal artery there. The recordings were repeated every
minute during the 3-min ischemia period and continued every 30 s for 2 min after deflation of the
air cuff. The scan parameters were field of view = 16×16 cm2, matrix = 160×160, flip angle = 25°,
TR/TE = 8.5/3.7 ms, and slice thickness = 0.5 cm. Only velocity in the superior-inferior direction
was measured, with velocity encoding (VENC) = 80 cm/s, resolved cardiac phases = 20, and views
per segment = 16. ROIs of the major arteries were manually drawn.
Results
In the tibial region, the cardiac cycle-averaged velocity well below 10 cm/s at rest was consistently
observed (Table 2.3). The velocities increased to between 10 and 20 cm/s during hyperemia. All
subjects exhibited triphasic flow at rest, with velocity in the range between -5 and 15 cm/s. During
the early post-cuffing period the retrograde phase vanished, and maximal systolic velocity reached
to a range between 17.0 and 36.6 cm/s in the subject group.
Table 2.3: Tibial blood velocity throughout the ischemia-reperfusion paradigm (n=4).
Blood velocity (cm/s) Baseline Ischemia Hyperemia Recovery
Anterior tibial artery 2.4 ± 1.3 0.4 ± 0.8 11.4 ± 7.5 2.3 ± 1.0
Posterior tibial artery 4.1 ± 1.4 0.9 ± 0.6 16.4 ± 5.8 3.3 ± 0.8
27
2.3.3 Simulation of pseudo-continuous labeling
The labeling efficiency is a scaling factor in the quantification of ASL perfusion. The previous
continuous ASL study [35] assumed a constant labeling efficiency of 0.9 throughout the ischemia-
reperfusion process. However, given the low mean velocity and triphasic profile observed in the
last sub-section, the labeling efficiency may be affected. Here, the labeling efficiency in pCASL
of the calf is investigated.
Methods
The resting popliteal velocity profile at the labeling plane, recorded in one subject using
aforementioned phase-contrast imaging, was used for simulation of pCASL labeling. The Bloch
simulator was implemented in MATLAB by modifying a simulator created by Dr. Brian
Hargreaves (www-mrsrl.stanford.edu/~brian/blochsim/). The simulation of adiabatic velocity-
driven inversion used a fine time step of 4 µs and T1B/T2B = 1600/275 ms for the blood relaxation.
At each time step the position of the spins was calculated from the interpolated velocity waveform,
whereas the RF and gradient waveform were retrieved from the pCASL sequence files. The
simulated result was compared with direct measurement of popliteal arterial signal intensity
acquired with pCASL for 10 pairs of labeled and control images, with the imaging plane prescribed
at 2 cm distal to the labeling plane and PLD = 100 ms.
Results
A popliteal velocity waveform recorded at rest (Figure 2.5a) was used as input for simulation of
pCASL labeling. The result in Figure 2.5b shows that effective inversion occurred initially when
the blood speed is above 5 cm/s independent of direction, but the change of flow direction in the
triphasic flow can reverse the inversion of magnetization. Depending on the cardiac phase at which
the labeling started, the degree of inversion varied significantly. The result of the pCASL
simulation (Figure 2.5c) indicated that the inversion in the labeled scan was highly variable due to
the triphasic flow characteristics, while in the control scan the magnetization remained
unperturbed. The variability of pCASL labeling measured at 2 cm below the labeling plane (Figure
2.5d) agreed with labeling variability seen in simulations.
28
Figure 2.5: Bloch simulation of velocity-driven adiabatic inversion. (a) The popliteal blood velocity demonstrated
a triphasic profile. The colored dots on the velocity profile indicated 5 different starting velocities of spins entering
the labeling plane, resulting in different degrees of inverted magnetization shown in (b). The longitudinal
magnetization (Mz) at 2 cm downstream of the labeling plane was simulated with the velocity profile broken into finer
steps (interpolated 10 times), shown in (c) where the edges of the boxes indicate the 25th and 75th percentiles and the
central lines indicate the medians. (d) The measured signal intensity of arterial pixels in pCASL.
29
2.4 Physiological influences on perfusion characteristics
Methods
All seven healthy subjects underwent 2, 3, and 5 min of cuff-induced ischemia. Five of these
subjects also participated in a 1-min ischemia experiment. The ischemic durations were applied in
random order and separated into two sessions on different days. The parameters of the FAIR
sequence to record reactive hyperemia included single-slice steady-state free precession
acquisition [50], with TE/TR = 1.7/3.8 ms, flip angle = 70°, and PLD = 1.4 s. The acquisition was
repeated every 3.35 s, starting 30 s before deflation of the air-cuff and continuing for at least 2 min.
Baseline perfusion at rest was recorded for 2 min before any stress on the first day. The field of
view, resolution, and slice thickness were identical to the settings in pCASL. Q2TIPS (Quantitative
Imaging of Perfusion Using a Single Subtraction II with thin-slice TI1 periodic saturation) was
applied to reduce the sensitivity to blood transit time, with the train of saturation pulses starting at
TI1=700 ms and ending 100 ms before image acquisition [51]. To suppress background signal, a
saturation pulse was applied at the imaging plane prior to labeling, followed by two non-selective
inversion pulses at 560 and 1200 ms after the saturation pulse. These time points were chosen to
achieve nulling for the largest range of T1 close to skeletal muscle based on theoretical calculation
of inversion recovery. Motion correction and determination of ROI were the same as described in
the pCASL experiment. To minimize the artifact of interleaved acquisition, both the label and
control series were linearly interpolated, followed by subtraction between time-matched label and
control data [52]. The difference signal was corrected for background suppression by multiplying
the difference signal by (1.06)2 [53]. The resulting difference signal ∆M was converted to perfusion
f :
f t =
λ2TI1
∆M(t)M0B
e-PLD/T1B ( 2.1 )
where M0B is the signal of fully relaxed blood, which was very close to the baseline signal of
muscle tissue acquired with SSFP before stress [50], and λ is the blood volume partition coefficient
assumed to be 0.9 mL/g.
Reactive hyperemia was characterized by the peak perfusion, time to peak perfusion (TTP), time
to recovery (TTR), and hyperemic flow volume (HFV). TTR was defined as the time perfusion
30
fell below 15mL/100g/min. HFV was the time integral of reactive hyperemia within TTR. The
perfusion curves were not processed by smoothing, scaling, shifting, or other manipulations.
Five healthy subjects participated in a repeatability session, where cuff experiments with 2-min
ischemia was conducted twice separated by 15 minutes of rest. The intraclass correlation
coefficient (ICC) and within-subject coefficient of variation (CV) were calculated.
Patient study
The patients with PAD underwent 2-min ischemia once for the more symptomatic leg. Calf
perfusion responses were acquired and characterized the same way as described in the study of
ischemic duration. The ASL sequence was started 1 minute before cuff inflation and repeated for
51 pairs of labeled and control images over 5 minutes and 41 seconds.
Statistical analysis
One-way analysis of variance (ANOVA) and Tukey's multiple comparisons test were used to
compare the effects of ischemic duration in the subjects who participated in all four durations of
ischemia. Two-sided equal variance Student’s t-test was used to evaluate the differences between
the healthy group and PAD. All data were assumed normally distributed as the Shapiro–Wilk test
did not indicate otherwise. Significance was considered at p <0.05.
Results
In the group participating in the experiment of varying ischemic duration, the baseline perfusion
ranged from 2.4 to 7.0 mL/100g/min, with a mean ± standard error of 4.8 ± 1.8 mL/100g/min. As
shown in Figure 2.6, the ischemic duration had the most impact on the temporal extent of the
response and also slightly enhanced the response magnitude when the ischemia duration was
increased. The rising phase of the responses had a similar slope regardless of ischemic duration. A
transient reactive hyperemia could be induced by just 1 min of flow interruption, but the peak was
more variable in the 1-min ischemia than in other longer durations of ischemia. All four
characteristics, including the peak, TTP, TTR, and HFV, were sensitive to the change of ischemic
duration (Table 2.4; ANOVA, p<0.05 for all characteristics), with the HFV being the most
sensitive and the peak the least sensitive. In the post-hoc test, the peak did not have a significant
31
difference between any pair of ischemic duration. 1-min and 3-min ischemia significantly differed
in TTP and HFV; 1-min and 5-min ischemia significantly differed in TTP, TTR, and HFV; 2-min
and 3-min ischemia significantly differed in TTP; 2-min and 5-min ischemia significantly differed
in TTR and HFV; 3-min and 5-min ischemia were not significantly different in any characteristic.
Figure 2.6: Reactive hyperemia induced by various durations of ischemia and recorded with FAIR.
0 20 40 60 80 1000
204060
Time (s)
Perfu
sion
(mL/
100g
/min
) 1−min ischemia (n=5)
0 20 40 60 80 1000
204060
Time (s)
Perfu
sion
(mL/
100g
/min
) 2−min ischemia (n=7)
0 20 40 60 80 1000
204060
Time (s)
Perfu
sion
(mL/
100g
/min
) 3−min ischemia (n=7)
0 20 40 60 80 1000
204060
Time (s)
Perfu
sion
(mL/
100g
/min
) 5−min ischemia (n=7)
32
Table 2.4: Characteristics of reactive hyperemia induced by different durations of ischemia.
Characteristics 1 minute (n=5)
2 minutes (n=7)
3 minutes (n=7)
5 minutes (n=7)
p value
Peak (mL/100g/min) 41.2 ± 14.1 44.7 ± 6.8 48.9 ± 8.1 50.2 ± 9.1 0.031
TTP (s) 5.6 ± 1.6 8.3 ± 1.7 15.0 ± 5.6 †§ 14.8 ± 3.2 § 0.002
TTR (s) 16.6 ± 4.8 20.2 ± 4.5 29.7 ± 8.9 42.3 ± 9.7 †§ 0.0015
HFV (mL/100g) 5.3 ± 2.5 8.2 ± 2.9 13.7 ± 5.0 † 20.2 ± 4.7 †§ 0.0006
† Significantly different from 1-minute ischemia § Significantly different from 2-minute ischemia
The repeatability assessment of 2-min ischemia showed that the peak perfusion was the most
repeatable characteristics (Table 2.5). All four characteristics had CV<25%. However, the ICCs
for TTR and HFV were relatively low.
Table 2.5: Intrasession repeatability of 2-min ischemia (n=5).
Characteristics First scan Second scan ICC CV
Peak (mL/100g/min) 40.1 ± 8.5 42.3 ± 7.2 0.92 5.6%
TTP (s) 10.5 ± 5.4 10.9 ± 6.7 0.93 15.1%
TTR (s) 24.5 ± 6.9 23.9 ± 7.3 0.35 22.8%
HFV (mL/100g) 10.8 ± 3.7 10.8 ± 3.8 0.51 23.3%
ICC intraclass correlation coefficient; CV mean within-subject coefficient of variation.
All patients completed the exam of 2-min ischemia without complaint. As an example shown in
Figure 2.7a, hyperemia was appreciable in PAD, but the magnitude was reduced when compared
to that in the healthy subjects. The time course of reactive hyperemia shown in Figure 2.7b
indicated that the peak perfusion was reduced in PAD, along with an extension of the time to
recover to the baseline. Only the peak perfusion was significantly different between the healthy
and PAD group (Figure 2.7; t-test, p<0.0001). TTP and TTR appeared longer in the patient group
than in the healthy group (p=0.1019 and 0.1652 for TTP and TTR, respectively), but these two
characteristics had a wide spread among the patient group. HFV was similar between the healthy
and patient group (p=0.9562).
33
Figure 2.7: (a) Baseline, (b) ischemia, and (c) hyperemia perfusion maps of a patient and (d) a hyperemia perfusion
map of a healthy subject with 2-min ischemia. Each perfusion map represents an average of 5 consecutive difference
images; hence the hyperemia maps represents the averaged perfusion of the first 33 s after ischemia. (e) The time
course mean and standard deviation and (f) characteristics of reactive hyperemia in the healthy and patient group,
respectively. * indicates p<0.0001.
34
2.5 Discussion
Peripheral dynamic perfusion may provide more information about the complex pathophysiology
in PAD than assessment of large vessels alone. However, two areas remain to be improved before
ASL reactive hyperemia can be more useful for clinical research of limb ischemia. The first area
is the standardization of ASL measurements. Comparisons across previous studies revealed that
an investigation of the inconsistent results for ASL reactive hyperemia was necessary. The second
area is the understanding of how physiological factors affect response characteristics. If a blunted
response mostly reflects the impact of arterial stenosis that standard diagnostic techniques already
capture, a test of reactive hyperemia may add very limited clinical value. Therefore, this paper
sought to provide mechanistic explanations for the discrepancies in the literature as well as
supporting data that may help connect the perfusion characteristics with vascular physiology.
2.5.1 Technical experiments of peripheral ASL
This study demonstrated that background signal contamination and variable labeling efficiency
might be the main sources of discrepancies in peripheral ASL studies. The ischemia-reperfusion
paradigm generated a large dynamic change of the baseline signal (Figure 2.4a), which was
associated with the oxygenation change of the static muscle [45]. The pairwise subtraction would
mix the signal resulted from the time derivative of this dynamic baseline with the signal difference
created by blood labeling, contaminating the quantification of perfusion. The contamination
explained the negative perfusion seen at the beginning of ischemia and later recovery in some
studies [35], [54] because the baseline signal had a negative slope at these time points. This
background artifact was only addressed in two studies [49], [54], where time-matched subtraction
was used. However, the interpolation approach in time-matched subtraction may not eliminate the
signal contamination completely. Ultimately, a more robust strategy to deal with background
artifact is needed. In fact, a solution has been well established and widely used in brain ASL, which
is to null the static tissue signal with background suppression pulses [53]. Therefore, background
suppression was incorporated in the later physiological experiments. The signal of static tissue was
reduced to below 5% of its unsuppressed level, which remarkably improved SNR and reduced
background artifact.
35
Both pulsed and (pseudo-) continuous labeling have been used in calf reactive hyperemia. Pulsed
labeling is known to be insensitive to blood velocity, with labeling efficiency close to 100% [55],
whereas continuous labeling relies on the mechanism of adiabatic fast passage (AFP) that
modulates the labeling with blood velocity [56]. Previous studies suggested a minimum of 10 cm/s
for optimal continuous labeling with efficiency above 90% [48]. This study indicated that the
triphasic popliteal velocity at rest might be suboptimal for AFP. Moreover, throughout the cuff
experiment, the blood velocity profile changed from triphasic at rest to monophasic during
hyperemia, then back to triphasic after recovery. These findings suggested that assuming a constant
labeling efficiency throughout the entire perfusion time course in the (pseudo-) continuous
schemes may result in scaling bias in the quantification of perfusion [35], [38]. On the other hand,
pulsed ASL does not have this issue of variable labeling efficiency.
To further optimized ASL for reactive hyperemia, tibial blood velocity across ischemia-
reperfusion was measured. The measured resting peak forward, peak reverse, and mean tibial
velocities, as well as the ratio of mean velocity between hyperemia and rest state, were consistent
with ultrasound measurement [3], [57]. The blood velocity affects not only the labeling but also
the transit time of the labeled blood. The configuration of pCASL required longer transit distance,
resulting in a significantly weaker difference signal between control and labeled images. Therefore,
after knowing that the tibial blood velocity, FAIR was deemed more appropriate for the subsequent
physiological studies.
In addition, because the labeling region should be filled with fresh blood to maximize the
difference signal, blood velocity was used to estimate the proper repetition rate of FAIR for the
slowest flow. Assuming a labeling region of 20 cm in length, a velocity of 10 cm/s would mean
that a separation of 2 s between two consecutive acquisitions would be ideal. The velocity results
suggested that a repetition time of less than 1 s was sufficient for refreshment of the labeling region
during peak reactive hyperemia, whereas more than 3 s was required at rest. Therefore, the FAIR
acquisition was repeated every 3.35 s to ensure reasonable recording of reactive hyperemia and
recovered/baseline perfusion.
Overall, this part of the study investigated the ASL signal behavior specific to calf muscle
perfusion throughout ischemia-reperfusion, which provided supporting data for optimal peripheral
ASL settings.
36
2.5.2 Characteristics of ASL reactive hyperemia
The influences of ischemic duration and PAD on calf reactive hyperemia were demonstrated using
the FAIR sequence. The responses to 5 minutes of ischemia were similar to a subset of previous
pulsed ASL results [37], [49], [58]. The whole-leg peak response of 50 mL/100g/min was also
consistent with perfusion measured by venous occlusion plethysmography [58]. No other ASL
studies have reported leg reactive hyperemia induced by 1- or 2-min ischemia. However, the
observed evolution of reactive hyperemia with ischemic durations was consistent with
measurements using red blood cell velocity and plethysmography in other skeletal muscle tissue
beds [59], [60]. This study is the first peripheral ASL that reported such consistency of
physiological effect with ischemic duration.
Protocols with 2-min and 5-min ischemia may yield similar results in peak perfusion and TTP,
with the benefit of being more tolerable for the 2-min ischemia protocol. The peak perfusion
showed excellent repeatability and a significant reduction in patients with PAD. TTP had
acceptable repeatability and was delayed in PAD (not significant). These findings are similar to
those in other studies using 5-min ischemia [36], [37]. On the other hand, the poor repeatability of
TTR and HFV may be caused by the ambiguity in determining the recovery time, which was
affected by the noise at baseline perfusion and the choice of threshold.
The characteristics of the rising and recovery phases of reactive hyperemia seem to be related to
different physiological indications. The rising phase of reactive hyperemia exhibited a similar
slope regardless of ischemic duration (Figure 2.6), and ability to detect the influence of arterial
stenoses was preserved when using shorter ischemia. These findings suggest that the rising phase
may be associated with the inflow rate, reflecting arterial properties such as the resistance and
compliance. Indeed, TTP had been shown to correlate with ABI inversely, and the patients with
ABI<0.5 had extremely delayed TTP [37]. On the other hand, the recovery phase of reactive
hyperemia characterized by the TTR and HFV changed consistently with the amount of ischemic
stress (Table 2.4). Therefore, the recovery characteristics may reflect microvascular function in
regulating perfusion response to stress. However, the implication and clinical relevance of the
recovery characteristics remain unclear. The only research group that reported recovery
characteristics of ASL reactive hyperemia, in their sequence validation study [61], did not continue
to do it in their larger, clinically oriented study [37]. No other study has attempted to quantify the
37
recovery characteristics. Therefore, there is no pre-existing opinion on how recovery
characteristics would differ between the healthy and patient groups in this study.
