University of Southern California Dissertations and Theses

Quantitative MRI for the measurement of cerebral oxygen extraction fraction in sickle cell disease

(USC Thesis Other)

Quantitative MRI for the measurement of cerebral oxygen extraction fraction in sickle cell disease

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Copyright 2024 Jian Shen
Quantitative MRI for the Measurement of Cerebral Oxygen Extraction Fraction in Sickle Cell
Disease
by
Jian Shen
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
Doctor of Philosophy
(BIOMEDICAL ENGINEERING)
May 2024

ii
Dedication
To my mom, dad, and Wenxi. Thank you for your endless love and support.

iii
Acknowledgments
This thesis not only signifies the culmination of my graduate studies but also embodies the
profound love and support I have received. Without this invaluable support, reaching this
milestone today would not have been possible. Though it takes much longer and harder than I
expected, I feel blessed that I have survived through all these.
To my advisor, Dr. John Wood: I extend my heartfelt gratitude to you for your unwavering
support, invaluable guidance, and constant encouragement throughout my doctoral journey. Your
expertise, commitment, and constructive feedback have been instrumental in shaping the
trajectory of my research and academic growth. I can never forget your passion for science and
creativity in engineering. I can never forget the day when you brought me back when I made a
mistake. I am truly fortunate to have had such a dedicated mentor. All these make me a better
man.
Thank you to my qualification and defense committee members: Dr. Michael Khoo, Dr.
Danny JJ Wang, Dr. Natasha Lepore, and Dr. Vasilis Marmarelis. Thank you for your diverse
perspective, critical advice, and generous support. Thank you for all the time and patience
throughout the preparation of my thesis.
Thank you to my core research family: Dr. Adam Bush, Dr. Eamon Doyle, Dr. Xin Miao, Dr.
Chau Vu, Dr. Soyoung Choi, Dr. Yaqiong Chai, Botian Xu, Clio Gonzalez Zacarias, Sudarshan
Ranganathan, Emma Carpenter, Sneha Verma, and Peter Chiarelli. Thank you for all the laughter
and tears we shared in the lab. I appreciate the countless insightful discussions, constructive
feedback, and shared triumphs that have characterized our time together. Thank you for being an
integral part of this transformative experience; your contributions have left an indelible mark on
both my academic and personal growth.
Thank you to my extended research family: Dr. Krishna Nayak, Dr. Justin Haldar, Dr.
Matthew Borzage, Dr. Jon Detterich, Dr. Thomas Coates, Dr. Brent Liu, Dr. Sharon O’Neil, Silvie
Suriany, Honglei Liu, Nathan Smith, Noel Arugay, Dr. Yi Guo, Dr. Wayne Chen, and William Yang. I
would also like to thank our collaborators at Amsterdam UMC: Dr. Aart Nederveen, Dr. Bart
Biemond, Dr. Koen Baas, and Lisa Afzali.
Special thanks to the family of Grace: Grace Ting, David Ting, James Ting, and Katherine
Ting. Thank you for making me feel at home when I first came to Los Angeles.
Most importantly, I would like to dedicate this thesis to my parents: Weiqian Shen and
Fang Liu. Your enduring encouragement, understanding, and patience have been instrumental in
navigating the challenges of pursuing a PhD. The sacrifices you have made and the consistent
belief you've shown in my abilities have been my constant motivation. Your love and support have
created a nurturing environment, providing me with the stability and strength needed to

iv
undertake this significant endeavor. This achievement is as much yours as it is mine, and for that,
I am profoundly grateful.
Thank you to my wife and my love, Wenxi Yu. It is my greatest luck meeting you through this
journey. Thank you for your patience and encouragement, guiding me through the darkness and
leading me to success. I can’t wait to hold your hands and start our new lives together.

v
Table of Contents
Dedication........................................................................................................................................ii
Acknowledgments...........................................................................................................................iii
List of Figures.................................................................................................................................vii
List of Tables ....................................................................................................................................x
Abstract .......................................................................................................................................xi
Chapter 1. Introduction .................................................................................................................. 1
1.1 Clinical motivation: Sickle cell disease................................................................................ 1
1.2 MR imaging for OEF measurement................................................................................... 10
1.3 MR validation using Monte Carlo Simulation ................................................................... 20
Chapter 2. Anemia Increases Oxygen Extraction Fraction in Deep Brain Structures but Not
in the Cerebral Cortex................................................................................................ 24
2.1 Introduction ...................................................................................................................... 24
2.2 Methods............................................................................................................................ 25
2.3 Results............................................................................................................................... 28
2.4 Discussion.......................................................................................................................... 35
2.4 Conclusion......................................................................................................................... 40
2.4 Supplemental figure.......................................................................................................... 41
Chapter 3. Oxygen extraction fraction measurements using Asymmetric Spin Echo during
hypoxic and hypercapnic gas challenges.................................................................... 43
3.1 Introduction ...................................................................................................................... 43
3.2 Methods............................................................................................................................ 44
3.3 Results............................................................................................................................... 50
3.4 Discussion.......................................................................................................................... 55
Chapter 4. Estimation of the Scaling Coefficient by Applying Realistic Models in Monte
Carlo Simulation.......................................................................................................... 59
4.1 Introduction ...................................................................................................................... 59
4.2 Methods............................................................................................................................ 60
4.3 Results............................................................................................................................... 64
4.4 Discussion.......................................................................................................................... 67

vi
Chapter 5. Conclusion................................................................................................................... 68
References .................................................................................................................................... 71

vii
List of Figures
Chapter 1
1.1 Red blood cell morphology in healthy and sickle cell disease subjects.......................... 1
1.2 Two types of strokes and silent cerebral infarct (SCI)..................................................... 3
1.3 Cumulative prevalence of SCI in children and young adults with SCD............................ 4
1.4 Illustration of oxygen flow though artery, capillary, and vein ........................................ 6
1.5 Schematic representation of cerebral blood flow (CBF) variations associated with
changes in mean arterial pressure (MAP), arterial carbon dioxide tension (PaCO2)
and arterial oxygen tension (PaO2).................................................................................. 7
1.6 Correlations between arterial oxygen concentration (CaO2) during 10%
FiO2 hypoxic challenge and (A) cerebral blood flow (CBF), (B) arteriovenous
O2 saturation difference (A-V O2), (C) cerebral metabolic rate of oxygen (CMRO2)....... 8
1.7 Mechanistic model for increasing stages of hemo-metabolic impairment in SCD ......... 9
1.8 The bovine T2-Y calibration model. (A) 3D mesh plot showing the dependence of
blood T2 on Y and Hct; (B) Exemplar T2-Y conversion curves extracted from the 3D
mesh plot....................................................................................................................... 11
1.9 The relationship between venous oxygen saturation and oxygen content measured
by three calibration models. The venous oxygen saturation increases with oxygen
content by bovine model (left). However, an opposite trending is revealed by the
HbS Model (right)........................................................................................................... 12
1.10 The cylindrical coordinate system for describing the magnetic field at a point (r, ,
). The cylinder is at an angle to the main B0 field.................................................... 13
1.11 Representative QSM processing steps from complex gradient echo MR data. (A)
Phase image. (B) Magnitude image. (C) ∆ map is generated from the phase map.
(D) Field map after removing background field (E) Final susceptibility map after
inversion. ....................................................................................................................... 15
1.12 (A) The cylindrical coordinate system for describing the magnetic field at a point (r,
, ). The cylinder, with radius a, is at an angle to the main B0 field. (B)
Schematic drawing shows the cross-section of a cylindrical object with radius a,
enclosed by four coaxial pseudo cylinders whose radii are R3, R3‘, R2, and R1. ........... 17
1.13 The Monte Carlo Simulation model. Arteries are indicated by red cylinders and
veins are blue cylinders. (Capillaries are not shown here)............................................ 22
Chapter 2

viii
2.1 Representative image and region of the internal cerebral vein (ICV, highlight by red
rectangle). (A) Magnitude in axial view. (B) Magnitude in coronal view. (C) Axial view,
and (D) Sagittal view of ROI in the max intensity projection of a representative
susceptibility map.......................................................................................................... 31
2.2 Group differences for OEF measurements by CISSCO, QSM and TRUST. (A) Boxplot
for OEF-QSM in internal cerebral vein for SCD, ACTL and CTL. (B) Boxplot for OEFCISSCO in internal cerebral vein for SCD, ACTL and CTL. (C) Boxplot for OEF-TRUST in
sagittal sinus for SCD, ACTL and CTL. (* denoted statistically significant p < 0.05; NS
denoted no significant difference) ................................................................................ 32
2.3 (A) Scatter plot of OEF-CISSCO and OEF-QSM with linear correlation line (solid) and
identity line (dashed) (r2 = 0.72, p < 0.001). (B) Bland-Altman plot for OEF-CISSCO
and OEF-QSM. The linear correlation line (solid) is shown (r2 = 0.61, p < 0.001) ......... 32
2.4 Relationship between OEF and O2 content in SCD, ACTL and CTL. (A) Scatterplot
between OEF-CISSCO and O2 content. The fitting reciprocal line is shown in black
with r2 = 0.86, p < 0.001. (B) Scatterplot between OEF-QSM and O2 content. The
fitting reciprocal line is shown in black with r2 = 0.48, p < 0.001.................................. 33
2.5 (A) Relationship between OEF-TRUST with left shifted hemoglobin. Linear
correlations are shown in blue line (r2 = 0.46, p = 0.0027) for ACTL and red line (r2 =
0.35, p = 0.0018) for SCD. The control group is shown as 36.8 ± 5.5 (mean ± std) in
green. (B) Relationship between corrected OEF-TRUST with O2 content..................... 34
2.6 Relationship between OEF and O2 content by CISSCO (OEF-CISSCO) and reference
data (OEF-ASE)............................................................................................................... 41
2.7 (A) Relationship between OEF-TRUST with left shifted hemoglobin using the mixture
model. Linear correlations are shown in blue line (r2 = 0.46, p = 0.0027) for ACTL and
red line (r2 = 0.27, p = 0.008) for SCD. The control group is shown as 36.8 ± 5.5 (mean
± std) in green. (B) Relationship between corrected OEF-TRUST with O2 content using
the mixture model. Figure S2. (C) Relationship between OEF-TRUST with left shifted
hemoglobin using the HbA model. Linear correlations are shown in blue line (r2 =
0.46, p = 0.0027) for ACTL and red line (r2 = 0.15, p = 0.0595) for SCD. The control
group is shown as 36.8 ± 5.5 (mean ± std) in green. (D) Relationship between
corrected OEF-TRUST with O2 content using the HbA model ....................................... 42
Chapter 3
3.1 Representative end tidal O2 (EtO2) and CO2 (EtCO2) for one subject. Five phases are
shown in the figure: baseline, hypoxia1, hypoxia2, hypercapnia, and recovery .......... 44
3.2 Representative ASE signal decay. The fitting line for the last 21 echoes are shown by
the dashed line. And the spin echo value (mean of the first 7 echoes) is shown by
the red dot. The slope of the fitted line is R2’, and the difference between spin echo

ix
value and the intercept of the fitted line is . Three figures of magnitude at
different echoes are also shown (at echo1, echo 9, and echo 27)................................ 45
3.3 The Monte Carlo Simulation model. (A) Arteries are indicated by red cylinders and
veins are blue cylinders. (Capillaries are not shown here) (B) The cylindrical
coordinate system for describing the magnetic field at a point (r, , ). The cylinder
is at an angle to the main B0 field.............................................................................. 48
3.4 EtO2 and EtCO2 for individuals during each phase are shown by dashed lines. Mean
values and standard deviations are shown by solid black line and error bar.
(Hypoxia1 and Hypoxia2 represent mild and severe hypoxia, respectively) ................ 51
3.5 SpO2 and TOI for individuals during each phase. Mean values and standard
deviations are shown by black line and error bar ......................................................... 51
3.6 Representative OEF maps during each phase. The average OEF values are indicated
in each map.................................................................................................................... 52
3.7 (A) OEF values measured by TRUST and ASE during each phase. (B) OEF values
measured by NIRS during each phase. Mean values and standard deviations are
shown in blue for TRUST, red for ASE, and magenta for NIRS. (* denotes statistical
significance p<0.01, compared with baseline) .............................................................. 53
3.8 Simulation results for both models. Actual (black solid line) represents the
experimental mean ASE values. Cylinder (gray solid line) represents the predicted
results by simple cylinder model. The rest four dashed lines represent predicted
results by two models under two assumptions, respectively ....................................... 54
Chapter 4
4.1 The Monte Carlo Simulation model. (A) Arteries are indicated by red cylinders and
veins are blue cylinders. (Capillaries are not shown here) (B) The cylindrical
coordinate system for describing the magnetic field at a point (r, , ). The cylinder
is at an angle to the main B0 field. (C)(D) Fingertip pulse oximetry (SpO2) for
desaturation and resaturation....................................................................................... 61
4.2 (A) The relationship between ∆2,
∗
and concentration of deoxygenated
hemoglobin ∆[] at the operating point for the simple cylinder model. The slope
of the linear fitting is . (B) One representative fitting curve between and
(1 − )
2
for Sharan model .......................................................................................... 64
4.3 Separate relationship between and Sa, OEF, CBV, and Hb ................................ 65

x
List of Tables
Chapter 1
1.1 Summary of main MRI techniques measuring OEF ....................................................... 20
Chapter 2
2.1 Subject demographics. Group averages and standard deviations are given ................ 29
2.2 Differences between hemoglobin F, hemoglobin S and OEF by three methods with
or without hydroxyurea (HU) for non-transfused SCD patients.................................... 30
2.3 OEF predictors by stepwise regression.......................................................................... 35
Chapter 3
3.1 Vessel configuration for vascular simulations in this study .......................................... 49
3.2 Simulation results for OEF changes during hypoxia ...................................................... 55
Chapter 4
4.1 Diameter, volume fraction, and oxygen saturation for the Sharan and Sava model.... 63
4.2 Calculated under normal condition, desaturation, and resaturation by different
models............................................................................................................................ 66

xi
Abstract
Sickle cell disease (SCD) manifests as a hereditary disorder characterized by anomalous
hemoglobin that undergoes polymerization in response to deoxygenation, resulting in the
formation of inflexible, sickle-shaped red blood cells. The recurring sickling of red blood cells
contributes to the development of anemia and vasculopathy, with the most severe repercussions
observed in cerebral tissues. Notably, silent cerebral infarction (SCI) emerges as a prevalent and
progressive complication in SCD, presenting substantial associations with heightened stroke risks
and neurocognitive impairments. Despite these associations, the precise prediction and
underlying pathogenesis of SCI remain enigmatic. This thesis endeavors to establish a cerebral
oxygenation marker for the prospective prediction of SCI and concurrently explores the
correlation between brain oxygen extraction fraction (OEF), SCD, and SCI. In addition, Monte Carlo
simulation is applied to demonstrate the details regarding microvascular architecture and blood
oxygenation. The logical framework of this research aims to enhance our comprehension of the
intricate interplay between these factors within the context of sickle cell disease.
Central to the oversight and control of stroke risk for patients with sickle cell disease is the
systematic examination of oxygen supply and utilization within the cerebral domain.
Consequently, the first chapter of this thesis introduces the clinical motivation for this thesis and
current non-invasive MRI techniques designed for the quantification of oxygen extraction fraction.
To quantify regional oxygen extraction, particularly deep structures of the brain, Chapter
2 proposes two susceptibility-based MRI techniques, Quantitative Susceptibility Mapping (QSM)
and Complex Image Summation around a Spherical or a Cylindrical Object (CISSCO), applied in the
internal cerebral vein (ICV). The results are also compared with the OEF from the same cohort
measured by T2-Relaxation-Under-Spin-Tagging (TRUST), indicating anemia worsens hypoxia in
deep brain structures while cerebral cortex appears to be spared.
To get more detailed information of OEF in brain tissues, Chapter 3 compares and crossvalidates two popular MR techniques, TRUST and Asymmetric Spin Echo (ASE), under hypercapnic
and hypoxic challenges. In addition, we apply a Monte Carlo Simulation of the ASE signal to model
the OEF changes by ASE during these gas changes to gain insights into any potential disparities
between the two methods. Both methods reveal that OEF decreases during hypercapnia. However,
OEF predictions diverge during hypoxia. TRUST OEF estimates decreased with both mild and
moderate hypoxia, while ASE OEF stayed the same or increased. This disparity is also validated by
the Monte Carlo simulation applied in two physiologically realistic models.
Chapter 4 extends the concept of Monte Carlo simulation and estimates the intrinsic tissue
relaxivity coefficients by using a realistic distribution of vessel sizes and oxygen saturations. We
investigate the effects of hemoglobin, OEF, CBV, and arterial saturation on the intrinsic tissue
relaxivity coefficientscaling R2* by BOLD to changes in deoxygenated hemoglobin concentration.

xii
We also derive an empirical model based on these parameters to allow the calculation of r2*tissue
for any combination of Sa, OEF, Hb, and CBV.
Overall, this thesis posits innovative non-invasive methodologies for the quantification of
cerebral oxygenation. In silico simulations with physiologically realistic models are also applied to
validate these findings. These techniques are subsequently employed within a patient cohort
afflicted with sickle cell disease to scrutinize their susceptibility to stroke. Comprehensive
advancements are imperative, encompassing both technical refinement and physiological
elucidation. This includes further validation across diverse pathologies and the incorporation of
routine hemodynamic screening within a comprehensive stroke prevention program, particularly
targeted at vulnerable populations.

1
Chapter 1: Introduction
1.1 Clinical motivation: sickle cell disease
1.1.1 Basics of sickle cell disease
Sickle Cell Disease (SCD) is the most common genetic disorder in the world. It is caused
by a single base pair mutation in the beta subunit of hemoglobin, a protein responsible for
carrying oxygen in the red blood cells [1], [2]. The mutation leads to the production of abnormal
hemoglobin S (HbS) and polymerization or aggregation of HbS after deoxygenation [3], [4]. The
polymerization of HbS distorts normally round and flexible red blood cells into a sickle or crescent
shape. These sickled red blood cells are less pliable and more fragile, leading to blood flow change
and vaso-occlusion (Figure 1.1). SCD complications include chronic anemia and cerebral
vasculopathy, resulting in a stroke, silent cerebral infarction (SCI), and cognitive dysfunction [5].
Figure 1.1. Red blood cell morphology in healthy and sickle cell disease subjects. Sickle
cells are less pliable and easier to block blood flow.
SCD affects 30,000 births each year worldwide and there are more than 100,000 SCD
patients in the United States [6]. The estimated life expectancy is about 54 years [7] for the SCD
cohort, while the non-sickle cell disease cohort has a projected life expectancy of 76 years.
1.1.2 Cerebrovascular Disease in SCD
Cerebrovascular affliction represents a prevalent and profoundly incapacitating
consequence of sickle cell disease (SCD) [1]. A fundamental contributory factor in the genesis of
cerebrovascular disorders within the SCD context resides in the aberrant sickling of red blood

2
cells (RBCs), although the underlying mechanisms exhibit marked heterogeneity. The propensity
of sickled RBCs for hemolysis, culminating in anemia, constitutes a pivotal facet. Furthermore,
these aberrant cells manifest anomalous adherence to the endothelium, thereby precipitating
vaso-occlusion phenomena [8] that predominantly manifest at the microvascular level. The
release of injurious bioactive substances from compromised sickle cells further impels a
hypercoagulable milieu and undermines vasomotor functionality. Sequential episodes of vasoocclusive incidents coupled with subsequent reperfusion instances potentiate incremental
vascular inflammation [9], culminating in consequential vascular impairments [10]. The ultimate
sequelae encompass multifaceted large-vessel vasculopathy and microvascular occlusion,
synergistically engendering a catastrophic clinical outcome intrinsic to SCD—namely, stroke [11],
[12].
Stroke manifests in distinct forms, mainly ischemic stroke, and hemorrhagic stroke [12],
[13], each bearing unique etiological underpinnings and pathophysiological signatures. Ischemic
stroke [14] arises from the occlusion or severe reduction of blood flow to a specific brain region,
often stemming from thrombotic or embolic events. This cessation of adequate perfusion triggers
neuronal ischemia, precipitating a cascade of excitotoxicity, cellular energy failure, and ultimately,
irreversible tissue damage. Conversely, hemorrhagic stroke [15], [16] materializes when the
integrity of cerebral blood vessels is compromised, leading to extravasation of blood into the
brain parenchyma or subarachnoid space. Subarachnoid hemorrhage, frequently attributed to
ruptured cerebral aneurysms, and intracerebral hemorrhage, associated with vessel ruptures
within the brain tissue, represent the cardinal subtypes.
The other common variant of stroke associated with sickle cell disease (SCD), recognized
as silent stroke or silent cerebral infarction (SCI) [17], [18], denotes instances of cerebral
infarction that manifest discernibly through MRI imaging but do not elicit evident focal
neurological deficits. Characteristically, these silent stroke lesions exhibit diminutive dimensions,
measuring at a minimum of 3 mm in their greatest linear extent, and are appreciable across at
least two planes of T2-weighted images [19]. This predilection for localization in the deep
supraventricular white matter aligns with the internal border zone [18]. Available data elucidates
a cumulative risk of 19.2% for silent stroke occurrence by the age of 8 years, ascending to 32.4%
by age 14 years, and further escalating to 39.1% by age 18 years, with the incidence persistently
exceeding 40% in the adult SCD population [20]. Despite the absence of overt neurological
manifestations, silent strokes in pediatric SCD cohorts are notably linked to neurocognitive
impairment [19], suboptimal academic performance, and portentous propensities towards
future overt stroke occurrences.