The poor repeatability of TTR and HFV does not diminish the importance of recovery
characteristics. Instead, it suggests that a better characterization scheme for the recovery phase to
reduce the variability caused by baseline noise is needed. In the next chapter, a novel physiological
model will be presented. The model will characterize reactive hyperemia in a curve-fitting fashion,
and the resulting parameters allowed us to interpret clinical data in the context of macrovascular
stenoses and microvascular dysfunction. In this regard, the current study, which standardizes the
perfusion measurement, is a critical first step for the determination of specific physiological
parameters in the theoretical model.
Limitation
This study investigated only the background artifact and the influences associated with blood
velocity specific to the leg. Other methodological differences among previous ASL studies, such
as the use of transmit knee coil, ROI sampling or combination of muscle groups, and assumptions
of blood transit time behind the choice of PLD, were not covered here. Other limitations of this
study include a suboptimal sequence for velocity measurement and lack of independent reference
of perfusion. Underestimation of systolic velocity was found when compared to rapid cardiac
phase-resolved measurement using projection imaging [49]. However, the VENC used in this
study was appropriate for the particular interest in slow flow, and the average velocities and
triphasic features were consistent with previous findings [3], [49], [57]. During hyperemia, the
tibial velocities should be interpreted as the average during the first phase-contrast acquisition
(about 20 s) of reperfusion, and not to be directly compared to rapidly acquired peak velocity at
popliteal artery just below the knee [49]. Despite the fact that there was no standard perfusion
measurement to which the results in this study could be compared to, the characteristics of
ischemic duration-dependent reactive hyperemia agreed with the results using other techniques
[58]-[60]. Detection of PAD has been available with standard diagnostic tools; hence, it was not
the goal of this study. Instead, the goal was to demonstrate the feasibility of ASL-based
characterization of calf perfusion in subject groups with known vascular conditions, which call for
further studies that may improve the understanding of pathophysiology in limb ischemia.
38
2.6 Conclusion
The current findings suggest that the mechanistic reasons for previous inconsistent peripheral ASL
measurements may primarily be background artifact and variable labeling efficiency. This study
demonstrated a technically justified FAIR for dynamic perfusion responses in the calf, with the
recorded reactive hyperemia changing consistently with the duration of ischemia in the healthy
subjects and showing a significant reduction of peak perfusion in the patients with PAD. The data
shed light on the complex physiology involved in reactive hyperemia. More advanced
characterization and physiological interpretation are needed for ASL reactive hyperemia to be
applied to clinical microvascular assessment, which may improve the understanding of the role of
the microcirculation for limb salvage.
39
Chapter 3
Model-based interpretation of reactive hyperemia This chapter is adapted from the article “A physiological model for interpretation of arterial spin
labeling reactive hyperemia of calf muscles” authored by Hou-Jen Chen and Graham Wright and
published in 2017 PLoS ONE 12(8): e0183259. doi: 10.1371/journal.pone.0183259. This chapter
describes the establishment of a physiological model to facilitate the interpretation of reactive
hyperemia.
3.1 Background
The pathophysiology of limb ischemia involves occlusive large-vessel disease and microvascular
dysfunction; the latter, however, is poorly understood. In the last chapter, the involvement of
microvascular function in regulating the calf muscle perfusion response to ischemia was
demonstrated by ASL reactive hyperemia. However, using ASL reactive hyperemia to detect
microvascular disease of PAD remains challenging. The exact vasodilatory mechanisms
responsible for leg reactive hyperemia have not been fully understood, which limits the
interpretation of abnormal responses observed in PAD. It is not clear if a blunted response implies
arterial stenoses, microvascular dysfunction, or a combination of these effects. The commonly
used protocol of 5-minute ischemia was established in the past to induce large and prolonged
responses based on bulk flow measurement [23], [60], [62], for which the presence of flow-limiting
stenoses would be easily detected. However, given that the microvascular relaxation usually occurs
tens of seconds after cuffing, increasing the sensitivity to microvascular dysfunction may require
a different protocol.
The goal of this work is to provide the physiological basis of calf reactive hyperemia for data
interpretation and establishment of a stress protocol that is more specific to microvascular
dysfunction. This work will develop a subject-specific model to extract physiologically relevant
features in individual responses, which may lead to microvascular assessment using ASL reactive
hyperemia and classification of patient groups based on more than the macroscopic lesions in their
arteries. The hypothesis is that a model-based analysis may help explain how the arterial stenoses
40
and microvascular dysfunction affect the overall shape of the ASL reactive hyperemia response in
the calf. The specific aims are: 1) to describe the tissue metabolic response during different cuffing
durations up to 5 min; 2) to simulate reactive hyperemia under normal conditions and conditions
of arterial stenoses and microvascular dysfunction; 3) to fit the acquired reactive hyperemia of
cuffing duration-varying experiments in healthy subjects; and 4) to demonstrate the model’s
potential in preliminary patient studies.
3.2 The model
To deal with the challenges in characterization and interpretation of ASL data and standardization
of the cuffing protocol, it is desirable to establish a model describing the essential physiological
factors involved in reactive hyperemia. Flow models essentially describe the integrated effects of
vascular segments connected in series, representing the arteries, microvasculature, and veins. Each
vascular segment can be described with a combination of flow resistances (R) and vessel
compliance (C). For dynamically regulated flow, the elements of resistance and compliance are
made variable to represent the microvascular responses to factors such as luminal pressure, shear
stress, and vasoactive substances. A theoretical framework to link the physiological factors,
including oxygen and metabolites, to the lumped circuit elements was laid out by Ursino et al.
[63], [64] for cerebral blood flow regulation. The work considered modeling of reactive hyperemia,
but the result was far off from experimental data [65], mainly due to the lack of measurement
technique to accurately characterize the rapid vasodilatory response mechanisms back then.
Adapting this framework, Spronck et al. [66] recently proposed a model suitable for fitting
individual cerebral flow responses to experimental maneuvers. Meanwhile, Secomb’s group
extensively studied the contribution of different flow regulatory mechanisms to the steady-state
skeletal muscle perfusion [67]. In addition, de Mul et al. [68] published a simple model which
characterizes PAD patients’ laser Doppler reactive hyperemia with just two time constants, for the
rise and decay of reactive hyperemia. Although the model did not take microvascular physiology
into account, it demonstrated that a simple RC block is sufficient to characterize the effect of
arterial stenoses on perfusion. Therefore, although a physiological model of skeletal muscle
reactive hyperemia had not been established, there are plenty of reference materials for use as a
41
starting point. We will create the skeletal muscle perfusion model by adapting Spronck’s model
and replacing the description of cerebral flow control with ischemia-induced vascular relaxation.
A lumped parameter model of three vascular segments, shown in Figure 3.1, describes the tibial
circulation. The model is a functional representation instead of a structural representation. The
popliteal segment represents the net upstream resistance and compliance for inflow, with fixed
values that do not change with ischemic duration. The arteriolar segment represents the time-
varying microvascular properties that change across ischemia-reperfusion. Capillaries may be
recruited during hyperemia and affect microvascular resistance, but the functional effect is
incorporated in the arteriolar segment. The venous segment represents the net downstream
properties for outflow.
Figure 3.1: The flow circuit of calf circulation. Pai , Pvo and Pim denote arterial input, venous output, and
intramuscular pressure, respectively. Qai and Qvo denote popliteal arterial inflow and venous output flow,
respectively. Qa and Qv are the flow to the arteriolar and venous circulation, respectively. The switches S1 and S2
represent the flow blocking effect of cuffing. Rp: popliteal resistance, Cp: popliteal compliance, Pp: end arterial
pressure for perfusion input to the arterioles; Ra: arteriolar resistance, Ca: arteriolar compliance, Pa mid-arteriole
pressure; Rv: venous resistance, Cv: venous compliance, Pv mid-vein pressure.
Perfusion-time waveforms can be simulated by solving the lumped parameters of the system as
functions of time. The flow rate in each segment can be derived from Ohm’s law. Specifically,
ASL perfusion fASL measured at the tissue level should correspond to Qv in the model:
!"#$ =&'
()=
*+ − *'
(.+ + .')/2 ( 3.1 )
where Vm represents the volume of the perfusion territory (Table 3.1) and other parameters are
defined in Figure 3.1 caption. The steady-state baseline flow Qr of the system is set to be the group
average of our healthy subjects. The segmental pressure distribution of the system at Qr is chosen
to match the reference values [69]. In the next section, the governing differential equations to
42
describe the system’s response to flow interruption with time-varying lumped parameters will be
derived.
Table 3.1: Constant parameters of the model.
Symbol Value Unit Description Pvo 14 mmHg Venous output pressure Pim 10 mmHg Intramuscular fluid pressure Qr 144 mL/min Resting flow rate to the lower leg Vm 3 L Estimated volume of one lower leg ηb 3×10-5 mmHg⋅s Blood viscosity ra,0 75 um Inner radius of unstressed arteriolar wall ha,0 25 um Wall thickness of unstressed arteriole σ1 11.19 mmHg Constant for elastic tension model σ2 52.51 mmHg Constant for elastic tension model ke 4.5 – Constant for elastic tension Tm,0 3 mmHg⋅s Maximal muscular tension of arterioles ra,m 128 µm Radius at maximal muscular tension ra,t 174 µm Constant for muscular tension model nm 1.75 – Constant for muscular tension model
3.2.1 The lumped parameters
The popliteal artery is modeled as a simple resistor-capacitor unit. The normal arterial resistance
is calculated assuming Poiseuille flow:
Rp= 8ηb
πrp4lp ( 3.2 )
where the arterial radius rp = 0.25 cm and length lp = 40 cm are estimated from population data,
resulting in a resistance of 0.013 mmHg⋅min/mL and an arterial pressure drop Pai−Pp = 1.8775
mmHg at Qr. The arterial capacitance Cp is chosen such that the arterial time constant matches the
reported experimental values τp=RpCp≈ 5 s [68].
43
Venous resistance is chosen such that a venous pressure drop is Pv−Pvo = 1 mmHg at Qr. With an
empirical choice of the venous time constant RvCv = 10 s, Cv can be calculated. Rv and Cv are taken
to be constant [66].
Because the arteriolar circulation is controlled by complex microvascular regulatory mechanisms
rather than apparent lumped parameters with fixed values, it is necessary to express Ra and Ca with
time-varying radius ra and consider the relationship between the pressure and radius based on the
mechanical properties of the arterioles [64], [66], [70]. In this regard, the mid arteriolar pressure
Pa is related to the wall tension T and radius ra by Laplace’s law. The wall tension consists of the
passive elastic component Te and active muscular component Tm scaled by the level of activation
A. Together these give:
3 = *+4+ − *5)(4+ + ℎ+) = 37 + 83) ( 3.3 )
where the arteriolar wall thickness ha is formulated by :
ℎ+ = 4+9 + 24+,;ℎ+,; + ℎ+,;
9− 4+. ( 3.4 )
The elastic tension is defined as:
37 = [>?@AB(CDECD,F)/CD,F − >9] ⋅ ℎ+ ( 3.5 )
where constants are given in Table 3.1. The muscular tension generated by the fully activated
arteriolar smooth muscle at a given radius is defined as:
3) = 3),;@
E|CDECD,J
CD,KECD,J|LJ
. ( 3.6 )
An arteriolar pressure drop ΔPa = Pp−Pv = 70 mmHg and mid-arteriole pressure Pa = 50 mmHg are assumed for Qr [69]. The resistance Ra and volume Va (related to Ca) expressed as a function of ra is shown in Appendix A.
44
With the lumped parameters defined, the governing equations of the flow system are expressed as:
M*N
MO=&+5 − &+
PN , ( 3.7 )
M4+
MO=*+(.' − 2.+) + *N(.+ − .') + *'.+)
4+QR.+(.' + .S), and ( 3.8 )
M*'
MO=&' − &'T
P'. ( 3.9 )
At any time point, a physiological status of the arteriolar segment will adjust the activation level
A of the arteriolar wall, which changes the pressure and radius of the arterioles as described in Eq.
3.3, resulting in changes of the entire flow system described by Eq. 3.7 – 3.9. Therefore, the next
section will describe how flow regulatory mechanisms change the activation level A.
3.2.2 Microvascular regulation
The vessel wall activation level A is defined as a sigmoidal function of the total regulatory
influence z:
8 =1
1 + @E9V. ( 3.10 )
The regulatory influences include myogenic response and shear stress response, which are
associated with the blood pressure and flow directly perturbed by flow interruption, as well as
ischemia-induced vasoactive substances with concentrations dependent on the ischemic duration.
To date, the mechanisms specific for ischemia-induced vasodilation have not been clearly
explained. However, evidence in the literature reveals that two microvascular regulatory
mechanisms are very likely to be tied to ischemia-induced vasodilation. First, it is found that ATP
released by red blood cells (RBCs) during hypoxia can activate the purinergic receptors on the
endothelial cells and triggers endothelium-dependent vasorelaxation [19]. Second, ischemia-
induced imbalance of energetic supply and demand would lead to an accumulation of interstitial
adenosine, which can activate the receptors on vascular smooth muscles and induces endothelium-
independent vasorelaxation [71]. Therefore, the total regulatory influence was defined as:
W = X5Y5
5
+ Y5Z5[ ( 3.11 )
45
where i = [myo, sh, ATP, ado] stands for the myogenic, shear stress-based, intravascular ATP-
mediated, and interstitial adenosine-mediated regulation, respectively. For each regulatory
mechanism, the regulatory influence is the regulatory state x multiplied by the respective gain g.
The parameter xinit sets the baseline arteriolar tension, radius, and resistance, which essentially
determines the dynamic range for flow increases in the system.
The regulatory state is perturbed by the associated stimulus and varies based on a certain time
constant (via a first-order approach):
MY5
MO=\5 − Y5
]5 ( 3.12 )
where τi stands for the time constant of each regulatory mechanism. The definition of the stimulus
function yi is explained in Appendix B. In short, ymyo is the deviation of current arteriolar wall
tension from the baseline tension; ysh is the deviation of current shear stress from the baseline shear
stress, and so on. Since the wall-derived regulatory mechanisms have been well characterized and
are known to be relatively weak when compared with metabolic flow regulation [67], the model
uses fixed values similar to a previous study [66] : gmyo = 1, gsh = 1, τmyo = 6 s, and τsh = 60 s.
The gain g and time constant ] above are parameters introduced from a systems perspective.
However, their physiological meanings are straightforward and obvious if we use the parameters
to express the change of wall tension explicitly. As explained in Appendix C, gATP is defined as
the sensitivity of the arterioles to ATP that scales the response of wall tension, whereas τATP is
defined as the response time to ATP that determines the quickness of wall tension response.
Likewise, gado and τado are defined as the sensitivity and response time of the arterioles to interstitial
adenosine.
The model assumes that intravascular ATP and interstitial adenosine are the main vasoactive
substances for metabolic regulations because their production depends on the level of oxygenation
in the capillary and tissue, respectively. Therefore, a two-compartment model for the transport of
oxygen is required. The current model adapted Lai’s model and used a constant permeability
surface area-product (D) for diffusion of oxygen between the capillary and tissue compartment
[72]. The differential equations of compartmental oxygen pressure are:
46
M*^9_
MO=!(^9+C[ − ^9_) − `a(*^9_ − *^9[)
b_c_ ( 3.13 )
M*^9[
MO=`a(*^9_ − *^9[) − (^9
b[c[ ( 3.14 )
where PO2c and PO2t are the oxygen partial pressure of the capillary and tissue compartment,
respectively. O2art and O2c are the oxygen concentration of the inflow arterial and capillary blood,
respectively. The solubility of oxygen in blood α is 1.46 µM/mmHg, and volume fraction of
capillary vc and tissue vt are 3% and 97%, respectively. The metabolic function VO2 and the
parameter γ=dO2/dPO2, varying with the level of oxygenation, are defined in Appendix B.
The concentration of intravascular ATP CATP can be calculated by considering the inflow arterial
ATP concentration CATP,in , oxygen saturation-dependent ATP release rate R, and degradation
by the endothelial surface:
MP"de
MO=!
b_
1 − ℎf
1 − ℎ[(P"de,5Z − P"de) +
ℎ[
1 − ℎ[. −
2Qf
4_(1 − ℎ[)P"de ( 3.15 )
where CATP,in = 0.1 µM, the capillary radius rc = 3µm, tube hematocrit ht = 0.4, discharge
hematocrit hd = 0.3, and degradation constant kd=2×10−4 cm/s [67]. The ATP release function R is
given by Arciero’s paper[67]:
.(*^9_) = .;(1 − .?
*^9_Z
*^9_Z+ *g;,hi
Z) ( 3.16 )
where R0 = 84 µM/min, R1 = 0.891, P50,Hb = 26.8 mmHg, and n = 2.7.
As ischemia progresses in the tissue compartment during cuffing, the skeletal muscle switches
from aerobic metabolism to anaerobic metabolism and the formation of interstitial adenosine Cado
is increased. On the other hand, interstitial adenosine is cleared slowly by cellular uptake back to
the tissue with the rate described by Michaelis-Menten equation [73]. Therefore, the dynamic of
Cado is formulated as:
47
MP+fT
MO= j+fT − ()+k,+fT
P+fT
P+fT + l),+fT ( 3.17 )
where Vmax,ado = 100 µM/min and Km,ado = 200 µM are chosen for the adenosine clearance rate
[74]. By referring to the concentration in various physiological conditions [75], [76], the adenosine
formation rates Fado in µM/min are assumed here to be:
j+fT *^9[ = 0.05*^9[ ≥ *^9_C
= 0.5 otherwise ( 3.18 )
where the critical oxygen pressure PO2cr for anaerobic metabolism is assumed to 10 mmHg [77].
The baseline interstitial concentration of adenosine is 0.1 µM.
3.2.3 Model behavior
Simulation of ischemia and reperfusion under normal and diseased conditions is presented in this
section to facilitate interpretation of the model behavior. However, this simulation was developed
based on the results of fitting the model parameters to the responses of healthy individuals, which
will be described in the methods and results sections.
During ischemia, the muscle tissue continues to consume oxygen at the baseline metabolic rate
until the PO2 is relatively low, as simulated in Figure 3.2a. The concentration of intravascular ATP
increases with deoxygenation in the capillary compartment, whereas the interstitial adenosine only
starts to increase when PO2cr is reached and anaerobic metabolism begins, as seen in Figure 3.2b.
The influences of each regulatory mechanism are drawn in Figure 3.2c. While the total regulatory
influence z continues to be enhanced by the accumulation of metabolites, the vascular smooth
muscle is almost de-activated after 2 minutes of ischemia, as shown by the level of activation A in
Figure 3.2d.
48
Figure 3.2: Ischemic responses of the tissue and arterioles. Modeling the responses during 5 minutes of ischemia
in the muscle, including (a) the oxygen tension, (b) concentrations of the vasoactive substances, (c) the influences of
vasoregulation mechanisms, and (d) the summed influence z and level of activation A of vascular smooth muscle.
The modeled reactive hyperemia is shown in Figure 3.3. Because of the rapid nature of
intravascular ATP-mediated vasorelaxation, an evident response can be induced with just one
minute of ischemia. As illustrated in Figure 3.3a, increasing ischemic duration would slightly
enhance the magnitude of the modeled response, but ischemia longer than 2 minutes would mostly
prolong the modeled reactive hyperemia. Arterial stenoses simulated by doubling Rp would reduce
the response magnitude regardless of ischemic duration, as depicted in Figure 3.3b, 3.3c, and 3.3d,
because it reduces the input perfusion pressure to the microcirculation. Doubled Rp would also
double the arterial time constant, resulting in longer rise time. Microvascular dysfunction does not
delay the rise time, on the other hand. Depending on the ischemic duration used, microvascular
dysfunction does not necessarily reduce the magnitude of the responses to a large extent. As
simulated by reducing gATP in the model, microvascular dysfunction appears to cause earlier
attenuation, suggesting that the dysfunctional arterioles tend to constrict. The influence of
microvascular dysfunction can be seen by comparing microvascular dysfunction alone with normal
0 1 2 3 4 5
10
20
30
Time (min)
PO
2 (mm
Hg)
(a)
CapillaryTissue
0 1 2 3 4 50
0.2
0.4
0.6
0.8
1
Time (min)
Con
cent
ratio
n (u
M)
(b)
ATPAdenosine
0 1 2 3 4 5−6
−4
−2
0
Time (min)
Eac
h in
fluen
ce (a
.u.)
(c)
myshATPAdo
0 1 2 3 4 5
−8
−6
−4
−2
0
Time (min)a.
u.
(d)
zA
49
function, or by comparing combined micro- and macro-vascular disease with stenoses alone. The
differences resulted from the presence of microvascular dysfunction are more evident in the
responses to 2-min ischemia than longer ischemia. This result of modeling may be explained by
the fact that as the concentration of interstitial adenosine increases in longer ischemia the vascular
smooth muscles eventually relax, regardless of the endothelium-dependent ATP signal. Therefore,
the model suggests that a protocol using shorter ischemic duration may be more sensitive to
endothelium-related microvascular dysfunction.
Figure 3.3: The effects of ischemic duration, arterial stenosis, and microvascular dysfunction on reactive
hyperemia. (a) Normal responses to various ischemic durations and diseased responses to (b) 2-min, (c) 3-min, and
(d) 5-min of ischemia. Arterial stenosis was simulated by doubling the popliteal resistance Rp, and microvascular
dysfunction was simulated by reducing gATP to 40%. "Combined" includes both effects at the same time.
3.3 Methods
The perfusion data used for simulation and fitting was the mid-calf ASL reactive hyperemia of 7
healthy subjects (age <30 yrs) and 9 patients with PAD, acquired as described in Chapter 2. The
0 20 40 60 80 100 120
0
10
20
30
40
50
Time (s)
Perfu
sion
(mL/
100g
/min
)
(a)
1−min2−min3−min5−min
0 20 40 60 80 100 120
0
10
20
30
40
50
Time (s)
Perfu
sion
(mL/
100g
/min
)
(b)
NormalStenosisMicro. dysf.Combined
0 20 40 60 80 100 120
0
10
20
30
40
50
Time (s)
Perfu
sion
(mL/
100g
/min
)
(c)
NormalStenosisMicro. dysf.Combined
0 20 40 60 80 100 120
0
10
20
30
40
50
Time (s)
Perfu
sion
(mL/
100g
/min
)
(d)
NormalStenosisMicro. dysf.Combined
50
healthy subjects underwent 1,2, 3, and 5 minutes of arterial occlusion via cuff inflation; the patients
only underwent 2-min cuffing for the more affect leg, which was determined based on self-reported
symptoms and physicians’ dictation. The median age of the patient group was 74.5 yrs (51 to 83).
The ABI ranged from 0.3 to 0.87. Seven patients had claudication, and the remaining two had rest
pain.
Modeling and fitting
The flow regulation system described previously, including the lumped parameters and oxygen
transport model, was implemented using MATLAB. Modeled reactive hyperemia was generated
by solving the governing differential equations Eq. 3.7 – 3.9 and Eq. 3.12 using the built-in solver
function ode15s. Subject-specific model parameters were identified as described in the next two
paragraphs. The values of these parameters were estimated by searching for the minimum least-
squares fit between the modeled and ASL-measured perfusion, using the built-in optimization
function lsqnonlin. The quality of fit was assessed by calculating the reduced chi-square:
qC7f9
=1
r − s − 1
(!(t) − !"#$(t))9
>9
u
5
( 3.19 )
where σ2 is the variance of baseline perfusion, N the data length, and n the number of free
parameters.
For the healthy group, [fr, gATP, τATP, gado, τado, τp] were chosen as the free parameters. The logic
lies in the approach to address the individual differences in the resting perfusion fr (= Qr/Vm) and
reactive hyperemia. First, we attributed the variation of resting perfusion in healthy individuals to
the baseline arteriolar tone, reflecting the variation in capillary permeability and tissue metabolism,
not the variation in resistances of the arteries or veins. Therefore, fr was varied to determine xinit
for the baseline arteriolar tone. The group average resting perfusion was aligned with a positive
xinit = 0.45 chosen empirically. Next, we assumed that the differences in reactive hyperemia result
from the variation in the sensitivities and time constants of ATP and adenosine in addition to the
arterial time constant (reflecting the difference in arterial compliance if the resistance is fixed).
Therefore, the steps of fitting the cuffing duration-varying reactive hyperemia are: 1) use [fr, gado,
51
τado, τp] with fixed gATP and τATP to fit the responses to 3- and 5-min cuffing; and 2) use [fr, gATP,
τATP] with the resulting gado, τado, and τp from step 1 to fit the responses to 1- and 2-min cuffing.
The parameters gATP and τATP were not included as free parameters in step 1 because their
influences on the response to long cuffing were extremely small and a wide range of their values
can be used, making the resulting values unstable and not meaningful. Likewise, only the more
sensitive parameters for short cuffing were included in step 2. The allowed range of free parameters
in the fitting process are: 2 ≤fr≤ 10 (mL/100g/min), 1 ≤gATP≤ 20, 1 ≤gado≤ 15, 6 ≤τATP≤ 24 (s), 12
≤τado≤ 60 (s), 1.5 ≤τp≤ 30 (s), determined empirically based on their influences on the response
and a previous modeling work [67].
For the patient group, [fr, gATP, τATP, Rp, τp] were chosen as the free parameters. Rp was included
as a free parameter because these PAD patients have various degrees of arterial stenoses affecting
reactive hyperemia. Adenosine parameters were not included as free parameters because of their
minimal influence on perfusion response to 2-min cuffing. Their values were fixed to the average
values of the healthy group. The allowed range of free parameters in the fitting process are: 3 ≤fr≤
8 (mL/100g/min), 1 ≤gATP≤ 12, 12 ≤τATP≤ 36 (s), 1.5 ≤τp≤ 25 (s), and 1 ≤Rp≤ 6 (starting from here
Rp is expressed as the ratio to the calculated standard value from Eq. 3.2). The ranges of these
parameters were narrower than those used in the healthy group because the physiological ranges
had been known after fitting the healthy responses.
Statistical analysis
The peak, TTP, and model parameters that resulted in the best fit to the responses were compared
between the healthy subjects and patients using a two-sided Students’ t test. A p-value of 0.05 was
assumed to indicate statistical significance.
3.4 Results
The model could generate ischemic-duration-dependent responses very similar to the measured
reactive hyperemia in healthy subjects. The model achieved reasonable fits with one set of
parameters for all responses to various durations of cuffing in each subject, as shown in Figure 3.4.
52
The quality of fit, measured by χ2red was affected by baseline drift, sudden jumps, and small
oscillations of the ASL signal in some cases. Noise level varied between recordings. Looking
through the healthy group, the model did not necessarily fit better to the large and long responses
(5-min cuffing) than the transient responses.
Figure 3.4: Fitting the responses of a healthy subject with the model. The resulting parameters are: [fr, gATP,
gado, τATP, τado, τp] = [5.59, 14.67, 3.49, 18.0, 44.5, 3.05].
The parameter values of the healthy group were summarized as means ± standard deviations: fr=
4.95 ± 0.92 mL/100g/min, gATP = 15.66 ± 2.05, gado = 4.18 ± 2.41, τATP = 15.86 ± 2.66 s, τado =
43.20 ± 21.11 s, and τp = 4.25 ± 1.16 s. The parameters related to adenosine had a broader range
than other parameters, with coefficients of variation (CV) 27.4% and 49.6% for gado and τado,
respectively. The rest of parameters were relatively consistent within the group (CV<20%).
The model fitting of diseased responses to 2-min cuffing was shown in Figure 3.5. Eight out of the
9 responses had χ2red ranged between 0.59 to 1.39, indicating reasonable fits; response 3 had a χ2
red
of 4.17, indicating a poor fit. Each recording had a different level of noise. Response 3 was very
0 20 40 60 80 100
0
20
40
60
Time (s)
Perfu
sion
(mL/
100g
/min
)
1−min ischemia
0 20 40 60 80 100
0
20
40
60
Time (s)
Perfu
sion
(mL/
100g
/min
)
2−min ischemia
0 20 40 60 80 100
0
20
40
60
Time (s)
Perfu
sion
(mL/
100g
/min
)
3−min ischemia
0 20 40 60 80 100
0
20
40
60
Time (s)
Perfu
sion
(mL/
100g
/min
)5−min ischemia
53
small and only three data points were two standard deviations above the baseline perfusion, which
was not sufficient to determine the five free model parameters.
The model parameters were more variable within the patient group than the healthy group. As
compared in Figure 3.6, responses in the patient group were significantly lower in peak perfusion,
but not in TTP (p=0.102), than those in the healthy group. With the model-based analysis through
fitting, the patient group showed higher Rp and lower gATP than the healthy group. The p-values
were 0.285, 0.098, and 0.461 for comparison of fr, τp, and τATP.
Figure 3.5: Fitting of patients’ responses to 2-min cuffing with the model-generated response. The fitting with
exceptionally large error is plotted as the blue line.
0 50 1000
20
40
Time (s)
mL/
100g
/min
Patient 1
0 50 1000
20
40
Time (s)
mL/
100g
/min
Patient 2
0 50 1000
20
40
Time (s)
mL/
100g
/min
Patient 3
0 50 1000
20
40
Time (s)
mL/
100g
/min
Patient 4
0 50 1000
20
40
Time (s)
mL/
100g
/min
Patient 5
0 50 1000
20
40
Time (s)
mL/
100g
/min
Patient 6
0 50 1000
20
40
Time (s)
mL/
100g
/min
Patient 7
0 50 1000
20
40
Time (s)
mL/
100g
/min
Patient 8
0 50 1000
20
40
Time (s)
mL/
100g
/min
Patient 9
54
Figure 3.6: Characteristics of reactive hyperemia induced by 2-min cuffing. Left: characterization using
apparent indices. Right: characterization using model-derived parameters. * denotes statistical significance (P <
0.05).
3.5 Discussion
3.5.1 Establishment of the model
The work in this chapter sought to provide an explanation to the dependence of reactive hyperemic
characteristics on ischemic duration using the model. Although reactive hyperemia is often related
to endothelial function, which is commonly assessed by flow-mediated dilatation via the effect of
shear stress, it is unlikely that shear stress plays a significant role in ischemia-induced vasodilation.
The first reason is that flow interruption should lead to shear-dependent vasoconstriction, not
vasodilation. Second, the change in shear stress induced by cuffing is independent of the ischemic
duration and hence should not enhance the perfusion response if ischemic duration increases. The
same arguments can be applied to the pressure-related myogenic regulatory mechanism. Indeed, it
has been suggested that these wall-derived mechanisms should be relatively weak and work against
the metabolic flow regulation [67]. To identify the mechanisms more relevant to ischemia, we
searched for mechanisms closely tied to the dynamics of the deoxygenation process during
ischemia. Studies of magnetic resonance spectroscopy have shown that myoglobin desaturation
begins at 100-120 s and plateaus at 300-420 s in the course of acute ischemia [78], [79]. The initial
delay of myoglobin desaturation in the tissue pool is due to the contribution of oxygen from the
blood pool. At the later stage of ischemia, high level of myoglobin desaturation corresponded to
the onset of anaerobic metabolism indicated by the change of phosphocreatine in the tissue. These
fr Rp τp gATP τATP0
5
10
15
20
25
Par
amet
er v
alue
Peak TTP0
20
40
60Normal (n=7)PAD (n=9)
Val
ue o
f res
pons
e in
dex
*
*
*
55
two-compartment deoxygenation dynamics were incorporated in our model (Figure 3.2) to control
the metabolite-derived regulatory mechanisms.
In the capillary compartment, the model linked vascular tension to the oxygen saturation-
dependent release of ATP by RBCs. The theory of RBCs being the oxygen sensor for local flow
regulation has long been established [19]. Our simulated concentration of ATP was in the
submicromolar range, consistent with intravascular measurements in human hypoxia experiments
[80]. The estimated time constant of ATP-mediated regulation τATP was 15.4 ± 2.7 s, which is very
close to the reported values (8-16 s) from direct observation of exposed arterioles in an animal
study [81]. The function of the ATP receptor on the endothelium was found to be attenuated in
type 2 diabetes [82]. This finding corresponded to a reduction in gATP in our model and was
simulated as microvascular dysfunction. Consistent with the simulation, reduced gATP was found
in the patient group. However, the model does not include the biological details of ATP signaling.
In addition to a change of receptor sensitivity, other reasons such as a reduction in the oxygen
consumption rate, in the total number of receptors due to vascular rarefaction, or in the ATP release
rate can all cause reduced gATP. After all, the current work only uses perfusion waveforms to look
at the disease and cannot reflect all the detail.
In the tissue compartment, the concentration of interstitial adenosine was modeled. Increased
formation of adenosine during hypoxia results from the elevated activity of ecto 5’-nucleotidase
converting muscle-released AMP into adenosine [83]. However, it is difficult to quantify the exact
reaction rate and concentrations of AMP and adenosine because a wide range of values was
reported, potentially because real-time interstitial measurements in vivo remain challenging [84].
Hence, this study only roughly estimated the adenosine formation rate and described it as a simple
2-mode function in Eq. 3.18. The high variability of gado and τado in the fitting of normal responses
may result from such simplification of adenosine formation. Furthermore, other metabolites, such
as lactic acid and hydrogen ions, may also be involved in the tissue compartment for extended
ischemia. Therefore, the current model may be less accurate in explaining/simulating the responses
to longer ischemia.
With intravascular ATP and interstitial adenosine incorporated, the model clearly demonstrated
ischemic duration-dependent responses. The rapid magnitude increase for responses to short
56
ischemia is very likely to be contributed by the ATP released by deoxygenated RBCs. On the other
hand, longer ischemia appeared to prolong the responses by accumulating adenosine in the tissue.
The relationship of the peak and TTP on ischemic duration in the modeled reactive hyperemia
agreed with the previously measured characteristics in the healthy group. From a statistical
perspective, however, some of the acquired responses did not entirely agree with the model. ASL
data contain not just metabolic perfusion response but also imaging noise, motion artifacts
introduced by cuff inflation and deflation, and higher order vessel properties causing flow
oscillation during recovery. These details were not considered in the model, which may be the
reason for poor fitting in some cases. Specifically, the mass of blood can be modeled as an
inertance that affects acceleration and deceleration of the blood flow (second-order). Oscillation
of blood pressure and flow can occur when the inertance is matched with compliance. Previously,
Spronck’s model was used to fit individual cerebral flow response averaged by repeated
recordings, which presumably decreased the experimental variability associated with these
influences. Therefore, our model may fit better to averaged individual responses, which should be
more representative of metabolic flow responses. The fitting may be very likely to fail when the
diseased response is very small. However, cases with such small responses may essentially indicate
severe microvascular disease and model analysis is not required.
3.5.2 Major findings
An important contribution of this work is that the model demonstrates the different effects of
arterial stenoses versus microvascular dysfunction on reactive hyperemia. Arterial stenoses can
limit response peak to a great extent. However, there are already many tools to assess arterial
lesions. The use of ASL reactive hyperemia was motivated by the desire to look at microvascular
function. According to the simulation in this study, a shorter ischemic duration can better detect
microvascular dysfunction. This finding reflects the often overlooked fact that the
microvasculature typically reacts quickly to ischemia and a short ischemic protocol can be used to
measure the quickness of reactivity. To date, 5-min cuffing is recommended for reactive hyperemia
measurement due to its lower variability in the peak and TTP when compared with shorter cuffing.