3
Figure 1.2. Two types of strokes and silent cerebral infarct (SCI).
Through the progression of cerebral imaging technologies and the culmination of rigorous
randomized clinical trials (RCTs) [21], [22] concerning both primary and secondary stroke
prevention, the inherent course of stroke within the context of sickle cell disease (SCD) is
undergoing transformative evolution. Randomized clinical trials (RCTs), including the Stroke
Prevention in Sickle Cell Anemia (STOP) and STOP II trials [23], [24], have definitively ascertained
the efficacy of periodic blood transfusion therapy, often administered on a monthly basis, in
achieving primary stroke prevention among pediatric patients afflicted with sickle cell anemia
(SCA) exhibiting elevated transcranial Doppler (TCD) velocities [25], [26]. Subsequently, after a
minimum duration of one year, hydroxycarbamide [27]–[29] has emerged as a viable substitute,
as underscored by the TCD With Transfusions Changing to Hydroxyurea (TWiTCH) trial [30], [31].
Notably, analogous high-income countries have also witnessed the conclusive outcomes of RCTs
delineating regular blood transfusion as the prevailing therapeutic regimen for secondary
prevention of infarcts in children diagnosed with SCA and strokes, exemplified by the Stroke With
Transfusions Changing to Hydroxyurea (SWiTCH) trial [32], [33], as well as for mitigating silent
cerebral infarcts, as illuminated by the Silent Infarct Transfusion (SIT) Trial [20], [34].
Despite the notable magnitude by which transfusion ameliorates the susceptibility to
primary strokes, individuals affected by SCD remain vulnerable to the occurrence of SCIs. The
prevalence of SCIs within the SCD demographic reaches 53% at the juncture of 30 years of age
[20], and intriguingly, this propensity does not exhibit a plateau as the cohort progresses into
adulthood (Figure 1.3).

4
Figure 1.3. Cumulative prevalence of SCI in children and young adults with SCD. The
cumulative prevalence of SCI increases 1-2% per age year with no plateau towards young
adulthood despite appropriate TCD screening.
Unlike the unequivocal linkage observed in cases of overt stroke, a discernible connection
between the existence of silent cerebral infarctions (SCIs) and aberrant transcranial Doppler (TCD)
measurements remains notably absent [35]. While investigations have demonstrated robust
correlations between SCIs and diminished hemoglobin levels [20], relative hypertension [19],
episodes of acute anemia [17], and anomalous magnetic resonance angiography (MRA) findings,
the utilization of these risk factors either in isolation or in combination still falls short of furnishing
precise prognostication of SCI occurrence. The core clinical impetus of this thesis resides in the
quest for more precise prognosticators of SCIs and a comprehensive comprehension of the
incremental evolution characterizing the trajectory of SCIs.
1.1.3 Current treatment for SCD
Current treatment of SCD [36]–[38] aims to alleviate symptoms and prevent
complications, encompassing a combination of medicines, transfusions, blood and bone marrow
transplant, and potential gene therapy treatments. Several medications have shown promise in
the treatment of SCD, and four representative medicines are introduced here.
❖ Voxelotor [39], [40] is used to treat SCD in adults and children over 4 years old,
and it functions as an inhibitor of hemoglobin S (HbS) polymerization, establishing
a binding interaction with HbS in a stoichiometric ratio of 1:1. It manifests a
predilection for selective partitioning within RBCs. Through augmentation of the
oxygen-binding affinity of hemoglobin, voxelotor engenders a dose-dependent
impediment of HbS polymerization.
❖ Crizanlizumab-tmca [41] is approved for adults or children over 16 years old with
SCD, targeting to reduce vaso-occlusive and pain crises by binding P-selectin on

5
the surface of activated endothelium and platelet cells blocks interactions
between endothelial cells, platelets, RBCs, and leukocytes.
❖ Hydroxyurea [27], [29] has been shown to be the most effective disease-modifying
therapy for both adults and children with SCD, and it can reduce or prevent
multiple complications. The underlying mechanism of hydroxyurea's effectiveness
in treating SCD lies in its ability to stimulate the production of fetal hemoglobin
(HbF), a type of hemoglobin that is typically present at birth but decreases as a
person grows older and HbF has a higher affinity for oxygen than the mutated
hemoglobin (HbS). This serves to dilute the concentration of the abnormal HbS
within red blood cells, thereby reducing the propensity of these cells to deform
and form the characteristic sickle shape.
❖ The FDA-approved L-glutamine [42], [43] mainly aims to reduce severe
complications and lower the number of pain crises by augmenting the ratio of
reduced nicotinamide adenine dinucleotides within erythrocytes affected by SCD.
Research shows that patients taking L-glutamine have fewer hospital visits than
patients taking a placebo. Despite the effectiveness of these medicines, there are
potential side effects, including headache, nausea, fatigue, chest pain, or even
severe symptoms.
Besides medicines, transfusion therapy [44], [45] plays an important role in the
comprehensive management of SCD. Transfusions, including chronic and acute transfusions, are
employed to ameliorate complications and mitigate the effects of SCD. Chronic transfusions
involve regular administration of packed red blood cells (HbA) to patients with SCD, targeting to
raise the hemoglobin level and reduce the risk of stroke and anemia. By diluting the
concentration of HbS lower than 30% in the bloodstream, chronic transfusions help prevent the
formation of sickle-shaped red blood cells. On the other hand, acute transfusions are used to
treat complications that cause severe anemia, especially in acute situations. Risks related to
transfusions include alloimmunization [33], which makes it hard to find future donors, infection,
and excess iron deposit.
A blood and bone marrow transplant [46], [47], also called a hematopoietic stem cell
transplant, is the only cure for SCD. Despite a high successful rate in children, many SCD patients
are hard to find a close genetic match. Potential genetic therapy treatments [48], [49] are also
explored to cure for SCD, by adding new DNA or changing existing DNA.
1.1.4 Clinical need for OEF measurements in SCD
The human brain constitutes 2% of the total body mass, but consumes 20% of the oxygen
and accounts for 15% of the cardiac output [50], [51]. This high oxygen demand of the brain is
due to its essential functions and continuous activity, emphasizing its importance in preservation
for the body’s survival. The cerebral metabolic rate of oxygen (CMRO2) plays a vital role in
maintaining optimal brain function, reflecting the balance between oxygen supply and demand
[52]. CMRO2 has been recognized as a fundamental parameter for assessing brain health and

6
function, providing insights into brain metabolism in various cerebrovascular diseases, including
anemia[53], [54], stroke [55], and Parkinson’s Disease [56], [57]. Understanding CMRO2 dynamics
and its relationship with cerebral blood flow (CBF) and oxygen extraction fraction (OEF) is crucial
for unraveling the intricate mechanisms underlying brain physiology. By Fick’s principle, CMRO2
can be modeled as:
CMRO2 = (CaO2 – CvO2) * CBF ≈ CaO2 * CBF * OEF,
where CaO2 and CvO2 are oxygen contents in arterial blood and venous blood. The right
hand is an approximation, neglecting the dissolved oxygen, which is valid under normal
conditions. The arterial and venous oxygen contents are related to the hemoglobin level (Hb),
oxygen saturation (Ya and Yv for arterial and venous respectively, Figure 1.4), and partial pressure
of oxygen (PaO2 and PvO2 respectively):
CaO2 = 1.34 * Hb * Ya + 0.003 * PaO2
CvO2 = 1.34 * Hb * Yv + 0.003 * PvO2
And OEF is defined as:
OEF = (Ya - Yv) / Ya
Often, the term oxygen delivery (DO2) is used to indicate the product of CBF and oxygen
content in arterial blood.
DO2 = CBF * CaO2
Figure 1.4. Illustration of oxygen flow though artery, capillary, and vein. The oxygen saturation
at artery and vein are Ya and Yv, respectively. OEF is calculated as OEF = (Ya - Yv) / Ya.
Many studies have shown that CMRO2 is kept stable during different brain activities [58]
and gas challenges [59], [60], on the order of 3 ml O2/100g tissue/min. Therefore, the balance
between oxygen supply and demand is guaranteed under normal condition. Regulating this

7
balance is extremely complex and is determined by a number of factors, such as arterial carbon
dioxide tension (PaCO2) [61], arterial oxygen tension (PaO2), viscosity of blood, the diameter of
cerebral blood vessels, the net pressure of the blood flow into the brain, known as cerebral
perfusion pressure (CPP) [62]. This interaction is primarily achieved by modulating CBF in
response to CO2. CBF is extremely sensitive to changes in arterial PaCO2, displaying marked
increases during moderate hypercapnia and reduction during hypocapnia. The relationship is
shown by the green dotted line in Figure 1.5.
Figure 1.5. Schematic representation of cerebral blood flow (CBF) variations associated
with changes in mean arterial pressure (MAP, red line), arterial carbon dioxide tension (PaCO2,
green dotted line) and arterial oxygen tension (PaO2, blue dotted line). Autoregulation is achieved
in the range of 60-150 mmHg by vasodilation (left) and vasoconstriction (right). Adapted from
[63].
In face of deranged oxygen delivery, CMRO2 is kept stable through two stages of
compensation, involving adjustments in both CBF and OEF [64], [65]. Oxygen delivery can be
disrupted by changes in oxygen content or CBF. Since oxygen content is just the product of
hemoglobin level and oxygen saturation, either anemia or hypoxia can decrease oxygen content.
In both cases, compensatory hyperemia happens to diminish this effect [66], [67]. It has been
shown by our lab and others that hyperemia is able to maintain normal levels of oxygen delivery
to the brain in chronically anemic patients [68], [69]. As shown in Figure 1.6, some researchers
found that the hyperemia under hypoxia is able to maintain the oxygen delivery and there is even
over-compensation.

8
Figure 1.6. Correlations between arterial oxygen concentration (CaO2) during 10%
FiO2 hypoxic challenge and (a) cerebral blood flow (CBF), (b) arteriovenous O2 saturation
difference (A-V O2), (c) cerebral metabolic rate of oxygen (CMRO2). Adapted from [70].
Oxygen delivery can also be disrupted by reversible or irreversible changes in CBF [71].
For example, fixed vascular stenoses may lower CBF and prevent its increase in response to stress
[72]. In this scenario, the only way to preserve CMRO2 is to increase oxygen extraction.
Reversible changes in oxygen delivery occur every time we change our head position with
respect to gravity or have spontaneous fluctuations of our blood pressure for other reasons. The
net pressure of the blood flow into the brain, known as cerebral perfusion pressure (CPP) [73],
which is calculated as the difference between mean arterial pressure (MAP) and intracranial
pressure (ICP) or the central venous pressure (CVP), whichever is the higher. Cerebral vasculature
possesses the capability to uphold cerebral blood flow (CBF) at a consistent level across a
substantial spectrum of mean arterial pressure (MAP) variations, spanning from 60 to 150 mmHg
[74]. This homeostatic equilibrium is achieved by modulating the diameters of cerebral vessels
through a physiological mechanism termed "cerebral autoregulation" (CA) [64], [75], [76]. In
physiological circumstances, an elevation in MAP engenders vasoconstriction, whereas a
decrement in MAP elicits vasodilation.
It has been show that patients with SCD have compromised red blood cells and arterial
oxygen content [77]. Hence, CBF is increased [69], [78] to compensate for the decreased oxygen
content through autoregulation. This CBF increase in SCD patients has been proved by several
MRI techniques, including phase contrast (PC) [68], arterial spin labeling (ASL) [79], [80] and
gadolinium-enhanced perfusion imaging [81].
However, the CBF increase in SCD patients comes with a cost. Cerebrovascular reserve
(CVR) is defined as the proportion that CBF can change in response to metabolic demands [82],
[83]. The increased CBF in SCD patients leads to a lower CVR [84], leaving the brain vulnerable to
ischemic insults or acute stresses [85].
It’s also worth noticing that oxygen delivery is not increased at the same level for the
whole brain of SCD patients, especially in the white matter and watershed zones of highest
incidence of SCI [86]. Our lab has postulated that a vascular steal phenomenon might exist, where

9
oxygen delivery to brain cortex is preserved at the expense of deep brain structures [86]. This
“cortical sparing” mechanism is further validated by the finding that deep brain structures have
higher OEF in SCD patients by using susceptibility-based MR methods [87]. And this is also
consistent with observed patterns of brain volume loss [88], [89] and silent infarction [18], [90].
Despite of the consensus on the CBF increase in SCD patients, there exists controversial
related to the CMRO2 and OEF change in SCD patients. Two PET studies [91], [92] have shown
that similar level of CMRO2 is observed between SCD patients without neurological symptoms
and healthy subjects. However, the PET study is limited by small number of participants (N = 6),
different SCD cohorts, and invasive imaging with high radiation. On the other hand, our lab has
demonstrated that both CMRO2 and OEF decrease in SCD patients by using more appropriate
calibrations for OEF [54], [93]. This finding is also supported by a compilation of historical CMRO2
datasets in a broad range of medical conditions [54]. Therefore, further investigation is needed
for the validation of measurements of CMRO2 and OEF in SCD cohorts.
It has been shown that the magnitude and direction of CBF change might be dependent
on the extent of autoregulatory capacity, in cases of notable reductions in cerebral perfusion
pressure and arterial steno-occlusion, which are frequently observed in advanced Moyamoya
associated with sickle cell anemia [94], [95]. Hence, the extent of anemia and vasculopathy can
lead to considerable heterogeneity in CBF alterations. Consequently, CBF alone may not provide
comprehensive indications of the severity of SCD. In the context of steno-occlusive disease, OEF
has been proposed as a more sensitive marker of critical tissue-level dysfunction, particularly for
cerebral ischemia and subsequent stroke. It surpasses CBF or cerebral blood volume (CBV) in its
ability to detect a broader spectrum of hemodynamic impairments [96], [97].
Figure 1.7. Mechanistic model for increasing stages of hemo-metabolic impairment in SCD.
Adapted from Jordan et al.
OEF has been recognized as a potential biomarker in various disease, such as Alzheimer’s
disease (AD) [98], [99], carotid steno-occlusive [100] and brain tumor [101]. However, studies on
the cerebral OEF in patients with SCD are scarce, compared with other oxygenation parameters
like CMRO2 and CBF. Furthermore, the results of OEF measurements are conflicting and there
has been no consensus. In summary, independent, and comprehensive study of OEF is strongly
needed.

10
1.2 MR imaging for OEF measurement in SCD
1.2.1 Three types of MRI-based techniques for OEF measurements
With recent advances in MRI, several techniques have been developed to quantify OEF
noninvasively, based on T2 of blood, susceptibility of blood, and R2’ of tissue. Within these
techniques, certain methods offer a comprehensive assessment of OEF on a global or wholebrain scale, while others focus on estimating OEF in specific regions of the brain.
As shown in the equation, OEF is dependent on the oxygen saturation on both arterial
and venous sides (Ya and Yv, respectively). Ya is easy to measure by pulse oximetry in the fingertip
[102], [103], and Ya is usually over 95% for healthy people [104]. Therefore, OEF measurement
mainly focuses on the measurement of Yv. Three representative and widely-used MRI techniques
are introduced in this section.
❖ T2-based
Oxygen saturation has a one-on-one relationship with blood T2 when the hematocrit is
fixed [105]–[107]. Therefore, Yv or OEF can be determined by the quantification of blood T2 [108].
It should be noted that pure blood signal is needed in this measurement because partial volume
effect from tissue might cause bias. This pure signal is usually achieved by increasing resolution
[109], or the design of specific sequence to isolate blood signal [110].
T2-Relaxation-Under-Spin-Tagging, or TRUST, is the most widely used T2-based MR
technique for the measurement of oxygen saturation [110], [111]. There are two steps in TRUST:
estimation of T2 and T2-Yv conversion. For the first step, pure signal from venous blood in the
superior sagittal sinus (SSS) is isolated through subtraction between control and labeled images.
T2-preparation, with varying number of refocusing pulses, is applied to quantify blood T2. Thus, a
series of T2-weighted images is acquired at effective TE (eTE), both with and without the blood
spin tagging. Venous blood T2 is estimated by a simple mono-exponential fitting of the blood
signal intensity as a function of eTE. The second step is to convert venous blood T2 into venous
blood saturation Yv through a calibration model. In studies involving individuals without
hematological disorders, the T2-Y calibration model proposed by Lu et al. [110], has emerged as
a widely adopted approach for both pediatric and adult populations. This calibration model is
based on in vitro experiments conducted on bovine blood samples. Bovine blood shares several
important characteristics with adult human blood, including comparable hemoglobin structure,
red blood cell (RBC) morphology [112], and water permeability [113]. The utilization of this
bovine model has demonstrated consistent quantifications of venous oxygen saturation (Yv) and
oxygen extraction fraction (OEF) when compared to 15O-PET measurements in healthy subjects
[114].

11
Figure 1.8. The bovine T2-Y calibration model. (A) 3D mesh plot showing the dependence of
blood T2 on Y and Hct; (B) Exemplar T2-Y conversion curves extracted from the 3D mesh plot.
Adapted from Dengrong et al.
The scan time of TRUST is about 1.2 mins, and it has presented high reproducibility and
reliability across multiple sites [115], [116], and different MR vendors [117]. TRUST has
undergone assessments of its sensitivity to changes in OEF through various physiological
challenges, including caffeine intake [118], hypercapnia [119], hypoxia and hyperoxia [120]. In
each of these challenges, TRUST consistently demonstrated the anticipated alterations in OEF,
affirming its capability to detect and measure OEF changes accurately.
However, the biggest limitation of TRUST is that the bovine model was calibrated with a
limited hematocrit range of 35%-55% [110]. While this range adequately encompasses the
majority of healthy subjects, it yields [121], [122] wildly inaccurate results when extrapolated to
anemic individuals with significantly low hematocrit (Hct) levels, even for normal red blood cells.
To counter this limitation, a calibration curve was derived from healthy human blood over a
hematocrit range of 10%-50% [93]. The bovine hand human T2 models agree well over
hematocrits of 35%-55% except for a small systematic bias, which likely results because bovine
RBCs are only half the size of human RBCs. While this human calibration is appropriate for many
forms of anemia, it does not accurately reflect T2 oximetry in patients with sickle cell disease.
Blood T2 is influenced not only by the concentration of deoxyhemoglobin in the blood but also
by factors such as red blood cell (RBC) cellularity and permeability [108]. Therefore, it is
inappropriate to use the bovine or healthy human model for SCD populations with abnormal red
blood cell morphology and permeability. Bush et al. from our lab developed sickle cell disease
specific TRUST calibrations and observed that it had dramatic impact on the predicted oxygen
extraction fraction values [93].

12
Figure 1.9. The relationship between venous oxygen saturation and oxygen content measured
by three calibration models. The venous oxygen saturation increases with oxygen content by
bovine model (left). However, an opposite trending is revealed by the HbS Model (right).
Adapted from Bush et al.
❖ Susceptibility-based
Deoxyhemoglobin is paramagnetic [123] due to the presence of unpaired electrons in its
molecular structure, allowing for the detection and characterization of venous oxygen saturation.
Spees et al. [124] performed calculations to determine the theoretical susceptibility value of a
red blood cell by considering the inherent susceptibility of deoxyhemoglobin, oxyhemoglobin,
and water molecules, as well as their respective volume fractions. The susceptibility of venous
blood can be modeled as:
= Δ × (1 − ) × + ∆ ×
where ∆ = -4pi * 0.03 ppm is the susceptibility shift of oxyhemoglobin, Δ = 4pi * 0.27 ppm
is the susceptibility difference between fully oxygenated and fully deoxygenated red blood cells.
A side note is that 0.27 is not the only value for Δ and some researchers use 0.18 for both
HbA and HbS [125]–[127]. For a healthy subject, if we assume a hematocrit value of 0.42 and a
venous oxygen saturation (Yv) of 0.6, the blood is expected to exhibit a susceptibility of
approximately 0.4 ppm. From the equation, Yv can be determined with the measurement of
susceptibility of blood.
There are mainly three kinds of susceptibility-based MR methods: susceptometry-based
oximetry (SBO), quantitative susceptibility mapping (QSM) and complex image summation
around a spherical or a cylindrical object (CISSCO).
SBO
SBO [128]–[131] is the most straightforward susceptibility-based model, which treats
blood vessels as infinite cylinder with uniform susceptibility. Blood susceptibility can be
calculated by measuring the magnetic shift of blood relative to the surrounding tissue. Deviation
from the assumption of a straight cylinder can occur when blood vessel possesses curvature,
branching, or non-circular cross-sections. To assess the impact of such deviations, numerical

13
simulations using realistic 3D models of the superior sagittal sinus demonstrated that the error
remained within 5% for vessel tilt angles below 30 degrees [128], [132]. Additionally, phantom
experiments conducted by Langham et al. [132] indicated an error of less than 2% resulting from
non-circular vessel cross-sections. Therefore, these findings lend support to the validity of the
assumption of the infinite cylinder model.
Upon introduction to a uniform magnetic field, B0, an object undergoes magnetization,
resulting in the creation of its own magnetic field that alters the initial external field. The
magnitude and distribution of the induced magnetic field are contingent upon the object's
geometry and orientation. Mathematically solving the partial differential equations with
boundary conditions is typically necessary to compute the induced field, often necessitating the
use of numerical methods [133]. However, an advantage arises in the case of ellipsoidal
susceptibility sources, as the induced magnetic fields can be described analytically. In the case of
an infinitely long cylinder inclined at an angle θ to the applied magnetic field (as depicted in Figure
1.10), the resulting induced magnetic field can be expressed as follows:
∆() = {
∆
6
(32 − 1)0 , ℎ
∆
2

2

2
2 cos 2 0 , ℎ
where cylindrical coordinates (r, , ) were used to describe the position relative to the
cylinder, and a is the radius of the cylinder.
Figure 1.10. The cylindrical coordinate system for describing the magnetic field at a
point (r, , ). The cylinder is at an angle to the main B0 field.
In the conventional 2D implementation, the short scan time (approximately 20 seconds)
of SBO allows for high temporal resolution in quantifying venous oxygen saturation [131]. When
interleaved with phase-contrast (PC) CBF measurement, SBO has been employed to rapidly

14
measure global CMRO2 both at baseline and during challenges such as hypercapnia and volitional
apnea [134]–[136].
However, one of the primary and biggest challenges encountered in SBO is the removal
of the background field induced by the air-tissue interface, commonly referred to as the
"background field". Various approaches have been proposed to address this issue, with high-pass
filtering being suggested to eliminate the low spatial frequency components of the background
field. However, the sensitivity of the measurement to the filter size has been observed [137]. An
alternative approach involves fitting the background field inhomogeneity to a second-order
polynomial, which has demonstrated reasonable results in measuring venous oxygen saturation
(SvO2) in the femoral vein [137]. However, the background field near the superior sagittal sinus,
situated at the tissue boundary, exhibits a more complex pattern, and complicates the problem.
Besides the difficulty in background correction, another limitation of SBO is that the
measurement is sensitive to the slice position, as revealed by Miao et al [138].
QSM
QSM [139]–[141] takes advantage of the susceptibility differences between tissues to
generate high-resolution maps that provide valuable information about tissue composition and
microstructure. By measuring the phase information of MRI data and applying sophisticated
algorithms, QSM can estimate the local magnetic susceptibility values, which are directly related
to the tissue's underlying composition, such as iron content, myelin content, and calcifications.
This technique has widespread applications in various diseases, including multiple sclerosis [142],
Parkinson’s [143], and Alzheimer’s disease [144]. It’s also used to investigate the brain iron
deposition [145] and measure cerebral OEF [146], [147].
Unlike the infinite cylinder assumption in SBO, QSM can calculate the susceptibility of
tissues with arbitrary geometry and orientation. This is realized by the summation of the dipole
fields generated by each element in the magnetization distribution. The induced magnetic field
at any position b(r) can be written as a convolution of the susceptibility distribution, (), and a
kernel function, d(r):
∆() = () ∗ () 0
where the kernel function (or “dipole kernel”) is:
() =
1
4
32 − 1

3
By applying a Fourier Transform, we can change ∆(), (r) and d(r) into ∆(), X(k) and
D(k), which correspond to a decomposition into their spatial frequencies k.
∆() = ()()
And

15
() =
1
3
−

2

2 +
2 +
2
A side note is that D(k) has zeros along the angle of 54.7 degrees (often called “magic
angle”), which makes the inverse problem more complex.
QSM usually involves several essential processing steps (Figure 1.11):
1. Data acquisition: 3D GRE sequence with high spatial resolution and multi-echo is usually
used for B0 field mapping. Both magnitude and phase information are captured.
2. Preprocessing: The acquired MRI data undergoes preprocessing steps to correct for
artifacts and enhance the image quality. This may involve motion correction, phase unwrapping
to remove phase ambiguities, and background field removal to eliminate non-local magnetic field
contributions.
3. Field-to-susceptibility inversion: The main computational step in QSM is the conversion
of the measured magnetic field to tissue susceptibility. This process involves solving an inverse
problem through a 3D dipole convolution to estimate the local susceptibility distribution based
on the acquired phase data.
Figure 1.11. Representative QSM processing steps from complex gradient echo MR data.
(a).Phase image. (b). Magnitude image. (c). ∆ map is generated from phase map. (d) Field map
after removing background field (e). Final susceptibility map after inversion. Adapted from
[148].
In this way, a pixel-wise susceptibility map can be generated bypassing the complex and
ill-conditioned inverse problem. However, there exist two main challenges in the QSM processing:
background field removal and field-to-susceptibility inversion, which might have a big effect on
the susceptibility measurements. The background field is originated from the global geometry,
air-tissue interfaces, and system imperfection. Three methods are proposed to remove the
background field: projection onto dipole fields (PDF) [149], sophisticated harmonic artifact

16
removal for phase data (SHARP) [150], regularization enabled SHARP (RESHARP) [151], and
Laplacian boundary value (LBV) [152].
In order to address the challenging nature of the inverse problem at hand, researchers
have proposed multiple approaches. These include the utilization of multi-orientation acquisition
techniques as well as the development of various regularization algorithms specifically designed
for single-orientation acquisition. There are mainly four algorithms: calculation of susceptibility
through multiple orientation sampling (COSMOS) [153], [154], truncated k-space division (TKD)
[155], L1-QSM [156], and morphology-enabled dipole inversion (MEDI) [157].
CISSCO
Susceptibility measurements of veins have been conducted using quantitative
susceptibility mapping (QSM) techniques. However, accurately quantifying the susceptibility of
small cylindrical objects becomes challenging as their size decreases [158], primarily due to the
effects of partial volume. To address this limitation, an alternative method known as CISSCO
(Complex Image Summation around a Spherical or Cylindrical Object) is developed by Ching-Yi
Hsieh et al [159]–[161]. The CISSCO method allows for the quantification of magnetic moment,
susceptibility, and the size of narrow cylindrical objects, even in cases where partial volume
occurs. Notably, the susceptibilities obtained through the CISSCO method may deviate by
approximately 5% from the expected values, whereas other QSM methods may exhibit deviations
of up to 30% [162].
The CISSCO method integrates the complex MR signals in three annuli around the cylinder
of interest. The complex sums were cast into equations containing three unknown parameters,
the susceptibility and radius of the vessel, and the proton spin density. The overall MR complex
signal S within a coaxial cylinder with radius R was :
= 0 ∫

2
0()

′
/2
+
20,

where a is the vessel radius, = −
∆
6
(32 − 1)0, ∆ is the susceptibility difference
between tissues inside and outside, is the slice thickness of the image, 0 and 0, are the
effective spin densities of the tissue outside and inside the object, is the effective magnetic
moment,
′ = (0.50∆) ∗ 2 is the extremum phase value the surface of the object, and
is the orientation of the cylinder.

17
Figure 1.12. (a). The cylindrical coordinate system for describing the magnetic field at a point (r,
, ). The cylinder, with radius a, is at an angle to the main B0 field. (b). Schematic drawing
shows the cross section of a cylindrical object with radius a, enclosed by four coaxial pseudo
cylinders whose radii are R3, R3‘, R2, and R1. Adapted from Heish et al.
The equation can be applied to all three annuli, allowing us to solve for the unknowns; complex
signal differences between any two annuli eliminate the second term, eliminating two variables.
First, the magnitude and phase images in coronal view are cropped to 64*64 and a 16*16
Gaussian high pass filter is applied to remove background phase. Next, is estimated from the
innermost annuli. It is usually close to 90 degrees since the internal cerebral vein is nearly
perpendicular to the direction of B0. Finally, after applying the equation to three coaxial cylinders,
the effective magnetic moment, the effective spin densities, and the susceptibility difference can
be solved sequentially. The resulting  is converted to oxygen saturation using equation 1, the
same as for QSM.
❖ R2’-based
In contrast to the T2-based and susceptibility-based methods that primarily consider the
impact of blood oxygenation on signals within blood vessels ("intravascular" signals), the R2’
model specifically examines the signal decay occurring in the surrounding "extravascular" space.
This decay is attributed to the local field inhomogeneities caused by the presence of
paramagnetic deoxyhemoglobin in the vessel network. This is similar to the BOLD effect, but the
effects of oxygenation level and blood volume can’t be separated in BOLD. Yablonskiy and his
colleagues proposed the MR signal behaves in the static dephasing regime [163] and assumes
the brain as a two-compartment model: intravascular and extravascular. And they also assumed
the intravascular space as an ensemble of randomly oriented cylinders with infinite length [163],
[164].

18
Based on these principles and assumptions, Dr. An and her colleagues at the University of
St. Louis developed the well-known MR technique called asymmetric spin echo (ASE) [165]–[167].
ASE can be viewed as a modification of SE in which the acquisition window is shifted in time
relative to the center of spin echo, increasing dephasing and creating signal decay with R2’
contrast. The signal decay can be calculated as:
S(τ) = (1 − )(, , )exp (−/2)(, 1, R)
where is the spin density, is the venous blood volume fraction, is the time shift from the
expected echo formation, and is the frequency shift induced by the microscopic susceptibility.
In the presence of randomly oriented cylinders containing de-oxygenated hemoglobin, can
be written as follows:
δw = 4/3π ∗ γ ∗ Δχ ∗ Hct ∗ B0 ∗ OEF
R2’ = δw * λ
where γ is the gyro-magnetic ratio, B0 is the main magnetic field strength, Hct is the fractional
hematocrit and Δχ is the susceptibility difference between fully oxygenated and fully
deoxygenated blood. The reversible relaxation rate, R2’, is simply the product of and . With
constant TE and a sufficiently long TR, the signal decay can be simplified to two relaxation regimes
separated by 2 = 1.5/:

() = ∗ exp(− ∗ ∗ 2 + )

() = ∗ exp (−0.3( ∗ 2)
2
)
R2’ and can be estimated calculating the slope and intercept of plot of log (SL(t)) versus
2. And OEF can be calculated through the above equations.
ASE holds promise across a wide spectrum of applications, spanning from clinical
evaluations of muscular disorder [168], [169] and renal assessments [170] to investigations of
cerebral hemodynamics [171], [172]. Notably, ASE has been extensively utilized to explore the
metabolic status of the brain in individuals with SCD [18], [27], [173]. Moreover, ASE-based
measurements of OEF have unveiled a co-localization between regions exhibiting high oxygen
extraction and deep white matter areas characterized by high sickle cell infarct (SCI) density [90].
This intriguing finding suggests that ASE has the potential to identify individuals at risk of stroke
and to detect perfusion territories most susceptible to SCI development.
However, there are several limitations of this method. First, the assumption of static
dephasing regime, during which diffusion effects are neglected, might not be valid. It has been
shown by Monte Carlo simulation that diffusion introduces additional signal decay depending on
the vessel size, thus leading to overestimation of venous blood volume and underestimation of
OEF [174], [175]. Second, the two-compartment assumption might be too simple for brain
modeling. Therefore, a novel method called quantitative BOLD (qBOLD) [176], [177] has been
proposed to take more compartments into consideration, including intravascular, extravascular

19
white matter, extravascular gray matter, and cerebrospinal fluid (CSF). However, prior knowledge
is needed about the brain tissue composition and various confounding factors may cause bias in
the OEF estimation [52], [176].
1.2.2 Summary and Comparison of Techniques
Three representative MRI techniques of OEF measurement were introduced in the
previous section. The advantages and limitations of each method, as well as the comparisons
between these methods, will be discussed in this section. Furthermore, the status of the
application of these techniques to the SCD population will also be discussed.
In summary, the T2-based TRUST, measures OEF in the SSS with high SNR, providing a
global OEF value, mainly reflecting the oxygen saturation in the cortex. However, TRUST can’t
provide the OEF information in deep structures, and it also can be affected by the blood flow.
The susceptibility-based CISSCO can provide regional OEF information in some specific veins
(usually large and perpendicular to the direction of magnetic field), like the internal cerebral vein
(ICV) or straight sinus (SS). The CISSCO method is not affected by blood flow, nor the partial
volume effect. But the processing procedures are complicated, and high SNR is needed to
guarantee the phase information can be extracted from the blood vessel. QSM is also
susceptibility-based, and it can provide pixel-wise OEF map for both tissue and blood vessels with
arbitrary shape or direction. However, QSM is easy affected by the blood flow, partial volume
effect, as well as streaking artifacts. In addition, OEF measurements by QSM are dependent on
the processing algorithms and steps. The R2’-based ASE technique can calculate tissue OEF maps.
Due to its complicated model, high SNR is required to separate the effect from OEF and venous
cerebral blood volume. R2’ is also influenced by magnetic field inhomogeneities from high water
content, air-tissue interfaces and system flaws. A summary of these techniques is outline in Table
1.1.
Presently, a consensus regarding venous oxygen saturation measurements across various
methods (as outlined in Table 1.1) remains elusive. While most measurements fall within an
acceptable range, only a few have been validated in vivo against established gold standards such
as positron emission tomography (PET) or direct jugular sampling via catheterization. To ensure
the clinical applicability of MRI-based oxygenation measurements, a systematic validation
process is imperative.

20
Table 1.1. Summary of main MRI techniques measuring OEF.
Methods Regions Pros Cons
TRUST SSS High SNR
Straightforward model
Short scan time
Flow effect
Only cortex OEF
Dependence on conversion
model
SBO SSS Straightforward model
Short scan time
High resolution
Background field
correction
Slice selection
QSM Large veins or
tissue
ROI specific
Voxel-wise
Partial volume effect
Flow effect
Streaking artifact
Complicated algorithms
CISSCO Large veins Vessel specific
Not affected by flow or
partial volume
ROI choice
Complicated processing
ASE Tissue Voxel-wise
Short scan time
Low SNR
Confounding factors
1.3 MR validation using Monte Carlo Simulation
1.3.1 Basics of Monte Carlo Simulation
Monte Carlo simulation [178], [179] is a computational methodology employed to
emulate intricate systems or processes by iteratively generating a multitude of random samples
or scenarios. This technique, named after the renowned Monte Carlo Casino, capitalizes on
probabilistic sampling to approximate the behavior of intricate systems that may be analytically
intractable.
In the realm of Monte Carlo simulation, intricate systems are characterized
mathematically through models and equations. Instead of seeking exact analytical solutions, the
approach orchestrates a substantial number of random iterations to estimate pertinent
outcomes, probabilities, or distributions.
The procedure entails the subsequent stages [175], [180]:

21
1. Model Formulation: Articulate a mathematical model encapsulating the essence of the
system or process in question. This encompasses defining variables, parameters,
equations, and their interrelationships.
2. Random Sampling: Stochastically sample input values from pre-defined probability
distributions pertinent to the variables within the model. These samplings are frequently
extracted from recognized distributions like uniform, normal, or exponential distributions.
3. Simulation Iterations: For each ensemble of sampled inputs, execute the model to
compute the targeted output or consequence. This output could encompass singular or
multiple values, an array of values, a probability distribution, or other pertinent metrics.
4. Aggregated Synthesis: Collect outcomes from the entirety of simulation runs and
scrutinize the resultant data. This analysis could encompass the calculation of means,
percentiles, or other summary statistical measures.
5. Interpretation and Inference: Exploit the synthesized outcomes to draw deductions,
formulate predictions, or discern patterns within the system. Monte Carlo simulations
furnish valuable insights into the spectrum of plausible outcomes, the likelihood of
distinct scenarios, and the susceptibility of the model to different input variables.
Monte Carlo simulation finds notable application in the realm of magnetic resonance
imaging (MRI) through its capacity to model and analyze intricate processes associated with MRI
data acquisition [181], image reconstruction [182], and quantitative analysis [183], [184]. This
methodology leverages stochastic sampling to emulate complex interactions between magnetic
fields, tissue properties, and imaging parameters, yielding insights that are challenging to attain
analytically. It has been used widely in signal formation and acquisition modeling, analysis of
noise and artifacts, image reconstruction, and quantitative parameter mapping. In this thesis,
Monte Carlo Simulation is used to investigate the signal generation by the ASE technique and
OEF calculation under different gas challenges.
The general procedures of Monte Carlo simulation in the signal generation of ASE are
[175], [180]:
1. A spherical or cubic simulation environment is generated containing vessels
represented by randomly-oriented cylinders with a fixed radius that span the entire simulation
volume. Both arteries (red) and veins (blue) are indicated in the figure. Capillaries were not
shown in the figure for clarity.

22
Figure 1.13. The Monte Carlo Simulation model. Arteries are indicated by red cylinders
and veins are blue cylinders. (Capillaries are not shown here)
2. The Monte Carlo random walk is performed simulating the diffusion of water protons.
Displacement in each direction (dx, dy, dz) taken by a proton followed the normal distribution
with 0 mean and standard deviation: = √2∆, D is the diffusion coefficient and ∆t is the time
step. Note that the standard devation of the resultant displacement has a magnitude of √6∆.
3. The phase accumulation at each time step was calculated by summing over the field
contributions from all vessels based on: ∆φ= γB(x,y,z)∆t, where the local field is assumed to be
constant over ∆t. The induced magnetic field by each blood vessel was calculated by the equation
∆() =
∆
2

2

2
220, where cylindrical coordinates (r, , ) were used to describe the
position of a proton relative to the blood vessel [180].
4. The signal decay curve was plotted by summing up all the protons with the calculated
phase. Only extravascular signals were simulated because both data in vivo and in silico indicated
that the intravascular signal can be neglected in ASE [175].
Typical simulation parameters were: number of protons N=10000, sphere radius=200 um,
vessel volume fraction=3%, TE=60 ms, τ=0-30 ms, D=10-9 m2
/s.
1.3.2 Realistic Models for SCD
Although ASE has been used in the SCD cohort, to our knowledge, there is no study
investigating the ASE-OEF estimation under hypoxia, a similar setting for the SCD patients. It has
been proved by Monte Carlo Simulation that there might be bias in the ASE-OEF calculation
caused by diffusion effect. Furthermore, there might be other sources of bias lying in the
theoretical model. Real cerebral microvasculature is more complicated mainly because arterial
blood has different oxygen saturation and magnetic susceptibility compared with venous blood.

23
In addition, the complexity of capillary bed also needs to be taken into consideration. Thus, it is
important to use a more realistic model to investigate the ASE-OEF calculation through Monte
Carlo Simulation.
In this thesis, we investigate two physiologically realistic distributions of vessel radii from
sheep and rat brains [185], [186]. Details are illustrated in Chapter 4, and only the key points are
summarized here. The Sharan model is originally formulated for oxygen transport modeling,
considering parameters such as hematocrit, vascular diameter, blood viscosity, blood flow,
metabolic rate, and P50. This compartmental model hasfive orders of arterial and venous vessels,
with a range of radii and relative volume fractions, and the radius of the capillary is fixed.
The Sava model is derived from two-photon spectroscopy estimates of microvasculature
size and oxygen saturation [185] in the mouse. Sava et al separate the perfusing vascular unit
into six segments with different radii. Vessel diameters and oxygen saturation are estimated by
exploiting the differential absorbance of oxygenated and deoxygenated hemoglobin. They show
that oxygenation decreases rapidly downstream along the arteriolar tree and provide
intravascular oxygen saturation as a function of arteriolar and venular diameters.