From a signal perspective, our simulation agrees with the viewpoint that the 2-min cuffing
response may be too short to provide enough data points to measure the exact peak difference
caused by arterial lesions. However, our model simulation suggested that the peak and TTP are
more useful for detecting the influence of macroscopic lesions than quantifying microvascular
57
dysfunction. The model clearly indicated that the unique potential of ASL reactive hyperemia for
microvascular assessment is highlighted when using 2-min cuffing. Therefore, this study proposed
the use of a 2-min ischemic protocol for more sensitive microvascular assessment.
The patient data in Chapter 2 was acquired after the establishment of this model. With 2-min
cuffing, the PAD group demonstrated lower peaks and slightly longer TTP than the healthy group
(Figure 3.6), which is similar to the previous findings using 5-min cuffing [37]. This finding is
consistent with the prediction of the model simulation that the presence of arterial stenoses should
be detectable in both protocols.
Curve fitting-based characterization of reactive hyperemia has been reported by other research
groups for skeletal muscle BOLD imaging, bulk flow measurement, and laser Doppler skin
perfusion [68], [85], [86]. These studies use combinations of known mathematical functions, such
as a gamma-variate function. While their functions may also fit ASL reactive hyperemia of the
healthy subjects, the diseased responses may not resemble a gamma-variate function, as
exemplified by response 9 in Figure 3.5. More importantly, an empirical function does not contain
physiological meaning, let alone explaining the dependence of reactive hyperemia on cuffing
duration or the effect of diseases on response characteristics. In contrast, the analysis results using
the current physiological model indicated that some of the model parameters not only clearly
differentiate the responses of the healthy and patient group but may also suggest direct
pathophysiological implications. First, the increased Rp may represent arterial stenoses. Second,
the arterial time constant τp is a function of both resistance and compliance. Since the two factors
are not necessarily tightly correlated and have the opposite effect on τp, a wide range of values in
patients is not surprising. Lastly, the gATP, representing the endothelial sensitivity to ATP, was
significantly lower overall in patients but also had greater variation within the patient group. This
result suggests that different levels of microvascular disease exist within the patient group and may
be assessed by ASL reactive hyperemia with our model-based analysis. The variation in the
microvascular parameter may have critical clinical implications.
It is known that a large portion of the patients with critical limb ischemia does not achieve limb
salvage 1 year after revascularization, which could be caused by more advanced microvascular
diseases. This work strongly demonstrates the potential of using ASL reactive hyperemia for
58
microvascular assessment in such a patient population, which may be useful to identify the
subgroup that requires more effective treatment for their microvascular disease.
Limitations
So far, the model explained only the data from small subject groups. Validation of the model is
lacking due to the fact that no standard microvascular assessment currently exists. Further study
of the microcirculation in particular patients, notably those with diabetes and PAD, may be
revealing. With more data accumulated, the patients can be stratified based on their clinical
indications, and the perfusion indices between subgroups can be compared. Also, the dependence
among model-derived parameters and the correlation between the parameters and clinical indices
may be further examined.
The model description of reactive hyperemia should not be generalized for exercise-induced active
hyperemia, as capillary recruitment in exercise can greatly change the permeability surface area-
product. Also, the model assumes a vascular system of an averaged human, with fixed values for
the arterial input pressure, venous output pressure, size of femoral artery vs volume of the lower
leg, and so on. Individual deviation of these parameters from the human average will cause errors
when we use the model to extract physiological parameters from the perfusion responses. Ideally,
recording the popliteal blood pressure and tibial metabolic rate simultaneously along with
perfusion across the period of ischemia-reperfusion and incorporating the data into the model
would be the best approach to minimize such errors.
Since the heterogeneity between calf muscle groups was not the focus here, perfusion of the whole
mid-calf cross section was used for the establishment of the model. Using perfusion data from
smaller ROIs with higher noise in ASL does not help extract consistent physiology for a general
description of the response to ischemia. Although the model fitted reasonably well to our data, the
quality of ASL data remains critical. Large SNR and minimum motion artifact can reduce bias in
the fitting. The experiments of 2-min thigh cuffing were well tolerated by patients, but repeated
recording could be affected by previous ischemia, especially in patients with reduced washout rate
of metabolites.
59
3.6 Conclusion
The physiological model established in this chapter provided a comprehensive explanation of the
flow regulatory mechanisms involved in ASL reactive hyperemia of calf skeletal muscles. The
model demonstrated that combining the dynamics of intravascular ATP released by RBCs and
interstitial adenosine can explain the evolution of response characteristics with increasing
durations of cuffing in healthy subjects. Model simulation suggested influences of microvascular
dysfunction distinct from those of arterial stenoses on reactive hyperemia. Microvascular
dysfunction may be detected more easily in a 2-min cuffing protocol rather than longer cuffing,
whereas the effect of arterial stenoses may be more consistently observed using 5-min cuffing.
Model-based analysis of the patient data showed that Rp and gATP were significantly different
between the patient and healthy group. While increased Rp in patients is normally expected, the
variation of gATP within the patient group may suggest that different amounts of microvascular
disease existed. Therefore, the next step will be to test the utility of ASL reactive hyperemia and
the model-based analysis in the patients with critical limb ischemia who require assessment of
their microcirculation in addition to their arterial lesions.
60
Appendix
A. The flow equations of the system
The arteriolar resistance Ra and volume Va are defined as:
.+ = lv/4+w ( 3.20 )
(+ = lR4+9 ( 3.21 )
where the constants KR and KV are chosen such that at ra=ra,0 (baseline), Ra=ΔPa/Qr, RaCa = 1 s,
and Va = CaΔPa.
The arterial flow equation Eq. 3.7 is derived from the definition of compliance being a change of
volume due to pressure change:
PN =
M(N
M*N=M(N
MO
MO
M*N ( 3.22 )
M(N
MO= &+5 − &+ ( 3.23 )
where Vp is the volume of the popliteal artery. The venous flow equation Eq. 3.9 is derived
likewise.
The arteriolar flow equation Eq. 3.8 is derived by taking the time derivative of Eq. 3.21 and
expressing the volume change as the difference between the input Qa and output flow Qv:
M(+
MO= lR4+
M4+
MO= &+ − &' =
*N − *+
2.+−
*+ − *'
2(.+ + .') ( 3.24 )
61
B. Modeling the regulatory influences
The stimulus functions of wall-derived regulatory mechanisms are taken from Spronck’s model
[66]. The myogenic stimulus is calculated based on the deviation of arteriolar wall tension T from
the baseline T0, multiplied by a scaling factors smy:
\)x = y)x(3 − 3;) ( 3.25 )
where smy = 3.33 (mmHg⋅cm)−1.
The shear stress is proportional to Q/r3. Therefore, the shear stress stimulus is defined as:
\z{ = yz{
&'
4+|− 1 ( 3.26 )
where ssh is chosen such that at rest the baseline flow and arteriolar radius will result in zero
stimulus.
Likewise, the stimulus functions of metabolite-derived regulatory mechanisms, including the
intravascular ATP and extravascular adenosine, are defined as the deviation from the baseline
concentrations CATP,0 and Cado,0 multiplied by scaling factors:
\"de = y"de(P"de − P"de,;) ( 3.27 )
\+fT = y+fT(P+fT − P+fT,;) ( 3.28 )
where sensitivity coefficients sATP and sado are empirically chosen to be 2.5 and 0.61 µM−1 ,
respectively.
The concentrations of oxygen in the capillary O2c and tissue compartment O2t are the sums of
freely dissolved oxygen and oxygen bound to hemoglobin and myoglobin, respectively:
^9_(*^9_) = 4Phi
*^9_Z
*^9_Z+ *g;,hi
Z + ~i`*^9_ ( 3.29 )
62
^9[(*^9[) = P�i
*^9[
*^9[ + *g;,�i+ ~[`*^9[ ( 3.30 )
where CHb = 2.33mM is the concentration of hemoglobin, CMb = 365µM the concentration of
myoglobin, wb = 80.95% the fraction of water in blood, and wt = 78% the fraction of water in the
tissue. The parameters γ=dO2/dPo2 for Eq. 3.13 and 3.14 are:
c_ = 4Phi
s*g;,hiZ*^9_
ZE?
(*^9_Z+ *g;,hi
Z)9+ ~i` ( 3.31 )
c[ = P�i
*g;,�i
*Ä9[ + *g;,�i+ ~[` ( 3.32 )
With O2t defined, the skeletal muscle metabolic function is calculated:
(^9(^9[) = ()+k,)
^9[
^9[ + l),ÅÇ
( 3.33 )
where Km,O2 = 0.7 µM, and Vmax,m is calculated from the individual baseline perfusion.
The change of intravascular ATP concentration Eq. 3.15 is derived by considering the total amount
of ATP in a single capillary with inflow, release and degradation:
É4_9Ñ_(1 − ℎ[)
MP"de
MO= Ö(1 − ℎf)(P"de,5Z − P"de) + É4_
9Ñ_ℎ[. − 2É4_Ñ_QfP"de
( 3.34 )
where the capillary length lc, flow rate of a single capillary q. Since capillary flow can be derived
from perfusion and capillary volume fraction, this equation can be simplified into Eq. 3.15.
63
C. Definition of microvascular regulatory parameters
This section derives an approximation describing the relationship between the wall tension and
microvascular regulation parameter g and ]. The approximation will lead to a more intuitive,
descriptive definition for the parameters. To start, the change of wall tension over the period of
flow interruption is the time derivative of Eq. 3.3:
M3
MO=M37
MO+M8
MO3) + 8
M3)
MO ( 3.35 )
As explained in the following, only the second term on the right-hand side will be non-zero. Based
on Eq. 3.5 and 3.6, the elastic37 and muscular 3) tension component are functions of the arteriolar
radius 4+. However, the radius is not allowed to change during flow interruption, i.e., M4+/MO = 0;
otherwise, there would be a vacuum created inside the vessels. Therefore, the elastic component
would be:
M37
MO=M37
M4+
M4+
MO= 0 ( 3.36 )
Likewise, the muscular component does not change over time during flow interruption, either.
Therefore, only the activation level of the vascular smooth muscles changes over time:
M3
MO=M8
MO3) =
M8
MW
MW
MO3) ( 3.37 )
To define the parameters related to ATP, we just need to consider the effect of intravascular ATP
on the wall tension. For simplicity, we can break the period of ischemia into multiple intervals, a
few seconds each (a time step smaller than τATP), and make the approximation in one small time
interval at a time. Within the small time window, Eq. 3.11, 3.12, and 3.27 together can be
approximated:
MW
MO= X"de
MY"de
MO≈ X"dey"de
P"de O − P"de(O = 0)
]"de, ( 3.38 )
where P"de(O) represents the concentration function of ATP, t = 0 the beginning of the time
interval, and y"de the sensitivity coefficient. Substituting MW/MO in Eq. 3.37 with Eq. 3.38 will
result in:
64
M3
MO≈ X"de
P"de(O) − P"de(O = 0)
]"de[y"de3)
M8
MW] ( 3.39 )
Because M8/MW changes very slowly when compared to ATP concentration, it can be seen as a
constant when we make the approximation in a small time window. Therefore, Eq. 3.39 essentially
says that the change of wall tension is linear with the change of ATP concentration whereas the
constants in the bracket do not vary with subjects. Hence, we can define X"de as the sensitivity of
the subject’s arteriolar wall to ATP that scales the wall tension response, and ]"de as the response
time to ATP that determines the quickness of wall tension response.
This approximation applies to the adenosine-mediated vasorelaxation as well. Note, however, the
approximation in this appendix is to facilitate the understanding of model behavior and not to be
implemented as such.
65
Chapter 4
Perfusion in different status of limb ischemia This chapter is adapted from the manuscript “Perfusion Measures for Symptom Severity and
Differential Outcome of Revascularization in Limb Ischemia — Preliminary Results With Arterial
Spin Labeling Reactive Hyperemia” authored by Hou-Jen Chen, Trisha Roy, and Graham Wright
and submitted to Journal of Magnetic Resonance Imaging in 2017. This chapter describes a study
that investigated the clinical relevance of previously developed perfusion measures.
4.1 Introduction
Saving limbs from amputation remains challenging in critical limb ischemia (CLI). The limb
salvage rate is less than 70% one year after revascularization in patients with CLI [4]. Meanwhile,
the rate of resolved ischemic pain is approximately only 50% [1]. Currently, no technique can
predict which patient will have a worse outcome. It is hypothesized that the burden of
microvascular disease may contribute to the variability of treatment response in CLI. To
investigate the role of microcirculation in PAD, developing an objective and reliable perfusion
assessment for the peripheral muscles is a critical step.
A variety of imaging techniques applied to limb perfusion has shown valuable clinical implications
recently [30], [46]. In patients with claudication, stress-rest contrast-enhanced ultrasound showed
less exercise perfusion and walking performance in PAD with diabetic microvascular
complications than in PAD alone [30], demonstrating the relationship between functional burden
and microvascular perfusion. In patients with CLI, BOLD reactive hyperemia was shown to
differentiate the post-intervention changes between patients with and without remaining distal
occlusion, which post-intervention ABI did not show [46]. These studies revealed the importance
of perfusion assessment, but quantifying microvascular disease remains challenging. Meanwhile,
other research groups using ASL reactive hyperemia only showed the perfusion differences
between healthy subjects and patients and between pre- and post-intervention [37], [38], which
mostly reflects macrovascular disease.
66
With the measurement and physiological model-based analysis of ASL reactive hyperemia
established in the last two chapters, it is possible to quantify both the macro- and microvascular
disease in PAD patients. A pictorial summary of the macro- and microvascular influences on the
shape of reactive hyperemia responses is shown in Figure 4.1. In brief, arterial resistance and
compliance affect the inflow, reflected in the rising phase of reactive hyperemia; the microvascular
sensitivity determines the overall scale of the response; and the microvascular response time
affects the rate of recovery. These complex characteristics of reactive hyperemia may be related
to different clinical indications of limb ischemia. Therefore, investigating the overall clinical
relevance of reactive hyperemia perfusion response is of great importance.
Figure 4.1: An illustration of physiological influences on the shape of reactive hyperemia, derived from Chapter
3. Reactive hyperemic perfusion is affected by the properties of inflow artery, including the resistance Ra and
compliance Ca, as well as by the microvascular sensitivity gATP and response time τATP to hypoxia induced by 2-min
flow arrest. The notations of arterial parameters are changed from Rp and Cp in the last chapter to Ra and Ca here, and
they should not be confused with arteriole parameters in the last chapter. Moreover, it is more straightforward to use
arterial compliance Ca than τa (=RaCa) as a measure to link to pathophysiology.
In this chapter, the goal is to demonstrate the utility of ASL measures of reactive hyperemia for
assessing the influences of macrovascular occlusive disease and microvascular dysfunction on
67
symptoms, response to revascularization, and short-term outcome in patients with PAD. This study
aimed to: 1) compare ASL measures of reactive hyperemia between the more and less affected
limbs of the same patient; 2) use ASL measures of reactive hyperemia to compare limbs with
different severity defined by standard diagnoses; 3) compare changes in ASL measures of reactive
hyperemia before and after successful peripheral revascularization; and 4) compare short-term
outcome with perfusion indices derived from ASL measures of reactive hyperemia.
4.2 Methods
This study was performed under a protocol approved by the institutional Research Ethics Board.
The patients were recruited from the vascular lab where the ABI and segmental blood velocity
profile provided a diagnosis of PAD. The participants gave their written informed consent. All
participants had culprit lesions in the leg arteries above the mid-calf, confirmed by duplex
ultrasonography and CT angiography. Exclusion criteria included contraindication to MRI and
pre-existing vascular stents in the thigh, which caused concerns about the safety of cuff-induced
compression.
All participants underwent at least one ASL scan for the more affected leg. The more affected leg
was identified based on self-reported symptoms and the physician’s dictation. In a subgroup of
patients, the contralateral less affected leg was also scanned to compare the perfusion
characteristics of the two legs for within-subject relative severity. Meanwhile, all the scanned legs
were categorized as asymptomatic, associated with claudication, or with CLI to compare the
perfusion characteristics for absolute severity.
In a subgroup of patients who were treated with revascularization, outcomes were compared with
pre-intervention reactive hyperemia response measures. Revascularization involved an
endovascular intervention of the iliac, common femoral, superficial femoral, or popliteal arteries.
Technical success of revascularization was determined by X-ray imaging demonstrating adequate
blood flow after restoration of target lesion lumen area. Cases with failure to cross the target lesion
were not followed. In those with technical success, the post-intervention outcome within 6 months
was evaluated based on the vascular lab report and self-reported symptoms, and by phone
interview if the patient did not return for a vascular lab test. Patients were identified as lost to
68
follow-up after 3 consecutive failed phone contacts. The outcome was classified as resolved
ischemia or showing limited improvement. Resolved ischemia was defined as no leg pain
associated with walking in daily life. On the other hand, a limited improvement was indicated by
remaining claudication or CLI after successful revascularization, and the exact change from
revascularization was documented in the follow-up vascular lab report. Meanwhile, pre- and post-
intervention perfusion were compared in patients who had successful interventions that did not
involve stent insertion in the thigh.
Reactive hyperemia
An illustration of the overall methodology was shown in Figure 4.2. The experimental settings for
the acquisition of reactive hyperemia were the same as described in Section 2.4. In brief, all
subjects underwent 2 minutes of flow arrest and were imaged with FAIR at 3 Tesla using a cardiac
receive coil array placed at the calf region. The recording started 1 minute before cuff inflation
and was repeated for 51 pairs of labeled and control images over 5 minutes and 41 seconds, with
a temporal resolution of 3.35 s. At the end of ASL, a baseline image for signal normalization was
acquired with the same readout sequence. The perfusion in the region of interest manually drawn
to cover the entire muscle region was quantified after image reconstruction in MATLAB.
69
Figure 4.2: An illustration of the acquisition and model-based characterization of ASL reactive hyperemia.