24
Chapter 2: Anemia Increases Oxygen Extraction Fraction in Deep Brain
Structures but Not in the Cerebral Cortex
2.1 Introduction
Sickle Cell Disease (SCD) is a genetic disorder characterized by a single base pair mutation
in the beta subunit of hemoglobin that causes the abnormal hemoglobin S (HbS) to polymerize
after deoxygenation leading to chronic hemolytic anemia and neurovascular complications [1].
SCD patients have an abnormally high and early risk for stroke [6]. The incidence of primary overt
stroke has been significantly reduced by Transcranial Ultrasound (TCD) screening and chronic
transfusion therapy [26]. However, silent cerebral infarction (SCI) is even more common and
there is lack of established relationship between SCI presence and TCD measurements [187].
Imaging of brain oxygenation could be a powerful tool to assess the risk of stroke and aid in its
prevention. The oxygen extraction fraction (OEF) has been recognized as an accurate and specific
marker for tissue viability and a predictor of misery perfusion in carotid artery disease [188],
[189]. However, compared with other markers such as cerebral blood flow (CBF), studies on the
oxygenation estimation in SCD, especially in deep brain structures, are scarce and results have
been conflicting [90], [122], [171], [190], [191]. The gold standard for oxygen metabolism is
Positron Emission Tomography (PET) imaging [192], [193]. However, PET is limited by its high cost,
invasiveness, long scan time, poor availability, and high exposure to radiation. Therefore,
noninvasive estimates of global and regional brain oxygenation are strongly needed.
Currently, T2-Relaxation-Under-Spin-Tagging (TRUST) is a widely used MRI technique to
quantitatively estimate global brain blood oxygenation via the measurement of pure blood T2
[110], [111]. Unfortunately, TRUST can only provide global saturation for the whole brain without
offering oxygenation information in deep brain structures. Furthermore, there exists uncertainty
in the proper calibration between T2 and oxygen saturation in SCD patients because red blood
cell (RBC) morphology and permeability are deranged in these patients [93]. Unlike T2 oximetry,
susceptibility-based oximetry (SBO) methods are based on the paramagnetic susceptibility of
venous blood. These methods usually measure magnetic susceptibility shift of a vein and there is
a linear relationship between magnetic susceptibility shift of blood and concentration of
deoxyhemoglobin. Quantitative Susceptibility Mapping (QSM) is a widely used technique to
derive a pixel-wise susceptibility map from its induced magnetic field based on the 3D dipole
convolution model [139]. Through multiple image processing steps, QSM allows quantification of
susceptibility for tissue with arbitrary geometry and orientation, which can be used to estimate
oxygen saturation in deep brain structures. An alternative susceptibility-based method called
CISSCO (Complex Image Summation around a Spherical or a Cylindrical Object) was introduced
to quantify the susceptibility of any narrow cylindrical object at any orientation using a typical
multi-echo gradient echo sequence [159]–[161]. The CISSCO method is based on the complex MR

25
signal whereas QSM calculation is based on the phase signal, and they can be both generated
from a typical multi-echo gradient echo scan. Despite the increasing applications of QSM and
CISSCO, neither has been used in patients with chronic anemia and in-vivo validation of these
two techniques remains lacking.
The primary purpose of this study was to compare oxygen utilization in deep cerebral
structures compared to oxygen saturation from the cerebral cortex. To accomplish this, we
performed compared QSM and CISSCO measurements of oxygen saturation in the internal
cerebral vein (ICV) with oxygen values derived from TRUST in the sagittal sinus in healthy subjects
(CTL), sickle cell anemia patients (SCD) and anemia patients with normal hemoglobin (ACTL). The
secondary objective was to cross-validate QSM and CISSCO measurements in the ICV.
2.2 Materials and Methods
2.2.1 Study Design
This study was approved by our Institutional Review Board (CCI#11-00083) at Children’s
Hospital Los Angeles, and all subjects provided written informed consent prior to participation.
Data from 28 SCD patients, 18 ACTL patients and 27 healthy control subjects were acquired.
Complete blood count, metabolic panel, and hemoglobin electrophoresis were measured at the
same day of their MRI scan. Four of the SCD and seven of the ACTL patients were receiving chronic
transfusion therapy; these patients were studied on the morning of their transfusion visit prior
to transfusion. Genotypes for the SCD patients were SS 25, Sb+ 1, and SC 3. ACTL patients
consisted of thalassemia major 6, hemoglobin H constant spring 3, hemoglobin H disease 2,
hereditary spherocytosis 3, Eb thalassemia major 1, aplastic anemia 1, and autoimmune
hemolytic anemia 1. Control subjects were age and ethnicity matched to the SCD population. 8
of these subjects had sickle trait, but prior work from our laboratory has demonstrated
indistinguishable cerebral blood flow and brain oxygenation patterns between hemoglobin AA
and AS subjects (Vu et al., 2021). The exclusion criteria for all subjects included: (1) pregnancy;
(2) hypertension; (3) diabetes; (4) stroke or other known neurologic insult; (5) seizures; (6) known
developmental delay or learning disability; and (7) hospitalization within the month prior to the
study visit.
Images were acquired on a clinical 3 T Philips Achieva system (Philips Healthcare, Best,
Netherlands) with a 32-channel RF coil and a digital receiver chain. The 3D gradient echo
sequence had parameters: TR = 30 ms; α = 25°; 2 echoes: TE1 = 4.94 ms, ΔTE = 5.2 ms; FOV = 210
* 190 * 120 mm3
; spatial resolution: 0.6 * 0.6 * 1.3 mm3
; SENSE acceleration rate = 2 in the phaseencoding direction; BW = 289 Hz/pix and total acquisition time = 6 mins 50 s. Flow-compensation
was added in the readout direction only, which was the anterior-posterior (AP) direction.
T2-Relaxation-under-Spin-Tagging (TRUST) images were acquired from the sagittal sinus
as previously described (Lu and Ge, 2008; Miao et al., 2019). Sequence parameters were as

26
follows: TR = 3000 ms; four effective echoes (eTE) at 0, 40, 80, 160 ms; CPMG τ = 10 m; voxel size
= 3.44 * 3.44 mm2
; FOV = 220 * 220 mm2
; matrix size = 64 * 64; inversion time (TI) = 1022 ms and
total scan time = 1 min 12 s.
2.2.2 QSM processing
For each subject, phase images were fitted to generate a B0 field map. Brain extraction
and phase unwrapping was performed using FSL (v6.0) [194]. Background field was removed
using projection onto dipole fields (PDF) [149]. Unreliable phase voxels were identified using the
condition of spatiotemporal smoothness of the GRE phase data and removed from the brain
mask for subsequent processing [150]. L1-regularized field-to-susceptibility inversion was
performed to derive the susceptibility map and a weighting parameter = 4 × 10−4 was applied
[156]. Venous oxygen saturation (SvO2) was computed based on:
= (1 − 2
)− + − (1)
where is the susceptibility measurement of the internal cerebral vein, − is the
susceptibility shift per unit hematocrit between fully oxygenated and fully deoxygenated
erythrocytes, and − is the susceptibility shift between oxygenated blood cells and water.
Values of 0.27 ppm and -0.03 ppm were used for − and − (Langham et al., 2009; Spees et
al., 2001).
The ROI mask of the internal cerebral vein was manually selected based on the
susceptibility map that was threshold at 0.1 ppm to avoid partial-volume effect. The angle
between ROI and AP axis was calculated manually from the 3D dataset based on the coordinates
of the two end points of the cylinder. Only the segment that had an angle below 30 degrees was
included. The purpose was to exclude regions that were susceptible to flow artifact.
2.2.3 CISSCO processing
A more detailed description of CISSCO method for susceptibility quantification of a
cylindrical object has been presented in [159]. Here we summarized with the major points and
equations. CISSCO integrates the complex MR signals in three annuli around the cylinder of
interest. The complex sums were cast into equations containing three unknown parameters, the
susceptibility and radius of the vessel, and the proton spin density. The overall MR complex signal
S within a coaxial cylinder with radius R was:
= 0 ∫

2
0()

′
/2 +
20,
(2)
where a is the vessel radius, the phase value inside the cylinder = −
∆
6
(32 −
1)0 , ∆ is the susceptibility difference between tissues inside and outside, is the slice
thickness of the image, 0 and 0, are the effective spin densities of the tissue outside and inside
the object, is the effective magnetic moment,
′ = (0.50∆) ∗ 2 is the extremum

27
phase value on the surface of the cylinder, is the orientation of the cylinder, and J0 is the zeroth
order Bessel function.
Equation 2 can be applied to all three annuli, allowing us to solve for the three unknown
variables; complex signal differences between any two annuli eliminate the second term,
eliminating two variables. First, the magnitude and phase images in coronal view were cropped
to 64*64 and a 16*16 Gaussian high pass filter was applied to remove background phase. Next,
was estimated based on the coordinates of the two endpoints of the innermost annuli. The
calculated θ was 82.3±5.6 degrees, revealing that the internal cerebral vein is nearly
perpendicular to the direction of B0. Finally, after applying the equation to three coaxial cylinders,
the effective magnetic moment, the effective spin densities, and the susceptibility difference can
be solved sequentially.
The resulting Dc was converted to oxygen saturation using equation 1, the same as for
QSM.
2.2.4 TRUST processing
Control–label difference images for each echo time were averaged and fit to a simple
mono-exponential function. In control and ACTL patients, the decay time constant was corrected
for T1 using an estimated calculated from hematocrit, assuming deoxygenated blood [195]. In
non-transfused SCD patients, venous T1 was estimated to be 1818 ms (Václavuu et al., 2016), and
for transfused SCD patients T1 was estimated using a simple mixture assumption based upon the
fraction of circulating hemoglobin S. In control subjects, the resulting T2-apparent was converted
to oxygen saturation using a calibration derived from human blood [197], [198]. In SCD patients,
a consensus calibration model [93] was used to convert T2-apparent to oxygen saturation.
Separate equations were used for transfused and non-transfused subjects, taking care to correct
T2-apparent for imperfections in the 180 degree pulse [93].
2.2.5 Physiological background
To gain physiological insight into predictors of oxygen saturation in the ICV compared with
the sagittal sinus, oxygen extraction fraction (OEF) was calculated separately for the two venous
locations (OEFICV, OEFSS) as follows:
= (2 − 2)/2 (3)
where 2 is the arterial saturation measured by pulse oximetry. We compared OEFICV
and OEFSS to O2 content using linear regression, with variable transformation when appropriate.
O2 content was derived as follows, neglecting the impact of dyshemoglobins [199]:
2 = 1.34 ∗ ∗ 2 (4)
To provide some physiological background, the equation between O2 content, cerebral
blood flow (CBF) and cerebral metabolic rate of oxygen (CMRO2) is also shown here:

28
2 = ∗ ∗ 2 (5)
Alternatively, equation (5) can be recast as follows:
= 2/( ∗ 2 ) (6)
Thus, OEF is expected to vary reciprocally with the product of CBF and O2 content, which
is also referred to as cerebral oxygen delivery.
2.2.6 Statistical analysis
Statistical analysis was performed in JMP (SAS, Cary, NC). Demographic and laboratory
variables were compared using Analysis of Variance (ANOVA) with Dunnett’s post hoc correction.
OEF values derived by QSM and CISSCO (OEF-QSM, OEF-CISSCO) were compared across study
groups using ANOVA with Dunnett’s post-hoc correction. Inter-modality comparison was
performed using Bland-Altman analysis, with bias assessed using a two-sided, one-sample T-Test.
Shapiro-Wilks tests of normality were applied to each variable, with transformation, outlier
exclusion, or nonparametric testing used when appropriate.
We examined predictors of OEFICV and OEFSS using linear regression, with variable
transformation when appropriate. Predictors included hemoglobin, hematocrit, oxygen
saturation, oxygen content, hemoglobin S%, left shifted hemoglobin %, LDH, reticulocyte count,
cell free hemoglobin, WBC, MCV, MCH, MCHC, WBC, platelets, and mean platelet volume. All
variables with p values greater than 0.05 were retained for stepwise regression. Models were
built iteratively (two variable models, followed by three variable models, etc), retaining variables
yielding the highest combined r2
. No nonlinear variable interactions were considered.
Given the collinearity between the three strongest predictors (hemoglobin, hematocrit,
and oxygen content), we also explored models where one of these three variables was “locked”
in the model to inform us about potentially important covariates.
2.3 Results
2.3.1 Demographics
Among the 73 volunteers participated in the experiment, data from 7 subjects were
discarded due to motion or low SNR. There were 25 SCD patients, 17 ACTL patients and 24 healthy
controls in the final data processing, and the demographics were shown in Table 2.1. Controls
were slightly older than either patient group, but the groups were well balanced for sex. Anemia,
corresponding erythropoietic and hemolytic markers, hemoglobin F and total left shifted levels
were increased in ACTL and SCD, but more severe in SCD. Oxygen saturation was not different
across groups, but diastolic blood pressure was 10% lower in both anemic groups. Forty percent
of the ACTL patients were transfused, compared with 16% of SCD, and none of the control
subjects.

29
Table 2.1. Subject demographics. Group averages and standard deviations are given.
Healthy controls ACTL patients SCD patients
N 24 17 25
Age 27.3±8.2 21.6±5.6 24.1±6.8
Sex 11F, 13M 10F, 7M 11F, 14M
Hematocrit (%) 40.8±4.1 35.2±5.9* 26.8±4.3*
Hemoglobin (g/dL) 13.6±1.4 11.5±2.8* 9.5±1.7*
HbS (%) 12.4±17.5 2.5±10.2 67.8±25.2*
HbF (%) 0 4.5±6.0* 12.1±10.5*
Systolic blood pressure (mmHg) 114.7±9.0 115.2±11.5 114.0±10.0
Diastolic blood pressure (mmHg) 69.0±7.5 62.1±8.1* 61±6.3*
Oxygen Saturation (%) 99.1±1.2 99.2±1.0 98.6±1.4
O2 content (ml O2/ml blood) 18.5±1.9 15.6±3.8* 12.9±2.4*
Lactose dehydrogenase (LDH) 548.2±86.4 635.2±250.0 863.0±432.2*

MCV (cubic um/ red cell) 85.5±7.4 79.5±9.7 96.1±16.6*

MCH (pg/ cell) 28.8±2.6 25.8±5.3* 34.0±6.7*

MCHC (g Hgb/ dl) 33.4±1.3 31.7±4.1 35.2±1.6*

Reticulocytes (%) 1.5±0.7 3.6±3.4* 8.4±4.9*
Transfused 0/24 7/17* 4/25
* Denotes p < 0.05 with respect to control, denotes p < 0.05 with respect to ACTL
Table 2.2 summarizes the Hb, HbS%, HbF%, and OEF values in non-transfused SCD
patients with and without hydroxyurea. Patients taking hydroxyurea had higher F%, higher OEF
CISSCO, lower S%, and lower sagittal sinus OEF measurements than patients not on hydroxyurea.
None of the controls or ACTL patients were taking hydroxyurea.

30
Table 2.2 Differences between hemoglobin F, hemoglobin S and OEF by three methods with or
without hydroxyurea (HU) for non-transfused SCD patients.
with HU without HU p-value
Hb 9.5 ± 1.4 12.1 ± 2.7 < 0.001
Hb F (%) 18.4 ± 8.8 1.4 ± 3.3 < 0.001
Hb S (%) 78.2 ± 8.3 18.2 ± 27.9 < 0.001
OEF-TRUST (%) 28.2 ± 6.1 35.3 ± 5.7 < 0.001
OEF-CISSCO (%) 42.9 ± 5.5 36.2 ± 5.2 < 0.001
OEF-QSM (%) 30.1 ± 4.4 28.0 ± 2.7 0.1034
Figure 2.1A,B,C show representative magnitude and phase images in both axial and coronal views.
The processed susceptibility map by QSM is shown in Figure 2.1D. ICV generally lies parallel to
the axial plane and typically has a quantifiable length around 11 mm by QSM. By CISSCO, the
calculated vessel radius, a, was 1.1±0.5 mm and was independent of disease state and
hemoglobin level.

31
Figure 2.1. Representative image and region of the internal cerebral vein (ICV, highlight by red
rectangle). (A) Magnitude in axial view. (B) Magnitude in coronal view. (C) Axial view, and (D)
Sagittal view of ROI in the max intensity projection of a representative susceptibility map.
2.3.2 Comparison of OEF measurements in ICV (by QSM and CISSCO) and SS (by TRUST) for
different groups
Figure 2.2A,B summarize the OEF measurements in the internal cerebral vein using QSM
(OEF-QSM) and CISSCO (OEF-CISSCO), respectively. Mean OEF-QSM measurements were 30.1%
in SCD, 28.3% in ACTL, and 26.6% in CTL (p < 0.01). On Dunnett’s post hoc correction, SCD was
different from CTL (p < 0.001), but ACTL was not (p = 0.1069). Mean OEF-CISSCO measurements
were 42.5% in SCD, 37.0% in ACTL, and 33.0% in CTL (p < 0.01), with both SCD (p < 0.001) and
ACTL (p = 0.007) significantly different from control subjects. Figure 2.2C shows the OEF
measurements in the sagittal sinus vein using TRUST (OEF-TRUST). Mean OEF-TRUST
measurements were 31.2% in SCD, 33.4% in ACTL and 36.8% in CTL (p < 0.01). After Dunnett’s
analysis, we found that SCD was different from CTL (p = 0.0034) and ACTL was not (p = 0.1291).

32
Figure 2.2. Group differences for OEF measurements by CISSCO, QSM and TRUST. (A) Boxplot
for OEF-QSM in internal cerebral vein for SCD, ACTL and CTL. (B) Boxplot for OEF-CISSCO in
internal cerebral vein for SCD, ACTL and CTL. (C) Boxplot for OEF-TRUST in sagittal sinus for SCD,
ACTL and CTL. (* denoted statistically significant p < 0.05; NS denoted no significant difference)
Figure 2.3 characterizes the bias between the two OEFICV measurements using linear
correlation (Figure 3A) and Bland-Altman analysis (Figure 2.3B). A strong linear relationship was
observed between OEF-QSM and OEF-CISSCO (r2 = 0.72, p < 0.001). The mean OEF-CISSCO is 9.3%
higher than OEF-QSM (p < 0.001). In addition, the bias was proportional to the mean value (r2 =
0.61, p < 0.001) and was larger in anemic subjects.
Figure 2.3. (A) Scatter plot of OEF-CISSCO and OEF-QSM with linear correlation line (solid) and
identity line (dashed) (r2 = 0.72, p < 0.001). (B) Bland-Altman plot for OEF-CISSCO and OEF-QSM.
The linear correlation line (solid) is shown (r2 = 0.61, p < 0.001).

33
2.3.3 Relationships between OEF measurements and O2 content in ICV
The relationship between OEF measurements in the internal cerebral vein and O2 content
is shown in Figure 2.4A,B. Both methods demonstrate a reciprocal relationship with O2 content,
but the variance is significantly less for OEF-CISSCO compared with OEF-QSM (p < 0.01 by F test).
Importantly, when this relationship is removed, all group differences in OEF disappear for both
techniques.
Figure 2.4. Relationship between OEF and O2 content in SCD, ACTL and CTL. (A) Scatterplot
between OEF-CISSCO and O2 content. The fitting reciprocal line is shown in black with r
2 = 0.86,
p < 0.001. (B) Scatterplot between OEF-QSM and O2 content. The fitting reciprocal line is shown
in black with r
2 = 0.48, p < 0.001.
2.3.4 Relationships between OEF, left shifted hemoglobin and O2 content in SS
Figure 2.5A demonstrates OEF-TRUST in the sagittal sinus as a function of left shifted
hemoglobin concentration. The left shifted hemoglobin included hemoglobin F in SCD patients
and fast moving hemoglobin in ACTL patients with alpha-thalassemia. It revealed that OEF-TRUST
in the sagittal sinus declined with increasing left shifted hemoglobin for both SCD (r2 = 0.35, p =
0.018) and ACTL (r2 = 0.46, p < 0.0027). The slope was statistically identical in SCD and ACTL, as
were the intercepts for all three groups. Thus, after controlling for inter-subject variability in the
left shifted hemoglobin, the corrected OEF in the sagittal sinus was independent of group and O2
content, as shown in Figure 2.5B.

34
Figure 2.5. (A) Relationship between OEF-TRUST with left shifted hemoglobin. Linear
correlations are shown in blue line (r2 = 0.46, p = 0.0027) for ACTL and red line (r2 = 0.35, p =
0.0018) for SCD. The control group is shown as 36.8 ± 5.5 (mean ± std) in green. (B) Relationship
between corrected OEF-TRUST with O2 content.
2.3.5 Predictors for OEF in ICV and SS
Table 2.3 summarizes the primary and secondary predictors for OEF in the ICV and sagittal
sinus. For OEF-CISSCO, there were two equivalent models predicting OEF in the ICV with a
combined r2 of 0.94 and 0.95 respectively. The dominant variable was either oxygen delivery (r2
= 0.86), hemoglobin (r2 = 0.85) or hematocrit (r2 = 0.92), which are intrinsically co-linear (r2 =
0.90). To explore what hematocrit reflected, beyond oxygen transport, we locked oxygen content
(or hemoglobin alone) into the model, displacing hematocrit but introducing mean corpuscular
hemoglobin concentration (MCHC) into the model. If hematocrit entered the model first,
hemoglobin, oxygen content, and MCHC were displaced. With either the primary or the
alternative model, oxygen saturation by pulse oximetry was positively associated with OEF-ICV.