Perfusion characterization
Similar to Section 3.3, all perfusion time curves were characterized by the apparent peak, time to
peak (TTP), and the five model parameters that resulted in the best fit between the model-curve
and measured response. The best fit was estimated by searching the parameter space for those that
yielded the least mean square error between the fit and the data for the first minute of perfusion
response, within the predefined parameter ranges: 1 ≤ Ra ≤ 5; 0.3 ≤ Ca ≤ 2; 1 ≤ gATP ≤12; 12 ≤ τATP
≤ 48 (s); and 3 ≤ fr ≤ 10 (mL/100g/min).
Statistical analysis
Statistical analysis was performed in Prism 7. The measurement indices, including perfusion
indices and ABI, were compared by two-sided Wilcoxon matched-pairs signed rank tests between
the pair of legs when both legs of the patient were scanned and between pre- and post-
revascularization on the leg receiving intervention when both measures were available. The
measurement indices in different severity categories were compared by a two-sided Kruskal-
Wallis test and Dunn’s post hoc test. The measurement indices between the outcome groups were
assessed by Mann-Whitney tests. In all comparisons, Holm-Sidak adjustment was used to maintain
70
a family-wise error rate of 0.05 for the families of empirical and model-derived perfusion indices.
A p < 0.05 was deemed statistically significant. Data are presented as means ± SD unless otherwise
stated. Meanwhile, the frequencies of having cardiovascular risk factors (hypertension,
hyperlipidemia, tobacco use, diabetes) and CLI were compared between the outcome categories
by a two-sided Fisher’s exact test.
4.3 Results
A total of 21 consecutive patients were enrolled for the study of ASL reactive hyperemia. Two
patients withdrew before the perfusion exam. The remaining 19 patients (demographics in Table
4.1) all completed the 2-minute ischemia test and tolerated it well without complaint of discomfort.
A diagram of participation is shown in Figure 4.3. The first nine patients had both legs scanned
for bilateral comparison; the rest only had the index leg examined. Due to a coil channel issue, the
responses in two index legs of the bilateral exams were not recorded in the desired settings and
were excluded from analysis and comparison, but their contralateral limbs were included in the
absolute severity comparison. One patient had intermittent leg motion during the perfusion scan,
and the associated data were excluded. In total, perfusion responses were properly recorded in 25
legs from 18 patients, including 7 pairs of legs.
Table 4.1: Demographics of the participants.
Patients (n=19)
Age, years (median, range) 71 (51-86)
Male to female ratio 9:10
Smoker 15
Hypertension 14
Hyperlipidemia 14
Diabetes 5
Severity of the index leg
Claudication 13
Critical limb ischemia 6
Endovascular revascularization was performed in 15 out of the 16 index legs with a proper prior
recording of reactive hyperemia; the remaining patient underwent exercise therapy. The time
71
between the perfusion exam and revascularization ranged from 0 days to 4 months (median time
= 1 day). All the target lesions were above the mid-calf, including two patients with occlusion in
the iliac arteries and the rest with lesions in the superficial femoral artery and/or popliteal artery.
Eleven legs had technically successful revascularization; in the remaining 4 legs, the
interventionist failed to establish blood flow across the target lesions in the endovascular
procedure. Eight patients had stent insertion, including the two with iliac occlusions. Five of the
successfully vascularized patients came back for the post-procedure perfusion exam within 3 days
to 5 weeks after their procedure.
Figure 4.3: A diagram of patient participation.
The follow-up outcome was collected in 10 of the 11 successfully vascularized patients at a time
ranging from 4 days to 6 months after the procedure, including 1 patient who reported excellent
recovery on the phone at 6 months and 9 patients who were assessed by the vascular lab exam and
self-reported their walking ability or leg pain (see Appendix Table 1).
72
Bilateral comparison
In the subgroup of seven subjects where both legs were scanned, five patients had claudication-
asymptomatic leg pairs, and the other two patients had CLI-claudication leg pairs. Greater
perfusion responses were consistently observed in the less affected legs (Figure 4.4a). The ABI
values in the more affected legs were significantly lower than the values in the contralateral legs
(Figure 4.4b). Reactive hyperemia also showed lower peak perfusion, model-derived resistance,
and baseline perfusion in the more affected legs than in the contralateral legs. Other perfusion
characteristics did not exhibit significant differences in the bilateral comparison.
73
Figure 4.4: Bilateral comparison. (a) The recorded reactive hyperemia. (b) ABI and perfusion characteristics
extracted from the recorded waveform. The native p values by Wilcoxon matched-pairs signed rank test are shown.
74
Comparison between severity categories
The ABI was found to decrease with increased symptom severity, and the peak perfusion and
model-derived macrovascular resistance agreed with this finding (Figure 4.5a). Other perfusion
indices did not show significant differences. When the arterial resistance was large, the perfusion
waveform became flat (Figure 4.5a), and the peak might appear at a random time within the
hyperemic interval due to noise, which led to large standard deviations of TTP in the groups.
Figure 4.5: Perfusion characteristics (a) and averaged waveforms (b) for different symptom severities. The
native p values by Kruskal-Wallis tests are shown.
75
Pre- and post-revascularization perfusion
The post-revascularization measurement was limited to the patients without vascular stents in their
thigh. As shown in Figure 4.6, the peak perfusion reached a fairly adequate level of 40
mL/100g/min in all five subjects after revascularization, which was close to the level in healthy
subjects (refer to Section 2.4). However, the changes associated with revascularization were very
small in patients who had fairly normal pre-intervention perfusion. None of the perfusion indices
were significantly different in the statistical comparison by Wilcoxon matched-pairs signed rank
tests, potentially due to the small number of subjects. However, the peak response and arterial
resistance had differences close to statistical significance (p = 0.0625 for the measures). The last
two patients shown in Figure 4.6 had low gATP pre- and post-intervention, and they also
experienced limited improvement at 6 months post-intervention.
Figure 4.6: Individual reactive hyperemia before and after revascularization.
0 20 40 60 80
0
10
20
30
40
50
s
mL/
100g
/min
0 20 40 60 80
0
10
20
30
40
50
s
mL/
100g
/min
0 20 40 60 80
0
10
20
30
40
50
s
mL/
100g
/min
0 20 40 60 80
0
10
20
30
40
50
s
mL/
100g
/min
0 20 40 60 80
0
10
20
30
40
50
s
mL/
100g
/min
Pre− Post−revascularization
76
Perfusion versus revascularization outcome
Outcome assessment was available in ten of the eleven successfully revascularized patients. In the
nine patients who returned for a follow-up vascular lab exam, the ABI significantly increased from
0.62 ± 0.14 before to 0.89 ± 0.23 after revascularization (Wilcoxon matched-pairs test; p = 0.0039).
Four patients had remaining leg pain associated with walking, including two with previous CLI.
Six patients reported excellent walking ability without limitation, including one with previous CLI.
The empirical perfusion indices, i.e. peak amplitude and TTP, did not show a clear separation
between the outcome groups (Figure 4.7). In contrast, model-based analysis showed a consistently
low microvascular sensitivity gATP in those with worse outcome. In other words, gATP derived from
pre-intervention reactive hyperemia predicted the revascularization outcome. Other perfusion
indices and ABI did not differentiate the outcomes (Table 4.2). The univariate analysis of
cardiovascular risk factors did not show a significant difference in the frequencies of diabetes,
hypertension, hyperlipidemia, and tobacco use between the two outcome groups (Fisher’s exact
test; p = 0.9999 for each risk factor). Based on the very small subject groups, having CLI did not
affect the outcome (Fisher’s exact test; p = 0.5), but the p value was smaller than those in the test
of risk factors. Gender and age were not significant between the two outcome groups.
Figure 4.7: A scatter plot of perfusion indices in the two outcome groups.
0 20 40 60 800
10
20
30
40
50
Time to peak (s)
Peak
per
fusi
on (m
L/10
0g/m
in)
Apparent characteristics
0 2 4 60
2
4
6
8
10
12
Arterial resistance (Ra)
Mic
rova
scul
ar s
ensi
tivity
(g−a
tp)
Model indices
Ischemia resolved Limited improvement
77
Table 4.2: Comparison of measurement indices between outcome groups
Measurement indices Ischemia resolved (n=6)
Limited improvement (n=4)
Adjusted p value
ABI 0.75 ± 0.20 0.67 ± 0.32 0.6095
Peak perfusion (mL/100g/min) 30.33 ± 5.82 28.42 ± 10.07 0.6095
Time to peak (s) 30.33 ± 29.51 7.80 ± 2.72 0.3600
Arterial resistance Ra 3.16 ± 1.14 2.50 ± 1.19 0.7284
Arterial compliance Ca 0.89 ± 0.65 0.50 ± 0.17 0.7284
Microvascular sensitivity gATP 8.72 ± 1.46 4.93 ± 0.91 0.0466 *
Microvascular response time !ATP (s) 27.80 ± 15.96 29.37 ± 16.52 0.7619
Baseline perfusion fr (mL/100g/min) 6.41 ± 2.68 4.07 ± 2.22 0.5286
Mann-Whitney test with Holm-Sidak adjustment for family-wise error rate. * indicates significance.
In Table 4.2, the ABI and peak perfusion had the same p values to the 4 decimals potentially
because the values were quantized due to the small sample sizes. On the other, the same adjusted
p values for Ra and Ca occurred because by definition it can occur. The adjusted p values in a
family of comparisons were calculated successively, starting from the comparison with the small
raw p value. The adjusted false positive rate of the subsequent comparison is the higher one
among the current and previous adjustment (see the following algorithm for multiplicity adjusted
p values).
According to the guide in the statistics software Prism:
The equations in the following are used to compute the adjusted p value padj(i) from the raw p
value p(i):
padj(1) = 1 - (1 - p(1))k
padj(2) = max ( padj(1) , 1 - (1 - p(2))k-1 )
..........
padj(k) = max ( padj(k-1) , 1-(1-p(k))k-k+1 ) = max ( padj(k-1) , p (k) ) ,
where the raw p values are sorted so p(1) is the smallest, k is the number of comparisons, and
max is a function that returns the larger of two values. Note that in some cases successive
adjusted p values will be identical, even when the raw p values are not.
78
4.4 Discussion
The study in this chapter is the first to demonstrate that ASL-measured calf muscle perfusion is
tied to the clinical indications of limb ischemia and has a discriminatory power down to the patient
subgroup level. The characteristics of reactive hyperemia contain multiple aspects of the
pathophysiology that can be disentangled by the model-based analysis. The peak perfusion and
model resistance reflected the macrovascular disease burden in accord with the results assessed by
the ABI. On the other hand, microvascular indices revealed additional variables in reactive
hyperemia that may be associated with the microvascular function. Measures of macrovascular
resistance were related to the manifested symptom severity, whereas a microvascular dysfunction
index seemed to reflect prognosis following revascularization.
There are two alternative MRI-based perfusion techniques, including the blood oxygenation level-
dependent (BOLD) and dynamic contrast-enhanced (DCE) imaging. ASL is advantageous
primarily in that it offers temporally resolved absolute perfusion, which can be directly interpreted
in the context of flow resistance and vascular reactivity. BOLD signal is a mixed result of perfusion,
metabolism, and oxygen extraction [45], [46]. DCE requires the use of gadolinium-based contrast
agents, which is a concern in patients with impaired renal function, and the perfusion is usually
estimated by tracer kinetic modeling [41], [87] for steady states instead of dynamic responses. The
major disadvantage of ASL is that the low SNR can affect the characterization of perfusion.
However, the novel model-based analysis strategy described in the last chapter has markedly
reduced the influence of noise on the characterization.
The arterial resistance increased with symptom severity in the bilateral comparison and in the
comparison between severity groups; the standard macrovascular assessment by the ABI agreed
with these results. Therefore, the ischemic symptom severity seemed to be directly linked to the
burden of macrovascular disease. The peak perfusion also changed consistently with symptom
severity, most likely due to the high sensitivity of reactive hyperemia to macrovascular disease.
On the other hand, the change of TTP with symptom severity was less consistent than the change
in resistance measures. The TTP may be a function of the resistance and compliance of the inflow
arteries. Since the compliance can be affected by the degree of calcification, which is not directly
related to the perceived symptom severity, a higher variability of TTP among the severity groups
is expected.
79
The reduction of macrovascular resistance by revascularization was observed in all patients with
follow-up ABI. The peak perfusion also increased in the subjects who participated in post-
intervention perfusion assessment. However, improvement of reactive hyperemia was small in two
patients with fairly adequate peak perfusion before revascularization. One possibility of having
adequate perfusion in patients with ischemic symptoms was that the site of ischemia did not
include the mid-calf. Although claudication is most common in the calves, it can also affect the
feet, thighs, hips, or buttocks. Also, there are other factors contributing to claudication in addition
to macrovascular obstruction, such as local inflammation, vascular dysfunction, altered muscular
metabolism, and myofibril atrophy [88], [89]. Therefore, although the relationship between
macrovascular disease and symptom severity was established at a group-wise level, individual
symptoms may be affected by other variables.
The microvascular indices are a novel characterization of reactive hyperemia derived from the
physiological modeling of the microvascular endothelial response to hypoxia. A lowered
microvascular sensitivity gATP may indicate microvascular endothelial dysfunction. The correlation
between microvascular dysfunction and poorer functional outcome discovered in this study reflects
the well-known prognostic relevance of endothelial dysfunction [90]-[92]. Systemic endothelial
dysfunction, typically measured by the brachial artery flow-mediated dilatation (FMD) [93], has
been shown to be a systemic independent risk factor for future cardiovascular events [91], [92].
More specifically, coronary microvascular dysfunction, identified by reduced flow reserve in
patients without overt arterial occlusion [13], [14] or by invasive coronary reactive hyperemia [25],
has been shown to predict adverse cardiac events. Therefore, the current finding that limb-specific
microvascular dysfunction predicted the functional outcome of peripheral revascularization seems
to be reasonably aligned with ongoing research of cardiac disease pathophysiology [94]-[96]. The
mechanistic reasons for the association of microvascular dysfunction and prognosis are not fully
understood. Microvascular dysfunction may reflect elevated inflammation, increased reactive
oxygen species, microvascular thrombosis, or reduced synthesis of vasoactive substances [97],
which can directly impair microvascular perfusion [98] or accelerate the progression of
macrovascular disease [99], ultimately leading to a worse outcome than for those without
significant microvascular dysfunction. We cannot rule out the possibility that unsatisfactory
improvement following intervention is due to residual macrovascular plaques in the more distal
arteries; however, the higher chance of having more distal lesions, with more calcification and
80
more severe symptoms at the first presentation of disease, may still be related to the contribution
of microvascular disease.
Since the results revealed a close relationship between gATP and the post-intervention functional
outcome, a future study specifically designed to validate the predictive value of ASL measures of
reactive hyperemia is desirable. The current functional outcome was a subjective measure reported
by the patients. A more definitive outcome can be assessed by hard measures including repeated
intervention, presentation of tissue loss, and amputation. Moreover, because the patients with CLI
are more likely to have a variable outcome, targeting CLI for a prognosis study may be more
clinically relevant [97], [100]. Therefore, a prospective study with a larger sample size exclusively
in CLI and with a longer follow-up period for composite outcome measures will be a valuable next
step following the current findings.
Limitations
This study was primarily designed to investigate the relationship between clinical indications and
perfusion characteristics and not to test the predictive power of ASL perfusion on prognosis. To
fully explore the predictive power, a clinical study with 100 patients and 5-year follow-up may be
required [100]. It is not clear if vascular stents are safe for cuffing. We recruited patients
consecutively without any presumption of clinical relevance to a specific population and lesion
types. Therefore, the resulting significant portion of patients with SFA stenting as part of the
revascularization procedure limited the availability of post-revascularization reactive hyperemia
studies. However, pre-intervention ASL reactive hyperemia was the main focus here. This study
did not segment the calf cross-section into muscle groups because the perfusion was compared
with ABI for the perceived symptoms and functional outcome of above-calf revascularization.
None of these factors differentiate the calf muscle groups. Also, the SNR of ASL can be reduced
significantly for smaller regions-of-interest, which may limit the quality of perfusion
characterization.
4.5 Conclusion
This work showed that the characteristics of reactive hyperemia reflect multiple aspects of the
pathophysiology. Measures of arterial resistance are related to the manifested symptom severity,
81
whereas microvascular dysfunction indices are related prognosis following revascularization. A
future study with a larger patient population and longer follow-up period will be required to
confirm the current findings regarding the disease prognosis.
82
Appendix Table 1: Successfully revascularized patients
Sex Age Disease description ABI Pre
ABI Post
Procedure Follow up
Outcome description Resolved
1 F 51 CLI, non healing ulcer, focal SFA occlusion
0.79 1.06 SFA angioplasty † 6 weeks Ulcer healed, calf claudication greatly improved. Y
2 F 83 Pain with 1/8 mile walking, bulky iliac lesion
0.52 - CIA stenting † 6 months Excellent improvement reported by phone Y
3 F 81 Pain with 5 min walking, mid SFA occlusion
0.64 1.13 SFA angioplasty † 10 weeks Walking without pain Y
4 F 73 Pain with few hundred yards walking, multilevel disease
0.71 1.16 SFA angioplasty, PA stenting 1 month Claudication resolved, excellent walking Y
5 F 77 CLI, rest pain, multilevel disease 0.31 0.45 EIA angioplasty and stenting † 6 months Improved but still had rest pain. 2nd intervention at 11 months
N
6 M 76 Pain with 100 m walking, SFA and tibial occlusions, calcification
0.71 0.91 SFA angioplasty † 5 months Pain improved, but walking limited to 100-150 m
N
7 M 63 Pain with 50 yards walking, multilevel disease
0.7 0.75 SFA and PA angioplasty and stenting
6 months Great improvement Y
8 M 60 Pain with 300m walking, long heterogeneous SFA lesion
0.53 0.97 SFA and PA angioplasty and stenting
1 month Walking without pain Y
9 M 85 CLI, rest pain, multilevel disease, calcification
0.67 0.97 CFA, SFA, and PA angioplasty, stenting
5 months Pain relieved, walking limited to 5-10 min N
10 M 70 Pain with 5-10 min walking, long heterogeneous SFA occlusion
0.58 0.68 SFA angioplasty and stenting 4 days Complaint of leg pain N
CIA = common iliac artery, EIA = external iliac artery, CFA = common femoral artery, SFA = superficial femoral artery, PA = popliteal artery. † denotes post-revascularization ASL recorded
83
Chapter 5
Summary and future work
5.1 Summary In Chapter 2, a series of experiments was performed to optimize ASL for the measurement of calf
muscle reactive hyperemia, followed by an investigation of the physiological relevance of
perfusion characteristics with varying durations of induced ischemia. The ASL signal behavior
and tibial blood velocity throughout an ischemia-reperfusion paradigm were examined in a group
of healthy subjects. The results showed that the pseudo-continuous ASL data was associated with
large background contamination and highly variable labeling efficiency. The tibial blood velocity
favored the use of pulsed ASL, which has labeling efficiency fairly independent of blood velocity.