35
Table 2.3. OEF predictors by stepwise regression.
Residual OEF-CISSCO and O2 content Residual OEF-TRUST and HbF%
Parameter r
2 p-value Parameter r
2 p-value
MCHC 0.34 <0.0001 MCH 0.16 0.008
Oxygen Saturation 0.19 0.0003 MCV 0.10 0.03
MCH 0.17 0.001 Height 0.10 0.03
HbS% 0.16 0.009 Weight 0.09 0.05
Weight 0.08 0.03
RBC 0.07 0.04
MCV 0.07 0.04
Final Stepwise Model OEF-CISSCO Final Stepwise Model OEF-TRUST
1/O2 content 0.85 <0.0001 HbF 0.34 0.0001
MCHC +0.05 <0.0001 MCV +0.13 0.004
Oxygen Saturation +0.04 <0.0001
Total r2 0.94 Total r2 0.53
Alternative Model OEF-CISSCO
1/Hematocrit 0.92 <0.0001
Oxygen Saturation +0.03 <0.0001
Total r2 0.95
After controlling OEF-TRUST for left shifted hemoglobin concentration, there was no
residual relationship with patient group, Hb, Hct, RBC or O2 content. There were also weak
associations with systolic blood pressure (SBP), and the collinear variables MCV and MCH (r2 =
0.92 with respect to each other). On stepwise analysis, left shifted hemoglobin, SBP and MCV
persisted with a combined r2 of 0.47.
2.4 Discussion
In this manuscript, we studied 66 subjects across a broad range of hemoglobin values and
identified an increase in OEF in deep brain structures in patients with chronic anemia by two
susceptibility-based methods. OEF in the ICV was reciprocally related to hematocrit, which
reflected a combination of oxygen content and mean cellular hemoglobin concentration. OEF in
the ICV was also directly proportional to peripheral oxygen saturation. In contrast, sagittal sinus
OEF was decreased in anemic subjects, proportionally to hemoglobin, hematocrit, and oxygen
content. However, after controlling for the impact of left shifted hemoglobin (hemoglobin F,
hemoglobin H, hemoglobin Barts), OEF was independent of patient group and oxygen content.
OEF by QSM and CISSCO were highly correlated, but QSM yielded systematically lower OEF
estimates.

36
Previous work from our laboratory, and others, has suggested disparate oxygen
extraction fraction estimates measured using sagittal sinus oximetry when compared to tissue
oximetry performed in deep brain structures [90], [197], [200]. In particular, tissue oximetry
suggests profound deep brain hypoxia with worsening anemia [90], [171], while sagittal sinus
oximetry suggests normal or even decreased oxygen extraction [197], [198]. We have previously
postulated that a vascular steal phenomenon may exist, where oxygen delivery to brain cortex is
preserved at the expense of deep brain structures [86]. Arterial transit time is decreased in sickle
cell disease patients [201], leading to decreased exchange of labeled spins into the
cerebrovasculature [191], [202]. However, this manuscript is the first to document deep brain
hypoxia in anemia subjects with concomitant preservation of sagittal sinus oxygen extraction.
The internal cerebral vein is one of the major deep cortical veins, and its oxygen saturation
provides insight into the oxygenation of the basal ganglia, corpus callosum and thalamus. OEF
estimates using either QSM or CISSCO demonstrated an inverse relationship with O2 content
(Figure 4), similar to whole brain estimates of OEF by Asymmetric Spin Echo (ASE) (Supplemental
Figure S1) [90]. Although there is a bias between the two datasets, both demonstrate a
comparable reciprocal relationship with O2 content. The similarity should not be surprising,
however, as ASE is dominated by tissue oxygenation in the white matter and deep gray structures;
in ASE, much of the signal from cortex is contaminated by susceptibility artifacts from superficial
veins and excluded from global OEF measurements. ASE oximetry measurements yield spatially
averaged tissue oxygenation that weigh grey matter and white matter equally, despite their
markedly disparate contribution to brain metabolism.
According to equation (6), OEF may be cast as the ratio of the cerebral metabolic rate of
oxygen divided by the product of cerebral blood flow and oxygen content (i.e. O2 delivery). On a
whole-brain basis, CBF increases reciprocally with O2 content in anemic subjects [68], [69],
preserving global oxygen delivery. However, compensatory hyperemia is blunted or eliminated
in deep watershed structures [86]. By equation (6), OEF in the ICV (and by ASE) should vary
inversely with oxygen content if deep brain blood flow does not augment appropriately.
The other two predictors of deep brain OEF also provide physiological insights. The
positive relationship between ICV OEF and SpO2 arises organically from equation (3). The brain
operates within a very narrow range of pO2, which translates to a much broader range of SvO2
because of inter-subject variability in the hemoglobin dissociation curve [203]. Powerful
physiological compensation mechanisms limit declines in SvO2 under hypoxic conditions. Thus,
inter-subject variability in SaO2, even within the normal range, introduces a positive relationship
with OEF by providing more “headroom” for oxygen extraction. The positive association of MCHC
and OEF in the ICV is more challenging to explain. MCHC contributes powerfully to RBC
deformability and viscosity. We speculate that MCHC could be modulating capillary transit time
through its impact on red cell rheology, however this would have to be independently confirmed.
In the sagittal sinus, oxygen saturation is dominated by cortical blood flow and supplydemand matching. Venous oximetry techniques follow Fick-principle (oxygen mass balance) and

37
are inherently flow-weighted rather than spatially weighted. Since grey matter has 3-4 times the
metabolic activity of white matter, venous oximetry techniques reflect grey matter oxygen
balance. Whole brain CBF rises inversely to oxygen carrying capacity such that oxygen delivery is
preserved in anemic subjects [68], [69]. There are also multiple other publications confirming
these observations [84], [204]. Regional flow assessment using ASL demonstrates that cortical
oxygen delivery is normal [86]. Preservation of OEF in the sagittal sinus, with normal or even
decreased OEF despite worsening anemia, is the natural consequence of compensatory
hyperemia[197]. The mechanisms behind this “cortical sparing” are unknown but work in mice
suggests that chronic hypoxia stimulates cortical capillary proliferation [205]. Duffin modeled
compensatory hyperemia using a “fail-safe” mechanism and proposed potential biochemical
mediators [206]. Regardless of the mechanism, observed patterns of brain volume loss [88], [89]
and silent infarction [90] are consistent with cortical sparing at the expense of the deep
watershed areas.
A second striking finding of this study was the powerful effect of hemoglobin F and other
high affinity hemoglobin molecules on OEF measured in the sagittal sinus. This undoubtedly
represents left-shift of the hemoglobin dissociation curve and resulting higher oxygen affinity of
hemoglobin. While this phenomenon is well known, the magnitude of the effect is striking, with
OEF decreasing 20 saturation points over the physiologic range of left shifted hemoglobin fraction
in anemic subjects. Importantly, all group differences in sagittal sinus OEF were eliminated once
differences in high affinity hemoglobin fraction was controlled for. These findings highlight the
critical physiologic principle that cerebral OEF is not a regulated variable. The brain varies
vascular tone to preserve capillary pO2 in a narrow range [203]. When hemoglobin is left-shifted,
less oxygen is delivered for any brain pO2. In SCD patients with high hemoglobin F concentration,
resting CBF increases to preserve tissue oxygen unloading in the cortex [201]. In contrast, the
deep structures have decreased compensatory hyperemia and are forced to operate at lower
pO2 (increased OEF compounded by tighter oxygen affinity).
So are hydroxyurea or other dissociation curve modifiers placing the deep structures of
the brain at risk? The left-shift of hemoglobin cannot be interpreted in isolation. With
hydroxyurea, the impaired oxygen unloading is at least partially compensated by a 10%-15% rise
in O2 content; to date no study has systematically evaluated regional oxygen delivery and
metabolism in patients prior to and following hydroxyurea initiation to evaluate that balance
between improved oxygen capacity and impaired oxygen unloading. Studies are currently
ongoing to explore these endpoints in voxelotor [39], [40], an allosteric modifier of hemoglobin
affinity. The present study emphasizes the need to examine both global and regional responses
to such therapies.
The powerful impact of HbF and other modulators of the hemoglobin dissociation curve
must also be considered when comparing OEF from modern hemoglobinopathy cohorts, with
historical data prior to the widespread use of hydroxyurea [54], [207]. TRUST studies from
modern SCD cohorts [54], [122], [201] consistently observed lower OEF values that reported in

38
historical cohorts [53], [207]. Hydroxyurea did not achieve widespread use until the last one to
two decades. Current recommendations favor introduction at 18 months of age, and dose
escalation to maximum tolerated dose, leading to robust hemoglobin F% induction in many
patients. In our cohort, OEF was 7 percentage points lower in patients taking hydroxyurea which
is consistent with OEF differences exhibited by current and historical cohorts.
It is important to consider whether our two principle findings, i.e. cortical sparing and the
powerful effect of left shifted hemoglobin fraction, could be artifacts resulting from the two
different oximetry techniques. MRI venous oximetry can be performed using magnetic
susceptibility or by R2, R2*, or R2’ relaxometries. If the calibration curves for these methods are
unbiased across the patient subgroups, then the techniques can be used interchangeably, and
our observed results are valid. The susceptibility calibration is considered the most robust
because it is independent of red cell integrity. The linear dependence of susceptibility on
hematocrit and oxygen saturation is incontrovertible but estimates of the intrinsic susceptibility
of deoxygenated hemoglobin vary from 0.18 – 0.27 [90], [126], [127], [132], [208]. Errors in this
parameter could introduce bias but not alter the direction of change observed. Hemoglobin S has
the same intrinsic susceptibility as hemoglobin A (Eldeniz et al., 2017) and there is little
biophysical reason to believe that other hemoglobins should be significantly different because
the electron shells of the heme moiety are identical (it’s just the supporting scaffolding that’s
different).
The TRUST calibration in normal subjects (so-called hemoglobin AA calibration) is also
fairly incontrovertible. Two independent laboratories have yielded superimposable results over
a very broad range of hematocrit values [93], [197], [198]. This observation is important because
the red blood cells in the ACTL group, and all transfused patients, predominantly contains normal
hemoglobin. Results from these patients yield identical findings compared to the SCD patients.
Thus, cortical sparing and the effect of left shifted hemoglobin cannot be attributed simply to a
faulty oximetry calibration.
The TRUST calibration has challenges in the SCD population that we have characterizing
for years[93], [197], [198]. The variation in calibration arises from damage to the red cell
membrane as well as changes in red cell density and shape [93], [197], [198], not from any
intrinsic magnetic difference in sickle hemoglobin (Eldeniz et al., 2017). Since patients with sickle
cell trait have normal appearing red blood cells, their blood follows the AA calibration curve, not
the SCD curve(Bush et al., 2018). The sickle calibration curve used in this study represents a
“consensus” calibration using data pooled from two laboratories to yield a more stable estimate
of the hematocrit interaction and to derive a model that accounts for the dilution effect of
transfusion (Bush et al., 2021). We believe that the absence of a group effect in Figures 4 and 5,
despite combining controls, transfused SCD, non-transfused SCD, transfused ACTL, and nontransfused ACTL, offers strong evidence that our TRUST calibrations not introducing systematic
bias that could be misinterpreted as physiologic change.

39
It is also reasonable to wonder about the potential impact of hemoglobin F on the T2
calibration in those patients with excellent response to hydroxyurea (e.g. >15%). The published
calibration curve for hemoglobin F cells is quite similar to the HbA calibration [122], [209, p. 2].
In SCD patients, hemoglobin F upregulation will cause some degree of hemoglobin F cells to
circulate (like hemoglobin A)…this we could potentially compensate for using the mixture model
(SS + FF is similar to SS + AA). To determine whether the increased HbF in some of the nontransfused SCD has an important “dilutional” effect, similar to transfusions, we reran all the
TRUST data using the mixture model and did not observe any significant qualitative differences
to our findings (see supplemental Figure S2A, B).
In the extremes of hemoglobin F expression (for example after gene therapy or hereditary
persistence of fetal hemoglobin), the sickling process is almost completely abrogated, and the
resulting red cells are morphologically normal. Under this extreme case (which does NOT reflect
even the highest F induction in this study), the hemoglobin A calibration would be appropriate.
As a result, we also compared the results using the hemoglobin A calibration for all subjects as a
“worst case” simulation. When we did this, the two principal findings (cortical sparing, OEF
negatively correlated with hemoglobin F) were maintained (supplemental Figure S2C, D).
However, OEF was systematically higher in SCD patients. This is nonsensical because it would
imply that cerebral metabolic rate is increased in SCD compared with controls and other anemic
subjects. Studies using gold standard techniques like Kety-Schmidt and PET have proven exactly
the opposite (see supplemental data in Vu et al., 2021).
Both the CISSCO and QSM methods reveal group differences in venous saturation
measurements and similar relationships with hemoglobin. However, there exists significant bias
between the two techniques with QSM exhibiting higher saturation values. There might be
several reasons for the underestimation of susceptibility for QSM. Firstly, the existence of
streaking artifact may affect the measurement of susceptibility indirectly [210]. Secondly, partialvolume effect causes susceptibility to be underestimated, particularly in small vessels like the
internal cerebral vein [147]. Lastly, QSM estimates of vein saturation are impacted by nonlinear
phase accrual in moving spins. This effect could worsen with anemia severity as blood viscosity
lessens and blood flow increases. We postulate that the latter two effects are responsible for the
systemic bias between QSM and CISSCO oximetry in these patients. CISSCO overcomes these
limitations by avoiding unstable dipole inversion and by inferring vessel susceptibility from its
effect on surrounding tissue rather than from the blood itself, similar to asymmetric spin echo
(ASE).
The CISSCO method is not without its own limitations. CISSCO is most accurate for veins
perpendicular to the main magnetic field, making it most effective for the straight sinus and
internal cerebral vein. It cannot be used for superficial veins like the sagittal sinus because a
complete annulus cannot be drawn across it. CISSCO requires high SNR inside the vessel (over
5:1) for accurate measurements, which limits the choice of echo times and can be impacted by
Gibbs ringing [159]. In practice, we first quantify the magnetic moment from the second echo

40
time as the uncertainty of measurement of magnetic moment at the longer echo time is smaller
[161]. Then we solve the susceptibility at the first echo time for a higher SNR inside the internal
cerebral vein, after scaling the calculated magnetic moment. In addition, improper choice of
radius of three annuli might produce up to 15% uncertainty to the calculated susceptibility
through error propagation [159]. However, despite these limitations, the tight relationship of
OEFICV with O2 content and the similarity of this relationship with OEF by ASE suggest that it may
be a better choice than QSM for venous oximetry in select vessels.
Other alternatives exist to quantify saturation in the deep veins. A variation of TRUST,
called TRU-PC [211], uses complex differencing rather than arterial spin labeling to eliminate
partial volume effects and has sufficient signal to quantify T2 in deep draining veins. It would be
instructive, in the future, to compare OEF estimates by TRU-PC and CISSCO in SCD patients as a
mechanism to cross-validate T2 and susceptibility-based oximetry in this patient population.
2.5 Conclusion
In summary, we have demonstrated that anemia worsens hypoxia in deep brain
structures while cerebral cortex appears to be spared, replicating observations using tissuebased oximetry ASE. Our findings resolve the apparent paradox observed between TRUST and
ASE measurements. We also demonstrate that hemoglobin F and other high affinity hemoglobin
molecules appear to be powerful modulators of OEF by TRUST and explain group differences in
OEF observed between anemic and control subjects. These observations were conserved across
diverse study populations, robust with respect to choice of sickle cell calibration, and
physiologically plausible.
Although CISSCO was ideal for the present context because is robust to partial volume
effects, independent of blood flow velocity, and unaffected by red cell properties, it is only
suitable for select cerebral veins, limiting its generalizability. QSM is more generalizable and
saturation estimates were qualitatively similar to CISSCO, but troubled by large biases from
partial volume and flow effects. Future oximetry in the deep brain may require techniques such
as TRU-PC.

41
2.6 Supplemental Figures
Figure 2.6. Relationship between OEF and O2 content by CISSCO (OEF-CISSCO) and reference
data (OEF-ASE) [90].

42
Figure 2.7. (A) Relationship between OEF-TRUST with left shifted hemoglobin using the mixture
model. Linear correlations are shown in blue line (r2 = 0.46, p = 0.0027) for ACTL and red line (r2
= 0.27, p = 0.008) for SCD. The control group is shown as 36.8 ± 5.5 (mean ± std) in green. (B)
Relationship between corrected OEF-TRUST with O2 content using the mixture model. Figure
S2. (C) Relationship between OEF-TRUST with left shifted hemoglobin using the HbA model.
Linear correlations are shown in blue line (r2 = 0.46, p = 0.0027) for ACTL and red line (r2 = 0.15,
p = 0.0595) for SCD. The control group is shown as 36.8 ± 5.5 (mean ± std) in green. (D)
Relationship between corrected OEF-TRUST with O2 content using the HbA model.

43
Chapter 3: Oxygen extraction fraction measurements using
Asymmetric Spin Echo during hypoxic and hypercapnic gas challenges
3.1 Introduction
Cerebral Oxygen Extraction Fraction (OEF) has been recognized as an important
physiological parameter of the brain’s metabolism, and it has been suggested to be an accurate
and specific biomarker in various diseases, including Alzheimer’s disease [98], [212], sickle cell
disease [197] and carotid artery diseases [189].
Magnetic Resonance Imaging (MRI) can provide non-invasive quantification of cerebral
OEF based on the T2 of blood, susceptibility of blood or tissue, or R2’ of tissue. T2-Relaxationunder-Spin-Tagging (TRUST) [110], [111] is a widely used T2-based MRI technique to
quantitatively estimate global brain blood oxygenation by measuring blood T2. Acquisition and
processing of TRUST data are easy and straightforward, and the results have shown to be in
excellent agreement with PET [116], [117]. TRUST measurements demonstrate normal [198] or
decreased OEF [197], [213], [214] in anemic subjects, whether suffering from sickle cell disease
or other rare anemias. However, OEF measurements by asymmetric spin echo (ASE) indicate the
opposite trend, namely increased OEF for lower hematocrit values[171]. Although both
techniques reflect OEF, the physical principles behind them are completely different. TRUST
represents a volumetric weighting of oxygen transport from the upstream vascular bed, governed
by the Fick Principle, and can be used to estimate cerebral metabolic rate [54]. Since oxygen
utilization is 3-4 times higher in grey matter compared to white matter, sagittal sinus TRUST
measurements are dominated by oxygen-supply demand relationships to the cerebral cortex.
In contrast, ASE estimates tissue-based OEF from water molecules diffusing near cerebral
microvasculature [165], [166], [176]. These measurements represent a spatial integration of
oxygenated and deoxygenated hemoglobin, independent of the overall tissue flow, analogous to
brain oxygen metrics derived using near-infrared spectrometry (NIRS) [215], [216].
Microvasculature in white matter and grey matter contribute equally to spatial OEF
measurements, despite their vastly different oxygen consumptions. Thus, ASE and NIRS
measurements cannot be used to infer cerebral metabolic rate.
Flow (TRUST) and spatially (ASE) weighted OEF measurements should generally agree
with one another when there is homogeneous oxygen extraction across the cerebral vasculature.
However, in conditions where there is frank vascular shunting or capillary transit time
heterogeneity [185], [217], differences could appear between these methods. Hypoxia and
hypercarbia represent physiologically-relevant challenges that alter total flow as well as
microvascular blood flow distribution [218]. Using both ASE [165] and TRUST [119], [136], OEF
has been shown to decrease with hypercapnia. Under hypoxia, OEF has been reported to be

44
unchanged [120], [219] using TRUST, depending on the severity of the stimulus. Unfortunately,
these studies have not controlled for end-tidal CO2, which decreases with hyperventilation,
causing cerebral vasoconstriction. Even less is known about the impact of hypoxia on ASE, with
no known published accounts of the response to isocapnic hypoxia.
The primary goal of this study was to investigate the bias between ASE and TRUST in
measuring OEF during hypercapnic and hypoxic challenges in healthy subjects. In addition, we
applied a Monte Carlo Simulation of the ASE signal [175] to model the OEF changes by ASE during
these gas changes to gain insights into any potential disparities between the two methods.
3.2 Methods
3.2.1 Study Design
This study with healthy volunteers was approved by the Institutional Review Board at
Children’s Hospital Los Angeles (CHLA); written informed consent and/or assent were obtained
from all subjects (CII#12-00338). This study was performed in accordance with the Declaration of
Helsinki. Exclusion criteria were contraindications for MRI or respiratory challenges.
The gas challenges were performed using a computer-controlled gas blender called
RespirAct (Thornhill Research, Toronto, Canada). RespirAct can prospectively target and
independently manipulate end-tidal O2 (EtO2) and end-tidal CO2 (EtCO2) partial pressures [220].
The scan protocol included five respiratory phases: baseline, mild hypoxia, severe hypoxia,
hypercapnia, and recovery (Figure 3.1). During baseline and recovery, all participants inspired
room air. During mild and severe hypoxia phases, EtO2 was targeted at 55 mmHg and 40 mmHg
respectively. During the two transitions, slightly lower EtO2 (40 mmHg and 35 mmHg respectively)
were applied for the first 60 seconds to accelerate the process to steady state oxygenation.
During hypercapnia, EtCO2 was increased by 10 mmHg compared to baseline. The baseline and
recovery phase lasted for 2 minutes each and the other three phases lasted for 12 minutes each.
Every subject held a signaling beacon allowing them to terminate the scan if they felt
uncomfortable.
Figure 3.1. Representative end tidal O2 (EtO2) and CO2 (EtCO2) for one subject. Five
phases are shown in the figure: baseline, hypoxia1, hypoxia2, hypercapnia, and recovery.

45
The phase contrast measurements, consisting of six dynamics of 30 second duration, were
initiated concurrently with the gas transitions. This captured the flow dynamics and provided
sufficient time to reach steady-state flow. During the steady state of each phase, TRUST and ASE
were acquired following the same order. Fingertip pulse oximetry (SpO2) was continuously
recorded. Cerebral tissue oxygenation index (TOI) was measured by NIRO 200 (Hamamatsu
Photonics, Japan) with a probe placed on the forehead.
3.2.2 Image Acquisition:
A total of 12 healthy subjects (4 females, age 36±10 years) were studied. All subjects underwent
an MR study using Philips 3T Achieva with a 32-element head coil. For each subject, ASE (Figure
3.2) was acquired with the following parameters: TR = 3000 ms, TE = 62 ms, resolution = 3 × 3 ×
6 mm3
, matrix size = 64 × 64, τ range = 10: 0.5: 20 ms. Diffusion gradients with b value = 100
s/mm2
in three directions were set to suppress the intravascular signal. Each scan was composed
of 28 dynamics, including 7 spin-echo scans (τ = 0) and the other 21 equally spaced τ values from
10ms to 20ms for each echo. The total scan time for the ASE sequence was about 3 minutes
during each phase.
Figure 3.2. Representative ASE signal decay. The fitting line for the last 21 echoes are shown by
the dashed line. And the spin echo value (mean of the first 7 echoes) is shown by the red dot.

46
The slope of the fitted line is R2’, and the difference between spin echo value and the intercept
of the fitted line is . Three figures of magnitude at different echoes are also shown (at echo1,
echo 9, and echo 27).
TRUST images were acquired from the sagittal sinus as previously described [111], [138], [221].
Sequence parameters were as follows: TR = 3000 ms; four effective echoes (eTE) at 0, 40, 80, 160
ms; τCPMG = 10 ms; voxel size = 3.44 × 3.44 mm2
; FOV = 220 × 220 mm2
; matrix size = 64 × 64;
inversion time (TI) = 1022 ms and total scan time = 1 min 12 s.
Phase Contrast scans were acquired using a 2D Fast Field Echo (FFE) readout: TR/TE = 22/14 ms,
flip angle = 10°, voxel size = 0.60 × 0.62 × 5.00 mm3
, FOV = 220 × 220 × 5.00 mm3
, and velocity
encoding gradient of 50 cm/s. Six dynamics of 30 second duration were acquired for a total scan
time of 3 minutes.
3.2.3 ASE Model:
The ASE model relies on the known relationship between R2’, OEF, and deoxygenated blood
volume fraction for a network of blood vessels approximated as randomly oriented cylinders with
infinite length [163], [164].
2
′ =
4
3
∗ ∗ ∗ ∗ 0 ∗ ∗
where γ is the gyro-magnetic ratio, B0 is the main magnetic field strength, Hct is the fractional
hematocrit, Δχ is the susceptibility difference between fully oxygenated and fully deoxygenated
blood, and λ is the venous blood volume fraction. With constant TE and a sufficiently long TR, the
signal decay can be simplified to two relaxation regimes separated by 2 = 1.5/ :
() = ∗ (−0.3 ∗ ∗ ( ∗ 2)
2
), 2τ < 1.5/

() = ∗ (− ∗ ∗ 2 + ) , 2τ > 1.5/
where and
stand for short and long time τ, respectively. R2’ and λ can be estimated from
equation [3] by calculating the slope and intercept in the plot of log (SL(t)) versus 2τ, after
substituting “c” derived from the t=0 images used in equation [2].
3.2.4 Image Processing:
For each subject, the ASE data were smoothed by a 3 voxel * 3 voxel Gaussian kernel to improve
SNR. For every voxel in the ASE dataset, the first 7 spin echo values were averaged as the spin
echo value, and the remaining 21 values with different τ were fitted linearly in log space to
generate R2’ and λ for each echo. To minimize the effects of large background magnetic field
inhomogeneities, regions with obvious artifacts or fitting r2 values under 0.85 were excluded
from the final OEF calculation. The remaining voxels were averaged together over the whole
brain.
[1]
[2]
[3]

47
For TRUST data, control–label difference images for each echo time were averaged and fit a
simple mono-exponential function. The decay time constant was corrected for T1 using an
estimated calculation from hematocrit, assuming deoxygenated blood [195]. The resulting T2-
apparent was converted to oxygen saturation using a calibration derived from human blood [122],
[198].
Whole-brain CBF was measured using phase contrast MRI. Regions of interest from the carotid
and vertebral arteries were drawn semiautomatically using a Canny edge detector and resultant
forms summated to yield total cerebral blood flow. Details of phase contrast MRI analysis have
been explained in our previous work [54], [197].
3.2.5 Monte Carlo Simulation:
The general procedure for Monte Carlo Simulation can be found in [175], [180]. Here we
summarized the key steps:
1. A spherical simulation environment (Figure 3.3A) was generated containing vessels
represented by randomly-oriented cylinders with a fixed radius that span the entire simulation
volume. Both arteries (red) and veins (blue) were indicated in the figure. Capillaries were not
shown in the figure for clarity. 2. The Monte Carlo random walk was performed simulating the
diffusion of water protons. Displacement in each direction ( x, y, z) taken by a proton followed
the normal distribution with 0 mean and standard deviation: = √2∆, D is the diffusion
coefficient and ∆t is the time step. Note that the standard deviation of the resultant displacement
has a magnitude of √6∆. 3. The phase accumulation at each time step was calculated by
summing over the field contributions from all vessels based on: ∆φ= γB(x,y,z)∆t, where the local
field is assumed to be constant over ∆t. The induced magnetic field by each blood vessel was
calculated by the equation
∆() =
∆
2

2

2
220, [4]
where cylindrical coordinates (r, , ) were used to describe the position of a proton relative to
the blood vessel [180] (Figure 3.3B). 4. The signal decay curve was plotted by summing up all the
protons with the calculated phase. Only extravascular signals were simulated because both data
in vivo and in silico indicated that the intravascular signal can be neglected in ASE [165], [175].
Therefore, the collisions between protons and vessel walls were neglected. Simulation
parameters were number of protons N=10000, sphere radius=200 um, vessel volume
fraction=3%, TE=60 ms, τ=0-30 ms, D=10-9 mm2
/s.

48
Figure 3.3. The Monte Carlo Simulation model. (A). Arteries are indicated by red cylinders and
veins are blue cylinders. (Capillaries are not shown here) (B). The cylindrical coordinate system
for describing the magnetic field at a point (r, , ). The cylinder is at an angle to the main B0
field.
Although the theoretical framework for the ASE equations is based on uniform, randomly
oriented cylinders having a fixed magnetic susceptibility, real cerebral microvasculature is more
complicated mainly because oxygen exchange is not limited to the capillary bed, causing
arterioles to contribute to spin-dephasing. This effect could be amplified by manipulations in
inspired gasses. We investigated a physiologically realistic distribution of vessel radii from sheep
brains [186]. The so-called Sharan model was originally formulated for oxygen transport
modeling, taking into account parameters such as hematocrit, vascular diameter, blood viscosity,
blood flow, metabolic rate, and P50. This compartmental model had five orders of arterial and
venous vessels, with a range of radii and relative volume fractions, and the radius of the capillary
was fixed. The Sharan brain vasculature model did not provide estimates of oxygen extraction at
each level, forcing us to approximate them from applied boundary conditions. The saturation in
the arterial segments was assumed to track the pulse oximeter, while the saturation in the
venular segments was assumed to track the measured sagittal sinus saturation by TRUST. The
intervening capillary compartment was calculated as a weighted average of the feeding arterial,
capillary, and draining venous saturations [222], [223].
Yc = (1-k)*Ya + k*Yv, k = 0.6 [5]

49
While this neglected the oxygen gradients between the three innermost layers of the
microvasculature, the simple admixture approximation was widely used in both MRI and optical
tissue oxygen saturation estimation [175], [222].
Since oxygen diffusion is known to occur across the whole perfusion unit, the next level of
complexity was to include estimates of oxygen desaturation in the arterioles and venules. The
so-called Sava model was derived from two-photon spectroscopy estimates of microvasculature
size and oxygen saturation [185] in the mouse. Sava et al. separated the perfusing vascular unit
into six segments. Vessel oxygen saturation was estimated by exploiting the differential
absorbance of oxygenated and deoxygenated hemoglobin. They showed that oxygenation
decreased rapidly downstream along the arteriolar tree and provided intravascular oxygen
saturation as a function of arteriolar and venular diameters. Thus, the Sava model used the same
vascular geometry as the Sharan model, but varied oxygen saturation values for blood vessels of
different sizes as summarized in Table 3.1.
Table 3.1. Vessel configuration for vascular simulations in this study
Sharan Model / Sava Model
Vessel Diameter (um) Relative Volume Fraction (%) Baseline Saturation (%)
Sharan / Sava
Arteriolar 1 60 4.3 97/97
Arteriolar 2 30 4.3 97/91
Arteriolar 3 15 4.1 97/87
Arteriolar 4 10 4.1 97/81
Arteriolar 5 5 4.0 97/77
Capillary 2.8 32.5 74
Venule 5 7.5 9.2 58/42
Venule 4 15 9.3 58/44
Venule 3 22.5 9.3 58/46
Venule 2 45 9.6 58/52
Venule 1 90 9.6 58/58

50
Saturation changes in hypoxia
Saturation(%) Baseline Hypoxia1 Hypoxia2
Ya 97 88 75
Yv 58 55 50
3.3 Results
Data from 12 volunteers (4 female, age 36 ± 10 years) were collected. All volunteers successfully
completed the complete scanning protocol. The mild hypoxia was hardly noticed by subjects, but
some found the severe hypoxia to be unpleasant. Most reported the hypercapnic stimulus as
being the most uncomfortable, describing it as being tolerable for the experimental duration but
not sustainable for much longer.
EtO2 and EtCO2 from all volunteers were shown in Figure 3.4. There was one subject failing to
reach the target EtO2 (green line) due to mask leakage. Note that the RespirAct successfully
maintains eucapnia during the hypoxic challenges. SpO2 and TOI values from all volunteers were
shown in Figure 3.5. Mean SpO2 was 95.7 ± 1.2% at baseline but significantly reduced to 87.3 ±
2.9% (p<0.01) and 75.5 ± 5.1% (p<0.01) during mild and severe hypoxia. SpO2 during hypercapnia
was similar to baseline: 96.6 ± 1.1% (p=0.09). The trend of TOI changes was similar to those in
SpO2 for all the phases. TOI decreased from 65.7 ± 6.1% at baseline to 62.7 ± 6.1% during mild
hypoxia (p=0.08), further fell to 55.7 ± 6.5% during severe hypoxia (p<0.01), increased to 69.2 ±
5.9% (p<0.01) during hypercapnia before normalizing during recovery.

51
Figure 3.4. EtO2 and EtCO2 for individuals during each phase are shown by dashed lines. Mean
values and standard deviations are shown by solid black line and error bar. (Hypoxia1 and
Hypoxia2 represent mild and severe hypoxia, respectively)
Figure 3.5. SpO2 and TOI for individuals during each phase. Mean values and standard
deviations are shown by black line and error bar.

52
Figure 3.6 shows representative OEF maps for the same slice during each phase derived from
ASE; the mean values over all valid voxels are displayed below each image. The changes are not
perceptible by the naked eye. There is no grey-white contrast and deep structures appear to have
normal to decreased OEF, unlike anemic subjects.
Figure 3.6. Representative OEF maps during each phase. The average OEF values are indicated
in each map.
Individual mean OEF values of all valid voxels measured by ASE and TRUST were shown in Figure
3.7A. For ASE, the mean OEF values at baseline and mild hypoxia were 38.1% and 38.4%. However,
OEF increased to 40.4% during severe hypoxia (p<0.01). During hypercapnia, OEF subsequently
decreased to 35.4% (p<0.01). On the other hand, mean OEF values measured by TRUST decreased
from 38.9% at baseline to 37.1% (p=0.09) during mild hypoxia, and 36.4% (p<0.01) during severe
hypoxia respectively. Mean OEF by TRUST decreased sharply to 22.9% (p<0.01) during
hypercapnia before recovering to slightly above baseline. Figure 3.7B showed mean OEF change
by NIRS during different phases. OEF decreased from 32.5% at baseline to 30.0% at mild hypoxia,
and 28.2% (p<0.01) at severe hypoxia. However, OEF rebounded to 30.1% during hypercapnia,
registering only 2.4% below baseline.

53
Figure 3.7. (A). OEF values measured by TRUST and ASE during each phase. (B). OEF values
measured by NIRS during each phase. Mean values and standard deviations are shown in blue
for TRUST, red for ASE, and magenta for NIRS. (* denotes statistical significance p<0.01,
compared with baseline)
Figure 3.8 and Table 3.2 summarized the simulation results for both the Sharan and Sava models;
Figure 8 also illustrates the observed OEF changes by ASE as well as predictions of the simple
cylinder model. OEF kept stable for the Sharan model and declined 1.1% for the Sava model
during mild hypoxia. In contrast, OEF increased by 2.3% and 3.3% respectively for the Sharan
model and Sava model during severe hypoxia. The simple cylinder model yielded OEF values
below baseline for both hypoxic conditions.

54
Figure 3.8. Simulation results for both models. Actual (black solid line) represents the
experimental mean ASE values. Cylinder (gray solid line) represents the predicted results by
simple cylinder model. The rest four dashed lines represent predicted results by two models
under two assumptions, respectively.
30%
35%
40%
Baseline Hypoxia 1 Hypoxia 2
OEF
Actual
Cylinder
Sharan
Sava

55
Table 3.2. Simulation results for OEF changes during hypoxia.
3.4 Discussion
In this work, OEF measurements by ASE and TRUST were compared across a wide range
of O2 saturation values using hypoxic and hypercapnic gas challenges. Both methods revealed
that OEF decreased during hypercapnia, which was in line with previous research [119], [136],
[224]. However, OEF predictions diverged during hypoxia. TRUST OEF estimates decreased with
both mild and moderate hypoxia, while ASE OEF stayed the same or increased. Interestingly, NIRS
OEF estimates (SpO2 – TOI)/SpO2 had similar directionality as changes by TRUST but were less
impacted by CO2. Taken together, these data demonstrate that while all three techniques provide
insights into cerebral oxygenation, they differ markedly in their response to physiologic changes
and cannot be used interchangeably.
Simple Cylinder Model
Gas Challenges Baseline Hypoxia 1 Hypoxia 1
OEF 34.2% 31.6% 33.2%
Sharan Vascular Model
Gas Challenges Baseline Hypoxia 1 Hypoxia 1
OEF 35.4% 35.3% 37.7%
Sava Vascular Model
Gas Challenges Baseline Hypoxia 1 Hypoxia 1
OEF 32.4% 31.3% 35.7%
Observed ASE
Gas Challenges Baseline Hypoxia 1 Hypoxia 1
OEF 38.1% 38.4% 40.4%

56
Previous work from our laboratory and others had suggested disparate OEF estimates
measured by TRUST when compared to OEF metrics recorded from deep brain structures [87],
[171], [197] in patients with chronic anemia syndromes. There are several potential reasons for
this disparity, physical, physiological, and technical. OEF can be estimated using venous oximetry
(TRUST, quantitative susceptibility mapping or QSM, simple susceptometry, complex image
summation around a spherical or cylindrical object or CISSCO) or by tissue oximetry (ASE and
NIRS). TRUST isolates the signal from blood in the superior sagittal sinus and estimates OEF based
on the T2 of blood. The predicted venous oxygen saturation is dominated by oxygen supplydemand balance in the cerebral cortex. TRUST follows Fick’s principle and is flow-weighted. Thus,
even though veins entering the sagittal sinus drain blood flow from both grey and white matter,
the higher metabolic activity and blood flow from grey matter dominate the resultant sagittal
sinus T2 value. Thus, TRUST OEF will be impacted primarily by CMRO2 and oxygen delivery to the
grey matter.
In contrast, the ASE derived OEF is based on the tissue R2’, instead of the blood T2. It
captures the signal decay in the extravascular tissue resulting from the paramagnetic
deoxyhemoglobin in the intravascular vessel network, independent of flow distribution within
the tissue, except as the flow distribution modifies the spatial distribution of deoxygenated
hemoglobin. For example, the “average” R2’ in the brain reflects roughly equal contributions of
grey and white matter, despite the 3-4 fold higher blood flow in the grey matter [225], [226].
Even on a microscopic level, spatial averaging and flow averaging of oxygen saturation differ
[185], [217], depending on the flow-distribution within the capillary network. Thus, while
spatially weighted techniques like ASE may provide plausible OEF measurements as well as
important clinical information, they cannot be used to infer CMRO2, nor will they necessarily
match flow-weighted OEF estimates.
R2’ techniques also reflect not only OEF but the arterial saturation as well (SpO2). Under
normoxic conditions, the bulk of deoxygenated hemoglobin is concentrated in the smallest
arterioles, venules, and capillaries. In this scenario, approximating the vascular bed as a random
distribution of uniform cylinders provides reasonable qualitative and quantitative description of
the ASE signal [163], [164]. However, during both mild and severe hypoxia, the total mass of
deoxygenated hemoglobin increases (especially the arterial side), regardless of oxygen utilization
or delivery, confounding the ASE estimate of OEF. This explains the increase in ASE-OEF observed
during severe hypoxia, even as TRUST-OEF was decreasing.
The change in predicted R2’ (and OEF) with hypoxia were validated by Monte Carlo
simulations. We are not the first to examine effects of a physiologically realistic vessel radius
distribution [175] on the ASE signal, but the first to model the impact of arterial desaturation.
Since the data on the vasculature of the human brain were scarce, we selected compartmental
models derived from sheep. It was logical and straightforward to replace a fixed arterial oxygen
saturation Ya, with the Ya’s recorded from the pulse oximetry under different situations. In a
similar manner, Yv from TRUST represented a logical saturation for the largest draining venules.