Hence, a pulsed ASL configuration named FAIR was used to characterize reactive hyperemia
induced by varying ischemia duration in healthy subjects (n=7). The healthy group showed
appreciable reactive hyperemia with just one minute of ischemia, with the peak responses slightly
increasing for longer ischemic durations. Calf reactive hyperemia was also characterized in
patients (n=9) with PAD where a 2-min period of ischemia was used. The patient group had
significantly lower peak responses to 2-min ischemia than the healthy group (26.9 ± 6.3 vs 44.7 ±
6.8 mL/100g/min, p<0.0001). The study in this chapter demonstrated the capacity of the FAIR
sequence to characterize reactive hyperemia that reflects microvascular regulation for ischemic
stress and to delineate the influences of PAD on calf perfusion.
In Chapter 3, a physiological model was established to interpret the ASL measures of reactive
hyperemia. The theoretical framework incorporated oxygen transport, tissue metabolism, and
vascular regulation mechanisms. The simulation demonstrated distinct effects between arterial
stenoses and microvascular dysfunction on reactive hyperemia and indicated a higher sensitivity
of 2-minute thigh cuffing to microvascular dysfunction than that of 5-minute cuffing. The recorded
perfusion responses in PAD patients were better differentiated from the normal subjects using the
model-based analysis rather than by characterization using the apparent peak and time-to-peak of
the responses. The analysis results suggested different amounts of microvascular disease within
the patient group. Overall, this work demonstrates a novel analysis method and facilitates the
84
interpretation of physiology involved in ASL measures of reactive hyperemia. ASL assessment of
reactive hyperemia with model-based analysis may be used as a noninvasive microvascular
assessment in the presence of arterial stenoses, allowing us to look beyond the macrovascular
disease in PAD. The findings in this chapter motivated the subsequent work to explore the clinical
relevance of the model-derived perfusion measures.
In Chapter 4, reactive hyperemia was characterized in a group of consecutively recruited patients
to study the relationship between the perfusion measures and clinical indications of limb ischemia.
Specifically, the relationship between indices of macrovascular disease and microvascular
dysfunction and symptoms, response to revascularization, and short-term functional outcome were
assessed. In total, ASL reactive hyperemia was properly recorded in 25 limbs from 19 patients,
including the recording of 7 pairs of limbs. The perfusion measures were found to be sensitive to
the relative severity of disease symptoms in the bilateral comparison, with the more symptomatic
leg having lower peak perfusion and higher arterial resistance (n=7). When categorizing all the
limbs as asymptomatic, claudication, and critical limb ischemia for comparison, the peak perfusion
decreased and arterial resistance increased with symptom severity (n=25). The ABI also decreased
with symptom severity in the bilateral and category comparisons.
Revascularization of the index limb was successful in 11 patients, validated by increased ABI. The
subsequent functional outcome in 10 of these patients within 6 months was also evaluated. Six
patients had ischemic symptoms resolved, whereas the other 4 patients had remaining exercise-
related leg pain. Only the model-derived microvascular sensitivity gATP pre-intervention showed a
significant difference, whereas the ABI, other perfusion measures, and cardiovascular risk factors
were not significantly different between the outcome groups. Therefore, the functional outcome
were related to model-derived microvascular measure obtained from perfusion before
revascularization.
Overall, this work proved that the characteristics of reactive hyperemia reflect multiple aspects of
the pathophysiology. Measures of arterial resistance are related to the manifested symptom
severity, whereas microvascular dysfunction indices may predict prognosis following
revascularization. Hypothetically, the subgroup of patients who will have a poor prognosis after
revascularization may be those with more severe microvascular disease; therefore, they may be
identified by the perfusion measures derived from ASL assessment of reactive hyperemia with the
85
model-based analysis. A future study with a larger patient population and longer follow-up period
will be required to confirm the current findings.
5.2 Experimental improvement The data quality of the ASL perfusion signal is very critical in this study and can be improved by
using better equipment. First, a dedicated peripheral coil with smaller elements can provide more
uniform sensitivity around the limb and higher SNR. The phased arrays used in this study had a
lower SNR for the lateral and medial portions than the anterior and posterior portions of the calf
muscles. With more uniform sensitivity and higher SNR, characterization of perfusion in separate
calf muscles will be possible. Moreover, higher SNR also provides room for acceleration of image
acquisition, which may be important in scenarios where a multislice acquisition is required.
Another improvement that can be made is the cuffing procedure. In this study, limb swelling and
motion artifact were inevitable because the homemade cuff system could not inflate or deflate
rapidly. A dedicated pneumatic tourniquet system (Hokanson Inc.) made to inflate and deflate
vascular cuffs within 1 second can be used to minimize the calf volume change throughout the
cuffing experiment. Lastly, thigh compression can induce knee flexion, so fixing the limb entirely
with additional devices can also be implemented to reduce motion artifact.
5.3 Future directions: variations of ASL for different applications An objective, noninvasive measure of tissue perfusion in the lower limb would be invaluable for
the management of PAD. Although this thesis focuses on the demonstration of reactive hyperemia
and quantification of microvascular disease, ASL assessment of perfusion may potentially be used
in different contexts. First, mapping the areas of poor perfusion with ASL may allow more
informed planning of the intervention, which is particularly useful in tibial or small artery
revascularization where the local distribution of blood flow, including the contribution of collateral
flow, is important. Second, mapping perfusion may allow selective targeting of injections of new
therapeutic agents, which can help, for instance, in the research and development of angiogenesis
therapies. Third, perfusion assessment to characterize post-intervention improvement may be more
objective and target-specific than assessment with perceived symptoms or bulk flow measures.
86
Lastly, when the cause of limb symptoms is uncertain due to co-existence of other diseases or
injuries, absolute perfusion by ASL may help in the diagnosis. That being said, depending on the
clinical purpose, ASL sequences may need to be optimized to suit the question.
5.3.1 Vascular territory mapping
In the context of revascularization of tibial or foot arteries for wound healing, it is important to
consider the angiosome theory. An angiosome is a perfusion territory fed by a single-source artery
[101]. Therefore, the vessels to be treated may be determined by the location of the wound in
addition to the morphological quality of the vessels on X-ray images. However, noninvasive
individual angiosome mapping is currently not available. Some ASL techniques used in cerebral
perfusion territory mapping may be adapted for angiosome imaging of the lower extremity. For
example, Golay et al. reported a pulsed ASL regional perfusion technique that had more selectivity
of the labeling region [102]. The sequence allowed for a prescription of the labeling region to
include one artery of interest at a time based on the pre-acquired MR angiogram. Translating such
a technique to peripheral applications seems straightforward. However, distinguishing the vascular
territories usually requires an experiment with multiple repetitions and assessment of perfusion in
a steady state. Whether the technique is sufficiently sensitive to the resting perfusion of the skeletal
muscles within an acceptable scan time remains unclear. Alternatively, mapping the territories
during hyperemia by interleaving the labeling of different vessels may be explored. There are also
pCASL-based territory mapping techniques [103], but as discovered in Chapter 2, continuous
labeling may not be reliable in the peripheral vessels.
5.3.2 Multislice peripheral perfusion
Since plaque distribution is not uniform and localized ischemia may exist in a particular segment
of the leg, it is desirable to probe perfusion in multiple segments of the calf and the foot. In this
case, sampling a few image slices over the entire lower leg may be more suitable than tightly
sampling many slices within one short section of the leg. However, most existing multislice ASL
methods are designed for the latter situation where multislice or volumetric readout is carried out
following blood labeling in a single region. This approach may be inadequate because the lower
87
leg blood velocity may be too slow for blood to reach the slices more distal from the same labeling
region. An ideal multislice ASL for peripheral perfusion would require simultaneous multiband
labeling and multislice acquisition. Multiband adiabatic inversion pulses have been designed,
primarily for use in MR spectroscopy, for B1-insensitive excitation [104], [105], and can be
translated to ASL for blood labeling. Simultaneous multislice acquisition techniques are also well
established in ASL [106]-[108] and provided by several vendors. Therefore, a multislice ASL for
peripheral use seems attainable.
5.4 Future directions: clinical studies
5.4.1 Validation of microvascular disease with invasive measurement
This thesis provided a novel methodology to assess peripheral macro- and microvascular disease.
While the macrovascular disease derived from perfusion was found to be consistent with the ABI,
the model-derived microvascular disease component was a novel measure. There is currently no
other comparable noninvasive measurement. To validate the result, we may compare the
perfusion-derived microvascular sensitivity with the index of microcirculatory resistance (IMR),
an invasive measure originally developed for assessing the coronary microcirculation independent
of coronary arterial stenosis [109], [110]. IMR is acquired by recording the coronary blood
pressure with pressure wires and the blood flow transit time derived from the thermodilution
technique. An elevated IMR represents microvascular dysfunction and predicts poor long-term
outcome [111]. Moreover, the pressure wire can also be used to measure the pressure gradient
across an arterial stenosis and to derive fractional flow reserve (FFR), the gold standard for the
influence of macrovascular disease. The fact that IMR and FFR measured in the same vessel reflect
different levels of vascular disease is very similar to the results using peripheral perfusion with the
model analysis in Chapter 3. Therefore, implementing an IMR measurement for the leg arteries at
the time of intervention and comparing this invasive measure to perfusion may be a way to validate
the findings of peripheral microvascular dysfunction and confirm its association with post-
intervention outcome.
88
5.4.2 The influence of diabetes on PAD
Diabetes is characterized by chronic hyperglycemia, dyslipidemia, and insulin resistance. These
conditions affect the functions of endothelial cells, vascular smooth muscle cells, and platelets,
thereby stimulating vasoconstriction, inflammation, and thrombosis [112] and fostering the
progression of atherosclerosis. The rates of PAD are increased by 2- to 4-fold in individuals with
diabetes, and their peripheral vascular disease is associated with greater severity, more diffuse
plaques, more calcification, and more distal lesions [113], [114]. Presumably, there is a different
microvascular disease burden in diabetic vs non-diabetic PAD. It would be interesting to test the
sensitivity of the perfusion measures to such a difference.
Diabetes also affects the outcome of limb ischemia, potentially because the diabetic vasculature is
more vulnerable to ischemic injury. CLI patients with diabetes have a higher rate of major
amputation than do non-diabetic CLI patients [115]. It has been shown that diabetic CLI patients
could benefit from early revascularization [116], but multiple procedures and close surveillance
are required. To better manage diabetic PAD in addition to revascularization, medical
interventions targeting specific pathophysiological mechanisms are being explored [112], [117].
Elevated fasting glucose was associated with lower patency and increased major adverse limb
event rates after revascularization [118]. Some trials showed that intensive glucose lowering had
very limited effect on cardiovascular outcome, including the rates of peripheral vascular events
[119], [120], but a recent trial that involved the use of a drug to increase urinary glucose excretion
showed beneficial effect on cardiovascular outcome (peripheral events were not considered). It is
not clear how glucose lowering, or other medical interventions, would control or improve diabetic
microvascular disease, with ischemic injury caused by PAD in particular. Applying a perfusion
assessment may help clarify the interactions between medical interventions and microvascular
disease. Medical interventions may improve microvascular function leading to long-term benefits.
5.4.3 Evaluation of therapeutic effect
While most patients with severe limb ischemia can be revascularized, the presence of irreversible
gangrene, the absence of a target vessel, and the lack of a venous graft can limit the effectiveness
of revascularization. Such a patient population is deemed to have no treatment option and will
likely be subjected to amputation eventually. Meanwhile, patients responding to revascularization
89
poorly, such as those with diabetes and CLI, may require additional treatments to control or
improve their conditions. Therefore, new therapies are necessary for no-option patients and
patients with poor outcome.
Pro-angiogenesis approaches, delivering vascular growth factors or bone marrow-derived cells
through intra-arterial infusion or intramuscular injections, have been developed by many research
groups and have undergone many trials [121]-[123]. As reviewed, these clinical trials failed to
demonstrate significant improvement in limb symptoms, including pain and wound healing [123].
These studies used hard outcome measures and conventional assessments such ABI and TcPO2,
which are poorly suited for detecting microvascular changes induced by angiogenesis if it ever
occurs. A more advanced microvascular assessment technique, such as the one developed in this
thesis, may provide greater clarity of the intermediate processes between the treatment and
outcome, which can be used to improve the therapies by optimizing the dose, frequency of
administration, and distribution of the therapeutic agents and for identifying more effective agents.
The methods described in this thesis may also be adopted to identify the patients with substantial
microvascular dysfunction who may be more responsive to microvascular therapies relative to
those with severe arterial disease.
90
5.5 Final words This thesis described methods to quantitatively evaluate the microcirculation in situations of limb
ischemia using MRI. Perfusion changes were found to reflect the combined effects of arterial
stenosis and microvascular dysfunction. A clinical study that involves more patients and a longer
follow-up period for definitive outcome measures is required to confirm the predictive power of
the perfusion-derived microvascular measure for revascularization prognosis.
The signal-to-noise ratio in ASL is critical to the overall methodology. Receive coils with higher
and more uniform sensitivity are required for imaging more details such as perfusion territory of
vessels and perfusion of individual muscles. Single-slice perfusion imaging can be extended to
multislice imaging by incorporating simultaneous multislice techniques for both the labeling and
acquisition. These advances may be valuable for broader evaluation of limb ischemia and
improving revascularization for small and distal vessels. Meanwhile, peripheral perfusion
assessment may be applied to other clinical studies to evaluate the efficacy of new therapies or to
identify the patients who are highly susceptible to microvascular disease. Finally, saving limbs
from amputation in patients with poor response to existing treatments or with no options may
require multidisciplinary approaches where a noninvasive assessment technique of perfusion
should play a crucial role in the conceptualization, development, and utilization.
91
References [1] L. Norgren, W. R. Hiatt, J. A. Dormandy, M. R. Nehler, K. A. Harris, F. G. R. Fowkes,
TASC II Working Group, K. Bell, J. Caporusso, I. Durand-Zaleski, K. Komori, J. Lammer, C. Liapis, S. Novo, M. Razavi, J. Robbs, N. Schaper, H. Shigematsu, M. Sapoval, C. White, J. White, D. Clement, M. Creager, M. Jaff, E. Mohler, R. B. Rutherford, P. Sheehan, H. Sillesen, and K. Rosenfield, “Inter-Society Consensus for the Management of Peripheral Arterial Disease (TASC II).,” presented at the European journal of vascular and endovascular surgery : the official journal of the European Society for Vascular Surgery, 2007, vol. 33, pp. S1–75.
[2] M. Lovell, K. Harris, T. Forbes, G. Twillman, B. Abramson, M. H. Criqui, P. Schroeder, E. R. Mohler, and A. T. Hirsch, “Peripheral arterial disease: Lack of awareness in Canada,” Canadian Journal of Cardiology, vol. 25, no. 1, pp. 39–45, 2009.
[3] A. Fronek, M. Coel, and E. F. Berstein, “Quantitative ultrasonographic studies of lower extremity flow velocities in health and disease.,” Circulation, vol. 53, no. 6, pp. 957–960, Jun. 1976.
[4] H. H. Dosluoglu, S. Davari-Farid, L. Pourafkari, L. M. Harris, and N. D. Nader, “Statin use is associated with improved overall survival without affecting patency and limb salvage rates following open or endovascular revascularization,” Vasc Med, vol. 19, no. 2, pp. 86–93, May 2014.
[5] D. T. Baril, L. K. Marone, J. Kim, M. R. Go, R. A. Chaer, and R. Y. Rhee, “Outcomes of endovascular interventions for TASC II B and C femoropopliteal lesions.,” J. Vasc. Surg., vol. 48, no. 3, pp. 627–633, Sep. 2008.
[6] D. T. Baril, R. A. Chaer, R. Y. Rhee, M. S. Makaroun, and L. K. Marone, “Endovascular interventions for TASC II D femoropopliteal lesions.,” J. Vasc. Surg., vol. 51, no. 6, pp. 1406–1412, Jun. 2010.
[7] B. H. Gray, A. A. Grant, C. A. Kalbaugh, D. W. Blackhurst, E. M. Langan III, S. A. Taylor, and D. L. Cull, “The Impact of Isolated Tibial Diseaseon Outcomes in the Critical Limb Ischemic Population,” Annals of Vascular Surgery, vol. 24, no. 3, pp. 349–359, Apr. 2010.
[8] M. Pardo, M. Alcaraz, F. Ramón Breijo, F. L. Bernal, J. M. Felices, and M. Canteras, “Increased transcutaneous oxygen pressure is an indicator of revascularization after peripheral transluminal angioplasty,” Acta Radiol, vol. 51, no. 9, pp. 990–993, Nov. 2010.
[9] Z. Wang, R. Hasan, B. Firwana, T. Elraiyah, A. Tsapas, L. Prokop, J. L. Mills, and M. H. Murad, “A systematic review and meta-analysis of tests to predict wound healing in diabetic foot.,” J. Vasc. Surg., vol. 63, no. 2, pp. 29S–36S.e1–2, Feb. 2016.
[10] J. J. Paszkowiak and A. Dardik, “Arterial wall shear stress: observations from the bench to the bedside.,” Vasc Endovascular Surg, vol. 37, no. 1, pp. 47–57, Jan. 2003.
[11] M. J. Fowler, “Microvascular and macrovascular complications of diabetes,” Clinical diabetes, 2008.