57
In humans, the distribution of oxygen saturation across the arteriolar-capillary-venular network
is unknown. We chose to model the oxygen exchange across this network two ways. The first
approach was simplistic, assuming a uniform distribution that was 60% of the arteriovenous
difference; this approximation has been used in Near Infrared Spectrometry (NIRS)[216].
Alternatively, we allowed the oxygen saturation to vary across the different levels of the
perfusion unit, using volume fraction, vessel size, and optically-determined oxygen saturation
derived from rat data [185]. All of the models (including the simple cylinder model) slightly
underestimated experimentally-observed OEF at baseline. There are many model imperfections
that could produce a bias, including uncertainty in the susceptibility value [124], [126], incorrect
diffusion coefficient, as well as inadequacies in the vascular models. However, the lower resting
OEF in the Sava model suggests that continuous drop of oxygen saturation across the perfusing
units increases magnetic homogeneity within the tissue. The simple cylinder model failed to
capture the increase in R2’ with desaturation because it neglects the contribution of desaturated
hemoglobin in arteriolar vessels. Both the Sharan and Sava models qualitatively captured the invivo OEF changes. These data indicate that OEF ASE estimates must be used with caution in
patients with arteriolar oxygen desaturation, such as congenital heart disease.
The physical differences between flow-weighted and spatially weighted OEF
measurements are amplified by changes in cerebral blood flow distribution in response to
physiological stimuli. Although OEF by TRUST and by ASE both decreased in response to CO2,
TRUST-OEF declined 14% (absolute units) compared with only 2% for ASE-OEF. The reason for
this disparity is that capillary transit time heterogeneity increases dramatically under vasodilatory
stimuli, with a relatively higher fraction of red blood cells passing through short capillary paths
having little oxygen extraction. Although not a vascular shunt in the classic sense, the more
oxygenated blood from these high-volume vessels produces a large OEF change in flow-weighted
measurements, creating a so-called physiologic shunt. However, much of the capillary network
does not experience significant hyperemia, retaining efficient oxygen extraction. The diffusing
proton in the extravascular space (97% of the ASE signal) is blind to the flow rate within the
capillary, so the decrease in OEF-ASE is much more modest than OEF-TRUST. Not surprisingly,
OEF-NIRS mirrored OEF-ASE, reflecting the spatial-weighting of optical techniques. Physiological
shunting is observed in other hyperemic conditions, including chronic anemia, and likely
contributes to the disparity of OEF measurements between TRUST and ASE in these subjects. The
physiological hyperemia also likely contributed to the OEF differences observed during severe
hypoxia, in addition to direct impact of oxygen saturation on R2’.
In addition to the role of microvascular physiological shunting, there may be flow
distribution heterogeneity on the macrovascular level as well. In chronic anemic syndromes,
deep watershed areas of the white matter are hypoxic by ASE [90], QSM [87], and CISSCO [87],
even while sagittal sinus venous oximetry by TRUST [122] and susceptometry [200] suggest
normal or decreased OEF. One potential explanation is that the compensatory hyperemia caused
by chronic anemia preferentially favors the cerebral cortex, at the expense of deep structures.
Chai et al tested this hypothesis using single-delay arterial spin labeling and observed that oxygen

58
delivery to deep white matter structures was markedly impaired compared to grey matter [86].
One could potentially model this phenomenon using the Monte-Carlo framework, but it would
require more information regarding cerebral metabolism and flow distributions than we had
available.
The last potential contributor to the differences between OEF-ASE and OEF-TRUST is the
calibration equations relating observed parameters (R2’, R2) to oxygen saturation metrics. Use of
inappropriate calibration curves can cause large errors. The original TRUST calibration was
performed using bovine blood over a hematocrit calibration range of 35%-55% [110]. Bovine red
blood cells (RBC) have half the volume of human RBCs and do not aggregate (unlike human RBCs)
but yield reasonable oximetry results in normal subjects. Subsequently, a calibration was derived
in normal human blood over a hematocrit calibration range of (10%-50%) [122] and
independently validated by a second laboratory [198]. Over the shared calibration intervals
(hematocrit 35% – 50%), the bovine and human calibrations agree well with one another except
for a fixed bias of 6% (bovine saturation less than human). However, attempts to extrapolate the
bovine calibration to anemic subjects leads to severe OEF overestimation when validated against
direct co-oximetry (Bush et al., 2017). TRUST calibrations also are also potentially impacted by
diseases that change the RBC size, shape, or hemoglobin density. Patients with sickle cell disease
exhibit a distinct TRUST calibration that yields lower OEF estimates than for normal hemoglobin
because the intrinsic R2 relaxivity of these dense, misshapen RBCs is higher than for normal RBCs
[197], [198]. The calibrations for ASE, CISSCO, QSM, and simple susceptometry all rely on the
fundamental linewidth between fully oxygenated and fully deoxygenated hemoglobin.
Theoretically, this difference should be invariant across individuals and disease states because it
depends on electron spin characteristics, independent of hemoglobin compartmentalization.
However, values between 0.18 [125], [126], [227] and 0.27 [124], [208] have been reported,
without consensus across laboratories, so this constant remains a potential source of uncertainty
between OEF-ASE and OEF-TRUST.
Taken together, our experimental studies demonstrate that TRUST, ASE, and NIRS reflect
different aspects of cerebral physiological compensation to changes in oxygenation and carbon
dioxide. While all three metrics have value, they cannot be used interchangeably. Our Monte
Carlo simulations demonstrate that details regarding microvascular architecture and blood
oxygenation impact OEF by ASE. Disparities between OEF-ASE and OEF-TRUST do not necessarily
reflect errors with either technique, rather they reflect the intrinsic physical differences between
spatially-weighted and volume-weighted oximetry, modulated by physiological responses to
hypoxia, carbon dioxide, or anemia.

59
Chapter 4: Estimates of Intrinsic Tissue Relaxivity Coefficients by Monte Carlo
Simulation Using Physiologically Realistic Models
4.1 Introduction
Dynamic Susceptibility Contrast (DSC) imaging employing gadolinium is a widely utilized
technique in magnetic resonance imaging (MRI) for the assessment of cerebral perfusion and
hemodynamics [228]–[230]. In gadolinium DSC-MRI, a bolus injection of gadolinium contrast is
administered intravenously, followed by rapid image acquisition to capture the contrast agent's
passage through the vascular system. The contrast introduces a paramagnetic susceptibility
effect on surrounding tissues, causing alterations in the local magnetic field. The susceptibilityinduced signal variations are particularly prominent in T2*-weighted images, enabling the
quantification of perfusion-related parameters using tracer kinetic modeling. Gadolinium DSCMRI has proven valuable in the characterization of various neurological conditions, including
brain tumors [231], ischemic strokes [232], and Alzheimer’s disease [233], offering insights into
tissue perfusion dynamics that are crucial for both diagnosis and treatment planning. Despite its
clinical utility, gadolinium DSC-MRI has some limitations. First, it is contraindicated in patients
with renal impairment due to an elevated risk of gadolinium-induced nephrogenic systemic
fibrosis [234]. Second, the accurate quantification of absolute cerebral blood flow may be
challenging due to factors such as the variable concentration of the contrast agent in the blood
and differences in individual vascular geometry. Another challenge lies in the need for short and
steep boluses because deconvolution of the brain signal is required. This can be hard to achieve
because of injection rate limitations in subjects with compromised venous access and gadolinium
dispersion in the vascular system.
Recently, Deoxygenation-based Dynamic Susceptibility Contrast (dDSC) MRI has been
introduced to serve as an endogenous contrast alternative to gadolinium injection [235]–[237].
Given the paramagnetic nature of deoxyhemoglobin, its presence within intravascular spaces
leads to interactions with the local magnetic field, inducing alterations in the phase of
surrounding tissue protons. Consequently, this interaction results in signal losses, analogous to
the effects observed with gadolinium [238]. This similarity in signal alterations facilitates the
application of tracer kinetic modeling to calculate parameters such as cerebral blood flow (CBF),
cerebral blood volume (CBV), and mean transit time (MTT). Initial studies used hyperoxia or
nitrogen exposure to create signal changes, however it was challenging to control the exposures
as well as secondary changes in end-tidal carbon dioxide concentration which could impact CBF.
To surmount these constraints, Poublanc et al. from the University of Toronto employed
a RespirAct commercial gas-blender system (Thornhill Medical, Toronto) to induce rapid changes
in deoxyhemoglobin concentration in their study [220], [239]. The RespirAct system
demonstrates versatility by allowing quick variations in inhaled oxygen, nitrogen, and carbon

60
dioxide across wide ranges, and targeting end-tidal O2 (EtO2) and end-tidal CO2 (EtCO2) partial
pressures. With the help of RespirAct, rectangular pulses inducing changes in oxygenation were
easily administered in both the negative (desaturation) and positive (resaturation) directions
[240]. Both methodologies produced credible estimations of relative cerebral blood flow and
blood volume. These results were further validated by numerical simulations of BOLD changes to
gadolinium DSC [241].
Based on these prior works, our group compared three gas challenges Desaturation
(transient hypoxia), Resaturation (transient normoxia), and SineO2 (sinusoidal modulation of
end-tidal oxygen pressures) in a cohort of 10 healthy volunteers (age 37 ± 11, 60% female). The
results showed moderate correlation and limits of agreement between dDSC and gadolinium DSC.
However, mean CBF levels were much lower during desaturation than resaturation (33.2±7.2 vs.
54.7±19.3 mL/100g/min). Relative CBF maps were highly concordant, suggesting that the
disparity either represented true flow differences or systematic bias in the intrinsic relaxivity of
deoxyhemoglobin. Observed differences were much larger than the expected hypoxic
vasodilation [242], indicating scale bias. Prior in-vivo and in-silico work by Schulman et al
demonstrated that the scaling constant between the BOLD signal and tissue deoxyhemoglobin
was higher in the presence of baseline hypoxia [241]. Since it is impractical to perform Monte
Carlo simulation for every conceivable vascular stimulus, we sought to derive generalizable
equations to calculate the scaling coefficient between BOLD and tissue deoxyhemoglobin
concentration for plausible variations in arterial oxygen saturation(Sa), oxygen extraction
fraction(OEF), hemoglobin(Hb), and CBV, using a realistic distribution of vessel sizes and oxygen
saturations.
4.2 Methods
For paramagnetic tracers, a linear relationship is observed between the BOLD signal and tissue
tracer concentration as follows [243]:
∆2,
∗
() = (
′
) × (), [1]
where represents the intrinsic contrast relaxivity coefficient and r’ which represents the
vascular geometry. Determination of this coefficient usually requires Monte Carlo simulation
methods. For dDSC, the contrast agent is the induced deoxyhemoglobin, [deoxyHb], produced
during hypoxia which varies with the vascular geometry r’, Hb, CBV, OEF, and Sa as follows.
∆2,
∗ = (
′
, , ,, ) [2]
In dDSC experiments, Sa is the primary variable being manipulated while the other variables serve
as constant, patient-specific, “operating points”. To calibrate the BOLD signal to the underlying
concentration of deoxygenated hemoglobin, it is necessary to calculate the partial derivative of
[2] with respect to [deoxyHb] as follows:

61
Δ2
∗ = (

[]
) [] [3]
Comparing equations [2] and [3] it follows that
= (

[]
) [4]
where [deoxyhemoglobin] can be calculated from Sa, OEF, Hb, and CBV. Note rtissue is known to
vary quadratically with Sa for large Sa fluctuations, thus equation [4] effectively represents a first
order approximation. The behavior of F with respect to CBV, Hb, OEF, and r’ has not been
previously characterized.
The general principles for Monte Carlo Simulation of this nature can be found in [175], [180].
Here we summarized the key steps specific to the present work.
1. An illustrative spherical simulation environment (Figure 4.1A) was generated to depict vessels
represented by randomly oriented cylinders with a fixed radius spanning the entirety of the
simulation volume. The figure delineates both arteries (depicted in red) and veins (depicted in
blue), with capillaries intentionally omitted for visual clarity. Within the time-scale of the imaging
experiment, the cylinders appear infinitely long. The volume fraction and magnetic susceptibility
of the different vessel radii was varied in a controlled manner as described later.

62
Figure 4.1. The Monte Carlo Simulation model. (A): Arteries are indicated by red cylinders and
veins are blue cylinders. (Capillaries are not shown here) (B): The cylindrical coordinate system
for describing the magnetic field at a point (r, , ). The cylinder is at an angle to the main B0
field. (C)(D). Fingertip pulse oximetry (SpO2) for desaturation and resaturation.
2. The Monte Carlo random walk was employed to simulate the diffusion of water protons. The
displacement in each direction (δx, δy, δz) undertaken by a proton adhered to a normal
distribution with a mean of 0, and a standard deviation calculated as = √2∆, where D
represents the diffusion coefficient, and ∆t denotes the time step. It is noteworthy that the
standard deviation of the resultant displacement attains a magnitude of √6∆.
3. The phase accumulation at each time step was computed by aggregating the field
contributions from all vessels, based on the equation ∆φ = γB(x, y, z)∆t, assuming the local field
remains constant over ∆t. The induced magnetic field by each blood vessel was determined using
cylindrical coordinates (r, θ, ϕ) to describe the proton's position relative to the blood vessel
(Figure 1B), as defined by the equation:
∆() =
∆
2

2

2
220, [5]
4. The signal decay curve was generated by summing up all protons with the calculated phase.
The simulation exclusively considered extravascular signals due to both in vivo and in silico data
indicating that intravascular signal contributions can be neglected in ASE [166], [175].
Consequently, collisions between protons and vessel walls were omitted. The main simulation
parameters encompassed a proton count of N = 10,000, a sphere radius of 200 µm, TE of 60 ms,
τ (time constant) ranging from 0 to 30 ms, and a diffusion coefficient D of D=10-9 mm2
/s.
The simplest vascular geometry was that all vessels had the same size, and we called it the
“simple cylinder model”. In this model, we only considered veins and the saturation was fixed.
The vascular geometry, r’, for Equation 1 was derived from previously published work by Sharan
in sheep brains [186]. This model reports vessel radii distributions and volume fractions of the
arteriolar, venous, and capillary compartments over eleven levels, which are summarized in Table
1. This anatomically-based model could not describe oxygen saturations across the vascular tree.
We first performed a simplified simulation with oxygen exchange limited to the capillary bed with
a capillary saturation reflecting a linear admixture of the cerebral arteriolar (Ya) and venous
saturations (Yv) as follows:
= (1 − ) × + × , = 0.6 [6]
This approximation is widely used in both the MRI and Near Infrared Spectroscopy literature
[175], [222]. For the simulations, Yv was calculated by Ya and OEF by the following relationship:
= × (1 − ) [7]

63
For in-vivo measurements, Ya can be set to pulse oximetry values (Sa) and Yv can be assumed to
track sagittal sinus saturation measured by T2-Relaxation-under-Spin-Tagging (TRUST) [87], [110],
[111], [197].
However, it is well known that oxygen exchange is not limited to the capillary. Using oxygensensitive two photon optical measurements of mouse cerebral microvasculature, Sava and
colleagues have characterized oxygen saturations across different vessel radii [218]. Using the
assumption that the relative oxygen changes were constant across Sa and OEF, we repeated the
entire Monte Carlo simulation framework using the Sharan geometry, but including more realistic
oxygen saturation distributions. Table 4.1, right hand column compares the saturations as a
function of radius for Ya = 97% and Yv = 58%, for the Sharan and Sava models, respectively.
Relative oxygenation changes from the Sava two-photon data were mapped to absolute
saturation values using linear interpolation.
Table 4.1. Diameter, volume fraction, and oxygen saturation for the Sharan and Sava model
Sharan Model / Sava Model
Vessel Diameter
(um)
Relative Volume
Fraction (%)
Baseline
Saturation (%)
Sharan / Sava
Arteriolar 1 60 4.3 97/97
Arteriolar 2 30 4.3 97/91
Arteriolar 3 15 4.1 97/87
Arteriolar 4 10 4.1 97/81
Arteriolar 5 5 4.0 97/77
Capillary 2.8 32.5 74
Venule 5 7.5 9.2 58/42
Venule 4 15 9.3 58/44
Venule 3 22.5 9.3 58/46
Venule 2 45 9.6 58/52
Venule 1 90 9.6 58/58
To determine the effects of Hb, OEF, CBV, and Sa changes on R2*, we varied these parameters
over the physiologically plausible range: Hb (0.04-0.18 g/L in steps of 0.01 g/L), CBV (1%-10% in
steps of 1%), OEF (10%-60% in steps of 5%), and Sa (60%-100% in steps of 5%). By Equation 1, our
simulations resulted in a complicated four-dimensional surface. To visualize changes of each
individual parameter, we linearized about a logical operating point, consisting of Hb = 0.13 g/L,
CBV = 4%, OEF = 40%, and Sa = 100%, representing normal values for healthy subjects.

64
4.3 Results
Figure 4.2 demonstrates the mean changes in oxygen saturation for the desaturation and
resaturation pulses in 10 subjects, respectively [244]. Average baseline saturations were 97% and
78%, respectively, with a change in saturation of 13%. Note the saturation change is more rapid
for resaturation than desaturation [244].
Figure 4.2. (A). The relationship between ∆2,
∗
and concentration of deoxygenated
hemoglobin ∆[] at the operating point for the simple cylinder model. The slope of the
linear fitting is . (B). One representative fitting curve between and (1 − )
2
for
Sharan model.
Figure 4.2A shows one representative pattern of simulated data points to derive the relationship
between ∆2,
∗
and concentration of deoxygenated hemoglobin ∆[] at the operating
point for the simple cylinder model. The change in deoxyhemoglobin concentration ∆[] was
calculated from the saturation ∆, volume fraction CBV, hemoglobin , and hematocrit as
∆[] = ∆ × ×

. There is a linear relationship between ∆2,
∗
and ∆[] and
the slope is . The ratio of Hb to Hct is known as the mean cellular hemoglobin concentration
and is close to 0.34 g/% in normal subjects.
Figure 4.3 demonstrates the marginal relationships between and Sa, OEF, CBV, and Hb
about the operating point (100%, 40%, 4%, 0.13 g/L) using the Sharan architecture. There was a
clear quadratic relationship between and Sa, while varied linearly with respect to
OEF. R2* relaxivity demonstrated weakly quadratic relationships with respect to both CBV and
Hb. We ignored the CBV nonlinearity because it was negligible but compared both linear and
quadratic fits to the rtissue – Hb relationship.

65
Figure 4.3. Separate relationship between and Sa, OEF, CBV, and Hb. Fitting line is also
shown in each figure.
For linear Hb, we collapsed OEF, CBV, and Hb to the product of these three parameters as follow:
= 0 + (1 − )
2 × (1 + 2 × × × ) [8]
To accommodate a quadratic relationship between rtissue and Hb, we modified equation [4] to
include an additional constant k3.
= 0 + (1 − )
2 × (1 + 2 × × × ( − 3)
2
) [9]
For both fitting methods, we simulated 10000 data points and used MSE as the loss function.
Detailed values for these constants in the equations were shown in Table 4.2. It clearly showed
that the MSE was ~ 30% smaller for quadratic fitting for both the Sharan and Sava models (Table
4.2), although the difference in the resultant rtissue value was minimal.
Table 4.2 also summarizes the predicted rtissue values under normal oxygenation, desaturation,
and resaturation by the Sava and Sharan models, using mean OEF and Sa values measured from
patients [244]. by Sava model was 1%-4% larger than Sharan model under all the

66
conditions. Baseline oxygenation had a much larger effect, with more than 25% higher for
resaturation for desaturation.
Table 4.2. Calculated under normal condition, desaturation, and resaturation by different
models.
r0 k1 k2 k3 MSE
Sharan/Linear 21.3 189.4 137.7 0.25
Sharan/Quadratic 21.0 175.4 1403.7 0.04 0.16
Sava/Linear 21.9 186.9 136.3 0.28
Sava/Quadratic 21.7 173.4 1345.6 0.04 0.20
Normal Desaturation Resaturation
OEF 38% 32% 29%
Sa 99% 90% 80%
r by Sharan/Linear 21.8 24.2 30.3
r by
Sharan/Quadratic
22.1 24.8 31.4
r by Sava/Linear 22.2 24.5 31.1
r by Sava/Quadratic 22.4 25.3 32.6

67
4.4 Discussion
In this work, we used Monte Carlo simulation to investigate the effects of hemoglobin,
OEF, CBV, and arterial saturation on the intrinsic tissue relaxivity coefficient scaling R2* between
BOLD and induced changes in deoxygenated hemoglobin concentration. We also derived an
empirical model based on these parameters to allow calculation of r2*tissue for any combination
of Sa, OEF, Hb, and CBV. We found that including realistic estimates of oxygen unloading across
the vascular bed had modest impact on r2*tissue, particularly compared with the impact of baseline
oxygen saturation. Similar to previous work, we found that r2*tissue exhibited an inverse quadratic
relationship with arterial saturation. The difference in the simulated under desaturation
and resaturation (24.5 vs. 31.3) is an important contributor to CBF differences previously noted
in dDSC experiments [245].
We are not the first to examine effects of a physiologically realistic vessel radius
distribution in MR signal simulation [175], nor the first to understand the contribution of baseline
saturation on rtissue [241], however, we are the first to combine the two approaches. More
importantly, we are the first to fit the resultant four dimensional surface to tractable empiric
calibration equations, similar to the approach for TRUST calibrations [93], [122]. These equations
allow patient and experiment rtissue estimation from measurements of arterial saturation, venous
saturation, and hemoglobin value. CBV would typically be assumed, however could be potentially
estimated using a bootstrap method from the dDSC experiment itself [245].
The models have some important limitations. Since data on the vasculature of the human
brain are scant and nonquantitative, we selected compartmental models derived from sheep
because they are morphologically more similar to humans than rodent brains. Nonetheless, we
do not know how close the sheep vessel sizes and distribution resemble those found in human
brains. Similarly, the distribution of oxygen saturation across the arteriolar-capillary-venular
network may differ between mice and men. Fortunately, increasing model complexity appears
to yield diminishing returns, compared to the key contribution of the oxygenation “set point” for
the dDSC challenges. In fact, the key failure of the simple cylinder model for dDSC modeling is
the inability to capture the paramagnetic effects of deoxygenated hemoglobin in the feeding
arteriolar bed. Conceptually, the “venous” cerebral blood volume is increased when arterial
blood is not fully saturated.
In conclusion, we have demonstrated that the tissue scaling coefficient for dDSC
experiments can be captured using simple parametric equations based on Sa, Hb, OEF, and CBV.
The first three parameters can be directly measured, while CBV must be assumed or estimated
using other MRI techniques. These equations improve agreement between dDSC CBF estimates
performed using different baseline oxygen saturation methods and may also improve
quantitative BOLD measurements in patient populations having baseline oxygen desaturation
(such as congenital heart disease). Further refinement of the vascular models is possible, but our
data suggests rapidly diminishing returns for additional complexity.