[12] B. B. L. Groen, H. M. Hamer, T. Snijders, J. van Kranenburg, D. Frijns, H. Vink, and L. J. C. van Loon, “Skeletal muscle capillary density and microvascular function are
92
compromised with aging and type 2 diabetes,” Journal of Applied Physiology, vol. 116, no. 8, pp. 998–1005, Apr. 2014.
[13] V. R. Taqueti, R. Hachamovitch, V. L. Murthy, M. Naya, C. R. Foster, J. Hainer, S. Dorbala, R. Blankstein, and M. F. Di Carli, “Global coronary flow reserve is associated with adverse cardiovascular events independently of luminal angiographic severity and modifies the effect of early revascularization.,” Circulation, vol. 131, no. 1, pp. 19–27, Jan. 2015.
[14] V. R. Taqueti, B. M. Everett, V. L. Murthy, M. Gaber, C. R. Foster, J. Hainer, R. Blankstein, S. Dorbala, and M. F. Di Carli, “Interaction of impaired coronary flow reserve and cardiomyocyte injury on adverse cardiovascular outcomes in patients without overt coronary artery disease.,” Circulation, vol. 131, no. 6, pp. 528–535, Feb. 2015.
[15] J. G. Betts, P. Desaix, J. E. Johnson, O. Korol, D. Kruse, B. Poe, J. Wise, M. D. Womble, K. A. Young, O. College, Rice University, Anatomy and Physiology. 2013.
[16] B. E. Carlson and T. W. Secomb, “A theoretical model for the myogenic response based on the length-tension characteristics of vascular smooth muscle.,” Microcirculation, vol. 12, no. 4, pp. 327–338, Jun. 2005.
[17] P. S. Clifford, “Local control of blood flow.,” Adv Physiol Educ, vol. 35, no. 1, pp. 5–15, Mar. 2011.
[18] Y. Hellsten, M. Nyberg, L. G. Jensen, and S. P. Mortensen, “Vasodilator interactions in skeletal muscle blood flow regulation,” J Physiology, vol. 590, no. 24, pp. 6297–6305, Dec. 2012.
[19] M. L. Ellsworth, C. G. Ellis, D. Goldman, A. H. Stephenson, H. H. Dietrich, and R. S. Sprague, “Erythrocytes: Oxygen Sensors and Modulators of Vascular Tone,” Physiology, vol. 24, no. 2, pp. 107–116, Apr. 2009.
[20] Y. Hellsten, M. Nyberg, and S. P. Mortensen, “Contribution of intravascular versus interstitial purines and nitric oxide in the regulation of exercise hyperaemia in humans.,” J Physiology, vol. 590, no. 20, pp. 5015–5023, Oct. 2012.
[21] T. J. Anderson, F. Charbonneau, L. M. Title, J. Buithieu, M. S. Rose, H. Conradson, K. Hildebrand, M. Fung, S. Verma, and E. M. Lonn, “Microvascular Function Predicts Cardiovascular Events in Primary Prevention: Long-Term Results From the Firefighters and Their Endothelium (FATE) Study,” Circulation, vol. 123, no. 2, pp. 163–169, Jan. 2011.
[22] D. H. J. Thijssen, M. A. Black, K. E. Pyke, J. Padilla, G. Atkinson, R. A. Harris, B. Parker, M. E. Widlansky, M. E. Tschakovsky, and D. J. Green, “Assessment of flow-mediated dilation in humans: a methodological and physiological guideline,” AJP: Heart and Circulatory Physiology, vol. 300, no. 1, pp. H2–H12, Jan. 2011.
[23] D. H. J. Thijssen, M. W. P. Bleeker, P. Smits, and M. T. E. Hopman, “Reproducibility of blood flow and post-occlusive reactive hyperaemia as measured by venous occlusion plethysmography.,” Clin. Sci., vol. 108, no. 2, pp. 151–157, Feb. 2005.
[24] A. Jeremias, S. D. Filardo, R. J. Whitbourn, R. S. Kernoff, A. C. Yeung, P. J. Fitzgerald, and P. G. Yock, “Effects of Intravenous and Intracoronary Adenosine 5'-Triphosphate as Compared With Adenosine on Coronary Flow and Pressure Dynamics,” Circulation, vol. 101, no. 3, pp. 318–323, Jan. 2000.
93
[25] T. C. Wu, J. W. Chen, C. I. Chen, G. Y. Mar, N. W. Hsu, Y. H. Chen, Y. A. Ding, S. P. Wang, and M. S. Chang, “Early alteration of coronary hemodynamics in late restenosis after coronary angioplasty.,” Jpn Heart J, vol. 40, no. 5, pp. 535–548, Sep. 1999.
[26] H. P. Ledermann, H.-G. Heidecker, A.-C. Schulte, C. Thalhammer, M. Aschwanden, K. A. Jaeger, K. Scheffler, and D. Bilecen, “Calf muscles imaged at BOLD MR: correlation with TcPO2 and flowmetry measurements during ischemia and reactive hyperemia--initial experience.,” Radiology, vol. 241, no. 2, pp. 477–484, Nov. 2006.
[27] J. T. Groothuis, L. van Vliet, M. Kooijman, and M. T. E. Hopman, “Venous cuff pressures from 30 mmHg to diastolic pressure are recommended to measure arterial inflow by plethysmography.,” Journal of Applied Physiology, vol. 95, no. 1, pp. 342–347, Jul. 2003.
[28] E. Cosson, F. Paycha, P. Tellier, R. N. Sachs, A. Ramadan, J. Paries, J. R. Attali, and P. Valensi, “Lower-limb vascularization in diabetic patients. Assessment by thallium-201 scanning coupled with exercise myocardial scintigraphy.,” Diabetes Care, vol. 24, no. 5, pp. 870–874, May 2001.
[29] C.-C. Lin, H.-J. Ding, Y.-W. Chen, W.-T. Huang, and A. Kao, “Usefulness of thallium-201 muscle perfusion scan to investigate perfusion reserve in the lower limbs of Type 2 diabetic patients,” J. Diabetes Complicat., vol. 18, no. 4, pp. 233–236, Jul. 2004.
[30] J. R. Lindner, L. Womack, E. J. Barrett, J. Weltman, W. Price, N. L. Harthun, S. Kaul, and J. T. Patrie, “Limb stress-rest perfusion imaging with contrast ultrasound for the assessment of peripheral arterial disease severity.,” JACC Cardiovasc Imaging, vol. 1, no. 3, pp. 343–350, May 2008.
[31] T. Bragadeesh, I. Sari, M. Pascotto, A. Micari, S. Kaul, and J. R. Lindner, “Detection of peripheral vascular stenosis by assessing skeletal muscle flow reserve.,” JAC, vol. 45, no. 5, pp. 780–785, Mar. 2005.
[32] J. A. Detre, J. S. Leigh, D. S. Williams, and A. P. Koretsky, “Perfusion imaging.,” Magn. Reson. Med., vol. 23, no. 1, pp. 37–45, Jan. 1992.
[33] E. C. Wong, R. B. Buxton, and L. R. Frank, “Implementation of quantitative perfusion imaging techniques for functional brain mapping using pulsed arterial spin labeling.,” NMR Biomed., vol. 10, no. 4, pp. 237–249, Jun. 1997.
[34] L. R. Frank, E. C. Wong, L. J. Haseler, and R. B. Buxton, “Dynamic imaging of perfusion in human skeletal muscle during exercise with arterial spin labeling.,” Magn. Reson. Med., vol. 42, no. 2, pp. 258–267, Aug. 1999.
[35] W.-C. Wu, E. Mohler, S. J. Ratcliffe, F. W. Wehrli, J. A. Detre, and T. F. Floyd, “Skeletal muscle microvascular flow in progressive peripheral artery disease: assessment with continuous arterial spin-labeling perfusion magnetic resonance imaging.,” Journal of the American College of Cardiology, vol. 53, no. 25, pp. 2372–2377, Jun. 2009.
[36] D. Lopez, A. W. Pollak, C. H. Meyer, F. H. Epstein, L. Zhao, A. J. Pesch, R. Jiji, J. R. Kay, J. M. Dimaria, J. M. Christopher, and C. M. Kramer, “Arterial spin labeling perfusion cardiovascular magnetic resonance of the calf in peripheral arterial disease: cuff occlusion hyperemia vs exercise.,” J Cardiovasc Magn Reson, vol. 17, p. 23, 2015.
[37] E. K. Englund, M. C. Langham, S. J. Ratcliffe, M. J. Fanning, F. W. Wehrli, E. R. Mohler, and T. F. Floyd, “Multiparametric assessment of vascular function in peripheral
94
artery disease: dynamic measurement of skeletal muscle perfusion, blood-oxygen-level dependent signal, and venous oxygen saturation.,” Circulation: Cardiovascular Imaging, vol. 8, no. 4, Apr. 2015.
[38] G. Grözinger, R. Pohmann, F. Schick, U. Grosse, R. Syha, K. Brechtel, K. Rittig, and P. Martirosian, “Perfusion measurements of the calf in patients with peripheral arterial occlusive disease before and after percutaneous transluminal angioplasty using MR arterial spin labeling.,” J. Magn. Reson. Imaging, vol. 40, no. 4, pp. 980–987, Oct. 2014.
[39] M. Raitakari, M. J. Knuuti, U. Ruotsalainen, H. Laine, P. Mäkeä, M. Teräs, H. Sipilä, T. Niskanen, O. T. Raitakari, H. Iida, R. Härkönen, U. Wegelius, H. Yki-Järvinen, and P. Nuutila, “Insulin increases blood volume in human skeletal muscle: studies using [15O]CO and positron emission tomography.,” Am. J. Physiol., vol. 269, no. 6, pp. E1000–5, Dec. 1995.
[40] D. C. Isbell, F. H. Epstein, X. Zhong, J. M. Dimaria, S. S. Berr, C. H. Meyer, W. J. Rogers, N. L. Harthun, K. D. Hagspiel, A. Weltman, and C. M. Kramer, “Calf muscle perfusion at peak exercise in peripheral arterial disease: Measurement by first-pass contrast-enhanced magnetic resonance imaging,” J. Magn. Reson. Imaging, vol. 25, no. 5, pp. 1013–1020, May 2007.
[41] R. B. Thompson, R. J. Aviles, A. Z. Faranesh, V. K. Raman, V. Wright, R. S. Balaban, E. R. Mcveigh, and R. J. Lederman, “Measurement of skeletal muscle perfusion during postischemic reactive hyperemia using contrast-enhanced MRI with a step-input function.,” Magn. Reson. Med., vol. 54, no. 2, pp. 289–298, Aug. 2005.
[42] S. Ogawa, T. M. Lee, A. S. Nayak, and P. Glynn, “Oxygenation-sensitive contrast in magnetic resonance image of rodent brain at high magnetic fields.,” Magn. Reson. Med., vol. 14, no. 1, pp. 68–78, Apr. 1990.
[43] M. Vöhringer, J. A. Flewitt, J. D. Green, R. Dharmakumar, J. Wang, J. V. Tyberg, and M. G. Friedrich, “Oxygenation-sensitive CMR for assessing vasodilator-induced changes of myocardial oxygenation.,” J Cardiovasc Magn Reson, vol. 12, p. 20, 2010.
[44] B. Jacobi, G. Bongartz, S. Partovi, A.-C. Schulte, M. Aschwanden, A. B. Lumsden, M. G. Davies, M. Loebe, G. P. Noon, S. Karimi, J. K. Lyo, D. Staub, R. W. Huegli, and D. Bilecen, “Skeletal muscle BOLD MRI: From underlying physiological concepts to its usefulness in clinical conditions,” J. Magn. Reson. Imaging, vol. 35, no. 6, pp. 1253–1265, May 2012.
[45] H. P. Ledermann, A.-C. Schulte, H.-G. Heidecker, M. Aschwanden, K. A. Jäger, K. Scheffler, W. Steinbrich, and D. Bilecen, “Blood oxygenation level-dependent magnetic resonance imaging of the skeletal muscle in patients with peripheral arterial occlusive disease.,” Circulation, vol. 113, no. 25, pp. 2929–2935, Jun. 2006.
[46] A. Bajwa, R. Wesolowski, A. Patel, P. Saha, F. Ludwinski, M. Ikram, M. Albayati, A. Smith, E. Nagel, and B. Modarai, “Blood Oxygenation Level-Dependent CMR-Derived Measures in Critical Limb Ischemia and Changes With Revascularization.,” Journal of the American College of Cardiology, vol. 67, no. 4, pp. 420–431, Feb. 2016.
[47] 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.,” Magn. Reson. Med., vol. 40, no. 3, pp. 383–396, Sep. 1998.
95
[48] W. Dai, D. Garcia, C. De Bazelaire, and D. C. Alsop, “Continuous flow-driven inversion for arterial spin labeling using pulsed radio frequency and gradient fields.,” Magn. Reson. Med., vol. 60, no. 6, pp. 1488–1497, Dec. 2008.
[49] E. K. Englund, Z. B. Rodgers, M. C. Langham, E. R. Mohler, T. F. Floyd, and F. W. Wehrli, “Measurement of skeletal muscle perfusion dynamics with pseudo-continuous arterial spin labeling (pCASL): Assessment of relative labeling efficiency at rest and during hyperemia, and comparison to pulsed arterial spin labeling (PASL).,” J. Magn. Reson. Imaging, Apr. 2016.
[50] 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.,” Magn. Reson. Med., vol. 62, no. 4, pp. 975–983, Oct. 2009.
[51] W. M. Luh, E. C. Wong, P. A. Bandettini, and J. S. Hyde, “QUIPSS II with thin-slice TI1 periodic saturation: a method for improving accuracy of quantitative perfusion imaging using pulsed arterial spin labeling.,” Magn. Reson. Med., vol. 41, no. 6, pp. 1246–1254, Jun. 1999.
[52] H. Lu, M. J. Donahue, and P. C. M. van Zijl, “Detrimental effects of BOLD signal in arterial spin labeling fMRI at high field strength,” Magn. Reson. Med., vol. 56, no. 3, pp. 546–552, Sep. 2006.
[53] D. M. Garcia, G. Duhamel, and D. C. Alsop, “Efficiency of inversion pulses for background suppressed arterial spin labeling.,” Magn. Reson. Med., vol. 54, no. 2, pp. 366–372, Aug. 2005.
[54] W.-C. Wu, J. Wang, J. A. Detre, S. J. Ratcliffe, and T. F. Floyd, “Transit delay and flow quantification in muscle with continuous arterial spin labeling perfusion-MRI.,” J. Magn. Reson. Imaging, vol. 28, no. 2, pp. 445–452, Aug. 2008.
[55] 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.,” Magn. Reson. Med., vol. 40, no. 3, pp. 348–355, Sep. 1998.
[56] D. S. Williams, J. A. Detre, J. S. Leigh, and A. P. Koretsky, “Magnetic resonance imaging of perfusion using spin inversion of arterial water.,” Proc. Natl. Acad. Sci. U.S.A., vol. 89, no. 1, pp. 212–216, Jan. 1992.
[57] M. Hirai and W. Schoop, “Clinical significance of Doppler velocity and blood pressure measurements in peripheral arterial occlusive disease.,” Angiology, vol. 35, no. 1, pp. 45–53, Jan. 1984.
[58] J. S. Raynaud, S. Duteil, J. T. Vaughan, F. Hennel, C. Wary, A. Leroy-Willig, and P. G. Carlier, “Determination of skeletal muscle perfusion using arterial spin labeling NMRI: validation by comparison with venous occlusion plethysmography.,” Magn. Reson. Med., vol. 46, no. 2, pp. 305–311, Aug. 2001.
[59] A. Toth, M. Pal, M. Intaglietta, and P. C. Johnson, “Contribution of anaerobic metabolism to reactive hyperemia in skeletal muscle,” AJP: Heart and Circulatory Physiology, vol. 292, no. 6, pp. H2643–H2653, Mar. 2007.
[60] I. Carlsson, A. Sollevi, and A. Wennmalm, “The role of myogenic relaxation, adenosine and prostaglandins in human forearm reactive hyperaemia.,” J Physiology, vol. 389, pp. 147–161, Aug. 1987.
96
[61] E. K. Englund, M. C. Langham, C. Li, Z. B. Rodgers, T. F. Floyd, E. R. Mohler, and F. W. Wehrli, “Combined measurement of perfusion, venous oxygen saturation, and skeletal muscle T2* during reactive hyperemia in the leg.,” J Cardiovasc Magn Reson, vol. 15, p. 70, 2013.
[62] A. Fronek, K. Johansen, R. B. Dilley, and E. F. Bernstein, “Ultrasonographically monitored postocclusive reactive hyperemia in the diagnosis of peripheral arterial occlusive disease.,” Circulation, vol. 48, no. 1, pp. 149–152, Jul. 1973.
[63] M. Ursino, P. Di Giammarco, and E. Belardinelli, “A mathematical model of cerebral blood flow chemical regulation--Part I: Diffusion processes.,” IEEE Trans. Biomed. Eng., vol. 36, no. 2, pp. 183–191, Feb. 1989.
[64] M. Ursino, P. Di Giammarco, and E. Belardinelli, “A mathematical model of cerebral blood flow chemical regulation--Part II: Reactivity of cerebral vascular bed.,” IEEE Trans. Biomed. Eng., vol. 36, no. 2, pp. 192–201, Feb. 1989.
[65] M. Ursino, “A mathematical model of overall cerebral blood flow regulation in the rat.,” IEEE Trans. Biomed. Eng., vol. 38, no. 8, pp. 795–807, Aug. 1991.
[66] B. Spronck, E. G. H. J. Martens, E. D. Gommer, and F. N. van de Vosse, “A lumped parameter model of cerebral blood flow control combining cerebral autoregulation and neurovascular coupling,” AJP: Heart and Circulatory Physiology, vol. 303, no. 9, pp. H1143–H1153, Nov. 2012.
[67] J. C. Arciero, B. E. Carlson, and T. W. Secomb, “Theoretical model of metabolic blood flow regulation: roles of ATP release by red blood cells and conducted responses,” AJP: Heart and Circulatory Physiology, vol. 295, no. 4, pp. H1562–H1571, Jul. 2008.
[68] F. F. M. deMul, F. Morales, A. J. Smit, and R. Graaff, “A Model for Post-Occlusive Reactive Hyperemia as Measured With Laser-Doppler Perfusion Monitoring,” IEEE Trans. Biomed. Eng., vol. 52, no. 2, pp. 184–190, Feb. 2005.
[69] S. Kurbel, M. Gros, and S. Maric, “Complexity of human circulation design: tips for students,” Adv Physiol Educ, vol. 33, no. 2, pp. 130–131, Jun. 2009.
[70] B. E. Carlson, J. C. Arciero, and T. W. Secomb, “Theoretical model of blood flow autoregulation: roles of myogenic, shear-dependent, and metabolic responses,” AJP: Heart and Circulatory Physiology, vol. 295, no. 4, pp. H1572–H1579, Jul. 2008.
[71] Y. Hellsten, D. Maclean, G. Rådegran, B. Saltin, and J. Bangsbo, “Adenosine concentrations in the interstitium of resting and contracting human skeletal muscle.,” Circulation, vol. 98, no. 1, pp. 6–8, Jul. 1998.
[72] N. Lai, H. Zhou, G. M. Saidel, M. Wolf, K. McCully, L. B. Gladden, and M. E. Cabrera, “Modeling oxygenation in venous blood and skeletal muscle in response to exercise using near-infrared spectroscopy,” Journal of Applied Physiology, vol. 106, no. 6, pp. 1858–1874, May 2009.
[73] Y. Hellsten, “The effect of muscle contraction on the regulation of adenosine formation in rat skeletal muscle cells.,” J Physiology, vol. 518, pp. 761–768, Aug. 1999.
[74] J. Lynge, C. Juel, and Y. Hellsten, “Extracellular formation and uptake of adenosine during skeletal muscle contraction in the rat: role of adenosine transporters.,” J Physiology, vol. 537, no. 2, pp. 597–605, Dec. 2001.
97
[75] F. Costa, J. Heusinkveld, R. Ballog, S. Davis, and I. Biaggioni, “Estimation of skeletal muscle interstitial adenosine during forearm dynamic exercise in humans.,” Hypertension, vol. 35, no. 5, pp. 1124–1128, May 2000.
[76] D. A. MacLean, L. I. Sinoway, and U. Leuenberger, “Systemic hypoxia elevates skeletal muscle interstitial adenosine levels in humans.,” Circulation, vol. 98, no. 19, pp. 1990–1992, Nov. 1998.
[77] T. Stumpe and J. Schrader, “Phosphorylation potential, adenosine formation, and critical PO2 in stimulated rat cardiomyocytes.,” Am. J. Physiol., vol. 273, no. 2, pp. H756–66, Aug. 1997.
[78] Z. Y. Wang, E. A. Noyszewski, and J. S. Leigh, “In vivo MRS measurement of deoxymyoglobin in human forearms.,” Magn. Reson. Med., vol. 14, no. 3, pp. 562–567, Jun. 1990.
[79] P. G. Carlier, D. Bertoldi, C. Baligand, C. Wary, and Y. Fromes, “Muscle blood flow and oxygenation measured by NMR imaging and spectroscopy.,” NMR Biomed., vol. 19, no. 7, pp. 954–967, Nov. 2006.
[80] S. P. Mortensen, P. Thaning, M. Nyberg, B. Saltin, and Y. Hellsten, “Local release of ATP into the arterial inflow and venous drainage of human skeletal muscle: insight from ATP determination with the intravascular microdialysis technique.,” J Physiology, vol. 589, no. 7, pp. 1847–1857, Apr. 2011.
[81] D. M. Collins, W. T. McCullough, and M. L. Ellsworth, “Conducted vascular responses: communication across the capillary bed.,” Microvasc. Res., vol. 56, no. 1, pp. 43–53, Jul. 1998.
[82] P. Thaning, L. T. Bune, Y. Hellsten, H. Pilegaard, B. Saltin, and J. B. Rosenmeier, “Attenuated Purinergic Receptor Function in Patients With Type 2 Diabetes,” Diabetes, vol. 59, no. 1, pp. 182–189, Dec. 2009.
[83] H. J. Ballard, “ATP and adenosine in the regulation of skeletal muscle blood flow during exercise.,” Sheng Li Xue Bao, vol. 66, no. 1, pp. 67–78, Feb. 2014.
[84] B. P. Ramakers, P. Pickkers, A. Deussen, G. A. Rongen, P. van den Broek, J. G. van der Hoeven, P. Smits, and N. P. Riksen, “Measurement of the endogenous adenosine concentration in humans in vivo: methodological considerations.,” Curr. Drug Metab., vol. 9, no. 8, pp. 679–685, Oct. 2008.
[85] K. Schewzow, M. Andreas, E. Moser, M. Wolzt, and A. I. Schmid, “Automatic model-based analysis of skeletal muscle BOLD-MRI in reactive hyperemia,” J. Magn. Reson. Imaging, vol. 38, no. 4, pp. 963–969, Nov. 2012.
[86] N. Olamaei, J. Dupuis, Q. Ngo, V. Finnerty, T.-T. Vo Thang, S. Authier, P. Khairy, and F. Harel, “Characterization and reproducibility of forearm arterial flow during reactive hyperemia,” Physiol. Meas., vol. 31, no. 6, pp. 763–773, Apr. 2010.
[87] K. L. Wright, N. Seiberlich, J. A. Jesberger, D. A. Nakamoto, R. F. Muzic, M. A. Griswold, and V. Gulani, “Simultaneous magnetic resonance angiography and perfusion (MRAP) measurement: initial application in lower extremity skeletal muscle.,” J. Magn. Reson. Imaging, vol. 38, no. 5, pp. 1237–1244, Nov. 2013.
98
[88] W. R. Hiatt, E. J. Armstrong, C. J. Larson, and E. P. Brass, “Pathogenesis of the limb manifestations and exercise limitations in peripheral artery disease.,” Circulation Research, vol. 116, no. 9, pp. 1527–1539, Apr. 2015.
[89] N. M. Hamburg and M. A. Creager, “Pathophysiology of Intermittent Claudication in Peripheral Artery Disease.,” Circ. J., vol. 81, no. 3, pp. 281–289, Feb. 2017.
[90] S. Verma, “Endothelial Function Testing as a Biomarker of Vascular Disease,” Circulation, vol. 108, no. 17, pp. 2054–2059, Oct. 2003.
[91] G. Brevetti, “Endothelial Dysfunction and Cardiovascular Risk Prediction in Peripheral Arterial Disease: Additive Value of Flow-Mediated Dilation to Ankle-Brachial Pressure Index,” Circulation, vol. 108, no. 17, pp. 2093–2098, Oct. 2003.
[92] A. L. Huang, A. E. Silver, E. Shvenke, D. W. Schopfer, E. Jahangir, M. A. Titas, A. Shpilman, J. O. Menzoian, M. T. Watkins, J. D. Raffetto, G. Gibbons, J. Woodson, P. M. Shaw, M. Dhadly, R. T. Eberhardt, J. F. Keaney, N. Gokce, and J. A. Vita, “Predictive value of reactive hyperemia for cardiovascular events in patients with peripheral arterial disease undergoing vascular surgery.,” Arteriosclerosis, Thrombosis, and Vascular Biology, vol. 27, no. 10, pp. 2113–2119, Oct. 2007.
[93] T. J. Anderson, A. Uehata, M. D. Gerhard, I. T. Meredith, S. Knab, D. Delagrange, E. H. Lieberman, P. Ganz, M. A. Creager, A. C. Yeung, and A. P. Selwyn, “Close Relation of Endothelial Function in the Human Coronary and Peripheral Circulations,” JAC, vol. 26, no. 5, pp. 1235–1241, Nov. 1995.
[94] F. Crea, P. G. Camici, and C. N. Bairey Merz, “Coronary microvascular dysfunction: an update,” European Heart Journal, vol. 35, no. 17, pp. 1101–1111, May 2014.
[95] D. G. Della Rocca and C. J. Pepine, “Some thoughts on the continuing dilemma of angina pectoris,” European Heart Journal, vol. 35, no. 21, pp. 1361–1364, Jun. 2014.
[96] L. Jespersen, A. Hvelplund, S. Z. Abildstrom, F. Pedersen, S. Galatius, J. K. Madsen, E. Jorgensen, H. Kelbaek, and E. Prescott, “Stable angina pectoris with no obstructive coronary artery disease is associated with increased risks of major adverse cardiovascular events,” European Heart Journal, vol. 33, no. 6, pp. 734–744, Mar. 2012.
[97] V. N. Varu, M. E. Hogg, and M. R. Kibbe, “Critical limb ischemia.,” J. Vasc. Surg., vol. 51, no. 1, pp. 230–241, Jan. 2010.
[98] J. C. Frisbee, A. G. Goodwill, S. J. Frisbee, J. T. Butcher, F. Wu, and P. D. Chantler, “Microvascular perfusion heterogeneity contributes to peripheral vascular disease in metabolic syndrome.,” J Physiology, vol. 594, no. 8, pp. 2233–2243, Apr. 2016.
[99] Y. Liu, B. P. Davidson, Q. Yue, T. Belcik, A. Xie, Y. Inaba, O. J. T. McCarty, G. W. Tormoen, Y. Zhao, Z. M. Ruggeri, B. A. Kaufmann, and J. R. Lindner, “Molecular imaging of inflammation and platelet adhesion in advanced atherosclerosis effects of antioxidant therapy with NADPH oxidase inhibition.,” Circulation: Cardiovascular Imaging, vol. 6, no. 1, pp. 74–82, Jan. 2013.
[100] T. Jämsén, H. Manninen, H. Tulla, and P. Matsi, “The final outcome of primary infrainguinal percutaneous transluminal angioplasty in 100 consecutive patients with chronic critical limb ischemia.,” J Vasc Interv Radiol, vol. 13, no. 5, pp. 455–463, May 2002.
99
[101] G. I. Taylor and J. H. Palmer, “The vascular territories (angiosomes) of the body: experimental study and clinical applications.,” Br J Plast Surg, vol. 40, no. 2, pp. 113–141, Mar. 1987.
[102] X. Golay, E. T. Petersen, and F. Hui, “Pulsed star labeling of arterial regions (PULSAR): A robust regional perfusion technique for high field imaging,” Magn. Reson. Med., vol. 53, no. 1, pp. 15–21, 2004.
[103] E. C. Wong, “Vessel-encoded arterial spin-labeling using pseudocontinuous tagging.,” Magn. Reson. Med., vol. 58, no. 6, pp. 1086–1091, Dec. 2007.
[104] G. Goelman and J. S. Leigh, “Multiband adiabatic inversion pulses,” Journal of Magnetic Resonance, 1993.
[105] G. Goelman, “Two methods for peak RF power minimization of multiple inversion-band pulses.,” Magn. Reson. Med., vol. 37, no. 5, pp. 658–665, May 1997.
[106] X. Li, D. Wang, E. J. Auerbach, S. Moeller, K. Ugurbil, and G. J. Metzger, “Theoretical and experimental evaluation of multi-band EPI for high-resolution whole brain pCASL Imaging,” Neuroimage, vol. 106, no. C, pp. 170–181, Feb. 2015.
[107] M. Barth, F. Breuer, P. J. Koopmans, D. G. Norris, and B. A. Poser, “Simultaneous multislice (SMS) imaging techniques,” Magn. Reson. Med., vol. 75, no. 1, pp. 63–81, Aug. 2015.
[108] D. A. Feinberg, A. Beckett, and L. Chen, “Arterial spin labeling with simultaneous multi-slice echo planar imaging,” Magn. Reson. Med., vol. 70, no. 6, pp. 1500–1506, Oct. 2013.
[109] W. Aarnoudse, “Epicardial Stenosis Severity Does Not Affect Minimal Microcirculatory Resistance,” Circulation, vol. 110, no. 15, pp. 2137–2142, Oct. 2004.
[110] M. K. C. Ng, “Invasive Assessment of the Coronary Microcirculation: Superior Reproducibility and Less Hemodynamic Dependence of Index of Microcirculatory Resistance Compared With Coronary Flow Reserve,” Circulation, vol. 113, no. 17, pp. 2054–2061, May 2006.
[111] W. F. Fearon, A. F. Low, A. S. Yong, R. McGeoch, C. Berry, M. G. Shah, M. Y. Ho, H. S. Kim, J. P. Loh, and K. G. Oldroyd, “Prognostic Value of the Index of Microcirculatory Resistance Measured After Primary Percutaneous Coronary Intervention,” Circulation, vol. 127, no. 24, pp. 2436–2441, Jun. 2013.
[112] M. A. Creager, T. F. Lüscher, F. Cosentino, and J. A. Beckman, “Diabetes and vascular disease: pathophysiology, clinical consequences, and medical therapy: Part I.,” Circulation, vol. 108, no. 12, pp. 1527–1532, Sep. 2003.
[113] J. A. Beckman, M. A. Creager, and P. Libby, “Diabetes and atherosclerosis,” JAMA, vol. 287, no. 19, pp. 2570–2581, 2002.
[114] T. Thiruvoipati, C. E. Kielhorn, and E. J. Armstrong, “Peripheral artery disease in patients with diabetes: Epidemiology, mechanisms, and outcomes.,” World J Diabetes, vol. 6, no. 7, pp. 961–969, Jul. 2015.
[115] M. I. Spreen, H. Gremmels, M. Teraa, R. W. Sprengers, M. C. Verhaar, R. G. Statius van Eps, J.-P. P. M. de Vries, W. P. T. M. Mali, and H. van Overhagen, “Diabetes Is Associated With Decreased Limb Survival in Patients With Critical Limb Ischemia:
100
Pooled Data From Two Randomized Controlled Trials,” Diabetes Care, vol. 39, no. 11, pp. 2058–2064, Oct. 2016.
[116] F. Dick, N. Diehm, A. Galimanis, M. Husmann, J. Schmidli, and I. Baumgartner, “Surgical or endovascular revascularization in patients with critical limb ischemia: Influence of diabetes mellitus on clinical outcome,” Journal of Vascular Surgery, vol. 45, no. 4, pp. 751–761, Apr. 2007.
[117] F. Paneni, J. A. Beckman, M. A. Creager, and F. Cosentino, “Diabetes and vascular disease: pathophysiology, clinical consequences, and medical therapy: part I,” European Heart Journal, vol. 34, no. 31, pp. 2436–2443, Aug. 2013.
[118] S. Singh, E. J. Armstrong, W. Sherif, B. Alvandi, G. G. Westin, G. D. Singh, E. A. Amsterdam, and J. R. Laird, “Association of elevated fasting glucose with lower patency and increased major adverse limb events among patients with diabetes undergoing infrapopliteal balloon angioplasty,” Vasc Med, vol. 19, no. 4, pp. 307–314, Jul. 2014.
[119] ADVANCE Collaborative Group, A. Patel, S. MacMahon, J. Chalmers, B. Neal, L. Billot, M. Woodward, M. Marre, M. Cooper, P. Glasziou, D. Grobbee, P. Hamet, S. Harrap, S. Heller, L. Liu, G. Mancia, C. E. Mogensen, C. Pan, N. Poulter, A. Rodgers, B. Williams, S. Bompoint, B. E. de Galan, R. Joshi, and F. Travert, “Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes.,” N. Engl. J. Med., vol. 358, no. 24, pp. 2560–2572, Jun. 2008.
[120] W. Duckworth, C. Abraira, T. Moritz, D. Reda, N. Emanuele, P. D. Reaven, F. J. Zieve, J. Marks, S. N. Davis, R. Hayward, S. R. Warren, S. Goldman, M. McCarren, M. E. Vitek, W. G. Henderson, G. D. Huang, VADT Investigators, “Glucose control and vascular complications in veterans with type 2 diabetes.,” N. Engl. J. Med., vol. 360, no. 2, pp. 129–139, Jan. 2009.
[121] E. C. Perin, M. P. Murphy, K. L. March, R. Bolli, J. Loughran, P. C. Yang, N. J. Leeper, R. L. Dalman, J. Alexander, T. D. Henry, J. H. Traverse, C. J. Pepine, R. D. Anderson, S. Berceli, J. T. Willerson, R. Muthupillai, A. Gahremanpour, G. Raveendran, O. Velasquez, J. M. Hare, I. Hernandez Schulman, V. S. Kasi, W. R. Hiatt, B. Ambale-Venkatesh, J. A. Lima, D. A. Taylor, M. Resende, A. P. Gee, A. G. Durett, J. Bloom, S. Richman, P. G’Sell, S. Williams, F. Khan, E. Gyang Ross, M. R. Santoso, J. Goldman, D. Leach, E. Handberg, B. Cheong, N. Piece, D. DiFede, B. Bruhn-Ding, E. Caldwell, J. Bettencourt, D. Lai, L. Piller, L. Simpson, M. Cohen, S. L. Sayre, R. W. Vojvodic, L. Moyé, R. F. Ebert, R. D. Simari, and A. T. Hirsch, “Evaluation of Cell Therapy on Exercise Performance and Limb Perfusion in Peripheral Artery DiseaseClinical Perspective,” Circulation, vol. 135, no. 15, pp. 1417–1428, Apr. 2017.
[122] M. Teraa, R. W. Sprengers, R. E. G. Schutgens, I. C. M. Slaper-Cortenbach, Y. van der Graaf, A. Algra, I. van der Tweel, P. A. Doevendans, W. P. T. M. Mali, F. L. Moll, and M. C. Verhaar, “Effect of repetitive intra-arterial infusion of bone marrow mononuclear cells in patients with no-option limb ischemia: the randomized, double-blind, placebo-controlled Rejuvenating Endothelial Progenitor Cells via Transcutaneous Intra-arterial Supplementation (JUVENTAS) trial.,” Circulation, vol. 131, no. 10, pp. 851–860, Mar. 2015.
101
[123] J. P. Cooke and D. W. Losordo, “Modulating the vascular response to limb ischemia: angiogenic and cell therapies.,” Circulation Research, vol. 116, no. 9, pp. 1561–1578, Apr. 2015.