68
Chapter 5 Conclusion
This dissertation contributes to the development, validation, and clinical translation of
quantitative MR techniques of measuring oxygen saturation in patients with sickle cell disease,
as well as healthy subjects. Project outcomes can advance the evaluation of brain oxygenation
and brain metabolism.
1. Two susceptibility-based MR techniques (QSM and CISSCO) were validated in the internal
cerebral vein (ICV) in quantifying oxygen extraction fraction (OEF). The results reflect the oxygen
saturation in the deep structures of brain, which were compared with the oxygen saturation in
the superior sagittal sinus (SSS) revealed by TRUST. Both susceptibility-based methods indicate
that OEF in ICV changed reciprocally with oxygen content. However, OEF in SSS increased with
oxygen content by the TRUST measurement. The opposite results suggest that anemia worsens
hypoxia in deep brain structures while the cerebral cortex appeared to be spared. The vascular
“steal phenomenon” may exist, where oxygen delivery to the cortex is preserved at the expense
of deep regions. There was also a bias between QSM and CISSCO, and the bias was larger in SCD
patients. The other striking finding was that left-shifted hemoglobin appears to be a powerful
modulator of OEF by TRUST and explained the group differences in sagittal sinus OEF.
2. TRUST and ASE were validated through gas challenges, including hypoxia and hypercapnia.
Both methods revealed that OEF decreases during hypercapnia, in line with other studies, but
the magnitude of change was much larger using TRUST, reflecting fundamental physical
differences between the signal loss mechanisms and their physiological interpretations. However,
contrary results were found during hypoxia. OEF by TRUST decreased with both mild and severe
hypoxia, while OEF by ASE stayed the same or increased. The OEF measurements by the optical
method NIRS have similar directionality as changes by TRUST, but were less impacted by CO2.
Taken together, we demonstrated that OEF by ASE, TRUST, and NIRS view tissue oxygenation
through different physical “lenses” and cannot be used interchangeably。
3. Monte Carlo Simulation was a strong tool to simulate the MR signal formation and investigate
the interactions between vessel geometry, oxygenation, and diffusion effects. The simple
cylinder model did not describe the OEF changes during hypoxia accurately, while two
physiologically realistic models (Sharan and Sava) captured the direction and magnitude of the
hypoxic response. Both the Sharan and Sava models were biased (3%-5%) compared with the
observed results, reflecting systematic deviation in our assumed parameters.
4. Monte Carlo simulation was also used to investigate the effects of hemoglobin, OEF, CBV, and
arterial saturation on the intrinsic tissue relaxivity coefficient scaling R2* by BOLD to changes in
deoxygenated hemoglobin concentration. We also derived an empirical model based on these
parameters to allow calculation of r2*tissue for any combination of Sa, OEF, Hb, and CBV. The results
indicated that r2*tissue exhibits an inverse quadratic relationship with arterial saturation, but a

69
nearly linear relationship to the other parameters. The difference in the simulated under
desaturation and resaturation (24.5 vs. 31.3) is an important contributor to CBF differences
previously noted in the dDSC experiments [245].
In conclusion, three different MR techniques for tissue oxygenation were cross-validated across
one another and against NIRs. They all can provide information regarding OEF but cannot be used
interchangeably. The widely used TRUST follows the Fick’s principle, and is flow-weighted, mainly
impacted by CMRO2. In contrast, ASE is spatially weighted, reflecting the average R2’ in brain
tissues. It is dominated by patient hematocrit, independent of whole brain OEF. ASE yielded
similar results to TRUST under resting conditions but diverged with TRUST during hypoxia and
hypercapnia. Although OEF by ASE declined with hypercapnia, it does not accurately reflect the
increased oxygen delivery from the resulting hyperemia. Furthermore, it failed to reflect changes
in tissue oxygen saturation under conditions of arterial desaturation. For deep brain tissues, ASE,
CISSCO, and QSM generated comparable results and suggested tissue hypoxia with anemia. QSM
and CISSCO yielded qualitatively similar results, but QSM was badly biased from partial volume
effects.
As a result, none of these OEF techniques we examined is fully generalizable. TRUST is suited
primarily to the superior sagittal sinus, providing a global OEF measurement. ASE requires high
SNR and careful estimation of venous blood volume to avoid confounding parameters. It also fails
during arterial desaturation and cannot be used to infer changes in CMRO2. CISSCO is limited to
veins orthogonal to the main field and require circumferential tissue sampling, preventing
assessment of superficial veins. QSM is heavily affected by partial volume effect and is also
challenging on the brain surface because of the removal of background field. A key limitation to
our research was the unavailability of a true reference standard such as jugular venous sampling
or oxygen positron-emitting-tomography.

70
Future work
This thesis introduces and validates several non-invasive MRI methodologies designed for the
assessment of oxygen extraction fraction. However, its applicability to various pathologies,
including overt strokes, brain tumors, Fontan, and neurodegenerative diseases, remains only
partially explored. Consequently, further clinical trials involving these diverse patient populations
are imperative to ascertain the clinical utility of this technique in both diagnostics and therapeutic
contexts.
Cares must be taken for the various MRI techniques in quantifying oxygenation. Venous
saturation imaging utilizing T2 oximetry has found widespread application across diverse
pathologies. Yet, it is crucial, even in retrospective analyses, to implement suitable and diseasespecific T2 and susceptibility-based oximetry calibrations, particularly in conditions such as
hemoglobinopathies where the magnetic properties of red blood cells deviate from the norm.
Following the acquisition of oxygen extraction fraction and cerebral metabolic rate of oxygen
(CMRO2) estimates using appropriate calibrations, a comprehensive biomarker analysis is
warranted. This analysis should identify predictors of decreased oxygenation and the risk of silent
cerebral infarctions (SCI) on a more extensive and heterogeneous cohort, ensuring a balanced
representation across subtypes of sickle cell disease and congenital anemias.
Lastly, there is a need to explore the neuropsychological correlates of impaired hemodynamics.
This exploration is essential for elucidating brain-behavior relationships and providing insights
for early prevention and intervention strategies tailored to individuals at risk.

71
References
[1] K. Ohene-Frempong et al., “Cerebrovascular accidents in sickle cell disease: rates and
risk factors,” Blood, The Journal of the American Society of Hematology, vol. 91, no. 1, pp. 288–
294, 1998.
[2] L. Pauling and H. A. Itano, “Sickle cell anemia a molecular disease,” Science, vol. 110, no.
2865, pp. 543–548, Nov. 1949, doi: 10.1126/science.110.2865.543.
[3] C. T. Noguchi, A. N. Schechter, and G. P. Rodgers, “Sickle cell disease pathophysiology,”
Baillieres Clin Haematol, vol. 6, no. 1, pp. 57–91, Mar. 1993, doi: 10.1016/s0950-
3536(05)80066-6.
[4] M. H. Steinberg, “Pathophysiology of sickle cell disease,” Baillieres Clin Haematol, vol.
11, no. 1, pp. 163–184, Mar. 1998, doi: 10.1016/s0950-3536(98)80074-7.
[5] S. S. Abdi, M. De Haan, and F. J. Kirkham, “Neuroimaging and Cognitive Function in
Sickle Cell Disease: A Systematic Review,” Children (Basel), vol. 10, no. 3, p. 532, Mar. 2023, doi:
10.3390/children10030532.
[6] K. L. Hassell, “Population estimates of sickle cell disease in the US,” American journal of
preventive medicine, vol. 38, no. 4, pp. S512–S521, 2010.
[7] D. Lubeck et al., “Estimated Life Expectancy and Income of Patients With Sickle Cell
Disease Compared With Those Without Sickle Cell Disease,” JAMA Network Open, vol. 2, no. 11,
p. e1915374, Nov. 2019, doi: 10.1001/jamanetworkopen.2019.15374.
[8] P. Shah et al., “Mental stress causes vasoconstriction in subjects with sickle cell disease
and in normal controls,” Haematologica, vol. 105, no. 1, p. 83, 2020.
[9] N. Conran and J. D. Belcher, “Inflammation in sickle cell disease,” Clinical hemorheology
and microcirculation, vol. 68, no. 2–3, pp. 263–299, 2018.
[10] I. Akinsheye and E. S. Klings, “Sickle cell anemia and vascular dysfunction: the nitric
oxide connection,” Journal of cellular physiology, vol. 224, no. 3, pp. 620–625, 2010.
[11] J. A. Switzer, D. C. Hess, F. T. Nichols, and R. J. Adams, “Pathophysiology and treatment
of stroke in sickle-cell disease: present and future,” The Lancet Neurology, vol. 5, no. 6, pp.
501–512, 2006.
[12] L. A. Verduzco and D. G. Nathan, “Sickle cell disease and stroke,” Blood, The Journal of
the American Society of Hematology, vol. 114, no. 25, pp. 5117–5125, 2009.

72
[13] E. H. Lo, T. Dalkara, and M. A. Moskowitz, “Mechanisms, challenges and opportunities in
stroke,” Nature reviews neuroscience, vol. 4, no. 5, pp. 399–414, 2003.
[14] M. S. Phipps and C. A. Cronin, “Management of acute ischemic stroke,” Bmj, vol. 368,
2020.
[15] L. Goldstein et al., “Hemorrhagic stroke in the stroke prevention by aggressive reduction
in cholesterol levels study,” Neurology, vol. 70, no. 24_part_2, pp. 2364–2370, 2008.
[16] E. E. Smith, J. Rosand, and S. M. Greenberg, “Hemorrhagic stroke,” Neuroimaging
Clinics, vol. 15, no. 2, pp. 259–272, 2005.
[17] C. S. Kase et al., “Prevalence of silent stroke in patients presenting with initial stroke: the
Framingham Study.,” Stroke, vol. 20, no. 7, pp. 850–852, 1989.
[18] A. L. Ford et al., “Silent infarcts in sickle cell disease occur in the border zone region and
are associated with low cerebral blood flow,” Blood, The Journal of the American Society of
Hematology, vol. 132, no. 16, pp. 1714–1723, 2018.
[19] S. T. Miller et al., “Silent infarction as a risk factor for overt stroke in children with sickle
cell anemia: a report from the Cooperative Study of Sickle Cell Disease,” The Journal of
pediatrics, vol. 139, no. 3, pp. 385–390, 2001.
[20] M. R. DeBaun, F. D. Armstrong, R. C. McKinstry, R. E. Ware, E. Vichinsky, and F. J.
Kirkham, “Silent cerebral infarcts: a review on a prevalent and progressive cause of neurologic
injury in sickle cell anemia,” Blood, vol. 119, no. 20, pp. 4587–4596, May 2012, doi:
10.1182/blood-2011-02-272682.
[21] E. J. van Beers, C. F. van Tuijn, P. T. Nieuwkerk, P. W. Friederich, J. H. Vranken, and B. J.
Biemond, “Patient-controlled analgesia versus continuous infusion of morphine during vasoocclusive crisis in sickle cell disease, a randomized controlled trial,” American journal of
hematology, vol. 82, no. 11, pp. 955–960, 2007.
[22] J. Howard et al., “The Transfusion Alternatives Preoperatively in Sickle Cell Disease
(TAPS) study: a randomised, controlled, multicentre clinical trial,” The Lancet, vol. 381, no.
9870, pp. 930–938, 2013.
[23] R. J. Adams, D. Brambilla, and Optimizing Primary Stroke Prevention in Sickle Cell
Anemia (STOP 2) Trial Investigators, “Discontinuing prophylactic transfusions used to prevent
stroke in sickle cell disease,” N Engl J Med, vol. 353, no. 26, pp. 2769–2778, Dec. 2005, doi:
10.1056/NEJMoa050460.

73
[24] M. T. Lee et al., “Stroke Prevention Trial in Sickle Cell Anemia (STOP): extended followup and final results,” Blood, vol. 108, no. 3, pp. 847–852, Aug. 2006, doi: 10.1182/blood-2005-
10-009506.
[25] R. J. Adams, F. T. Nichols, R. Figueroa, V. McKie, and T. Lott, “Transcranial Doppler
correlation with cerebral angiography in sickle cell disease.,” Stroke, vol. 23, no. 8, pp. 1073–
1077, 1992.
[26] R. J. Adams, “TCD in sickle cell disease: an important and useful test,” Pediatric
radiology, vol. 35, no. 3, pp. 229–234, 2005.
[27] M. E. Fields et al., “Hydroxyurea reduces cerebral metabolic stress in patients with sickle
cell anemia,” Blood, The Journal of the American Society of Hematology, vol. 133, no. 22, pp.
2436–2444, 2019.
[28] S. J. Nevitt, A. P. Jones, and J. Howard, “Hydroxyurea (hydroxycarbamide) for sickle cell
disease,” Cochrane Database of Systematic Reviews, no. 4, 2017.
[29] O. S. Platt, “Hydroxyurea for the Treatment of Sickle Cell Anemia,” New England Journal
of Medicine, vol. 358, no. 13, pp. 1362–1369, Mar. 2008, doi: 10.1056/NEJMct0708272.
[30] M. Gwam and C. Nwosu, “Treating sickle cell anaemia: the TWiTCH trial,” The Lancet,
vol. 388, no. 10048, pp. 960–961, 2016.
[31] C. T. Quinn, “TWiTCH: itching to find an alternative to transfusions for prevention of
stroke in sickle cell anemia,” The Hematologist, vol. 10, no. 6, 2013.
[32] K. J. Helton et al., “Magnetic resonance imaging/angiography and transcranial Doppler
velocities in sickle cell anemia: results from the SWiTCH trial,” Blood, The Journal of the
American Society of Hematology, vol. 124, no. 6, pp. 891–898, 2014.
[33] R. E. Ware et al., “Stroke With Transfusions Changing to Hydroxyurea (SWiTCH): a phase
III randomized clinical trial for treatment of children with sickle cell anemia, stroke, and iron
overload,” Pediatric blood & cancer, vol. 57, no. 6, pp. 1011–1017, 2011.
[34] J. M. Kawadler, C. A. Clark, R. C. McKinstry, and F. J. Kirkham, “Brain atrophy in
paediatric sickle cell anaemia: findings from the silent infarct transfusion (SIT) trial,” British
Journal of Haematology, vol. 177, no. 1, pp. 151–153, 2017.
[35] P. L. Croal, J. Leung, P. Kosinski, M. Shroff, I. Odame, and A. Kassner, “Assessment of
cerebral blood flow with magnetic resonance imaging in children with sickle cell disease: A

74
quantitative comparison with transcranial Doppler ultrasonography,” Brain and Behavior, vol. 7,
no. 11, p. e00811, 2017.
[36] M. H. Steinberg, “Management of sickle cell disease,” New England Journal of Medicine,
vol. 340, no. 13, pp. 1021–1030, 1999.
[37] P. T. McGann, A. C. Nero, and R. E. Ware, “Current Management of Sickle Cell Anemia,”
Cold Spring Harb Perspect Med, vol. 3, no. 8, p. a011817, Aug. 2013, doi:
10.1101/cshperspect.a011817.
[38] H. F. Bunn, “Pathogenesis and treatment of sickle cell disease,” New England Journal of
Medicine, vol. 337, no. 11, pp. 762–769, 1997.
[39] L. B. Herity, D. M. Vaughan, L. R. Rodriguez, and D. K. Lowe, “Voxelotor: a novel
treatment for sickle cell disease,” Annals of Pharmacotherapy, vol. 55, no. 2, pp. 240–245,
2021.
[40] E. Vichinsky et al., “A phase 3 randomized trial of voxelotor in sickle cell disease,” New
England Journal of Medicine, vol. 381, no. 6, pp. 509–519, 2019.
[41] K. I. Ataga et al., “Crizanlizumab for the prevention of pain crises in sickle cell disease,”
New England Journal of Medicine, vol. 376, no. 5, pp. 429–439, 2017.
[42] N. E. Cieri-Hutcherson, T. C. Hutcherson, E. E. Conway-Habes, B. N. Burns, and N. A.
White, “Systematic review of L-glutamine for prevention of vaso-occlusive pain crisis in patients
with sickle cell disease,” Pharmacotherapy: The Journal of Human Pharmacology and Drug
Therapy, vol. 39, no. 11, pp. 1095–1104, 2019.
[43] A. Sadaf and C. T. Quinn, “L-glutamine for sickle cell disease: Knight or pawn?,”
Experimental Biology and Medicine, vol. 245, no. 2, pp. 146–154, 2020.
[44] S. T. Chou, “Transfusion therapy for sickle cell disease: a balancing act,” Hematology
2013, the American Society of Hematology Education Program Book, vol. 2013, no. 1, pp. 439–
446, 2013.
[45] S. T. Chou and R. M. Fasano, “Management of patients with sickle cell disease using
transfusion therapy: guidelines and complications,” Hematology/Oncology Clinics, vol. 30, no. 3,
pp. 591–608, 2016.
[46] M. C. Walters et al., “Bone marrow transplantation for sickle cell disease,” New England
Journal of Medicine, vol. 335, no. 6, pp. 369–376, 1996.

75
[47] “Bone Marrow Transplantation for Sickle Cell Disease | NEJM.” Accessed: Jun. 16, 2023.
[Online]. Available: https://www.nejm.org/doi/full/10.1056/nejm199608083350601
[48] M. D. Hoban, S. H. Orkin, and D. E. Bauer, “Genetic treatment of a molecular disorder:
gene therapy approaches to sickle cell disease,” Blood, The Journal of the American Society of
Hematology, vol. 127, no. 7, pp. 839–848, 2016.
[49] J.-A. Ribeil et al., “Gene therapy in a patient with sickle cell disease,” New England
Journal of Medicine, vol. 376, no. 9, pp. 848–855, 2017.
[50] E. R. Kandel et al., Principles of neural science, vol. 4. McGraw-hill New York, 2000.
[51] D. Rolfe and G. C. Brown, “Cellular energy utilization and molecular origin of standard
metabolic rate in mammals,” Physiological reviews, vol. 77, no. 3, pp. 731–758, 1997.
[52] J. Cho et al., “Cerebral metabolic rate of oxygen (CMRO2) mapping by combining
quantitative susceptibility mapping (QSM) and quantitative blood oxygenation level-dependent
imaging (qBOLD),” Magnetic Resonance in Medicine, vol. 80, no. 4, pp. 1595–1604, 2018, doi:
10.1002/mrm.27135.
[53] A. Heyman, J. L. Patterson, T. W. Duke, and others, “Cerebral circulation and metabolism
in sickle cell and other chronic anemias, with observations on the effects of oxygen inhalation,”
The Journal of Clinical Investigation, vol. 31, no. 9, pp. 824–828, 1952.
[54] C. Vu et al., “Reduced global cerebral oxygen metabolic rate in sickle cell disease and
chronic anemias,” American Journal of Hematology, vol. 96, no. 8, pp. 901–913, 2021.
[55] J.-M. Lee et al., “Magnetic resonance cerebral metabolic rate of oxygen utilization in
hyperacute stroke patients,” Annals of Neurology: Official Journal of the American Neurological
Association and the Child Neurology Society, vol. 53, no. 2, pp. 227–232, 2003.
[56] P. Borghammer, M. Vafaee, K. Ostergaard, A. Rodell, C. Bailey, and P. Cumming, “Effect
of memantine on CBF and CMRO2 in patients with early Parkinson’s disease,” Acta neurologica
scandinavica, vol. 117, no. 5, pp. 317–323, 2008.
[57] P. Borghammer et al., “Cerebral oxygen metabolism in patients with early Parkinson’s
disease,” Journal of the neurological sciences, vol. 313, no. 1–2, pp. 123–128, 2012.
[58] P. L. Madsen et al., “Cerebral O2 metabolism and cerebral blood flow in humans during
deep and rapid-eye-movement sleep,” J Appl Physiol (1985), vol. 70, no. 6, pp. 2597–2601, Jun.
1991, doi: 10.1152/jappl.1991.70.6.2597.

76
[59] K. M. Sicard and T. Q. Duong, “Effects of hypoxia, hyperoxia, and hypercapnia on
baseline and stimulus-evoked BOLD, CBF, and CMRO2 in spontaneously breathing animals,”
NeuroImage, vol. 25, no. 3, pp. 850–858, Apr. 2005, doi: 10.1016/j.neuroimage.2004.12.010.
[60] J. J. Chen and G. B. Pike, “Global cerebral oxidative metabolism during hypercapnia and
hypocapnia in humans: implications for BOLD fMRI,” Journal of Cerebral Blood Flow &
Metabolism, vol. 30, no. 6, pp. 1094–1099, 2010.
[61] R. L. Grubb Jr, M. E. Raichle, J. O. Eichling, and M. M. Ter-Pogossian, “The effects of
changes in PaCO2 cerebral blood volume, blood flow, and vascular mean transit time,” Stroke,
vol. 5, no. 5, pp. 630–639, 1974.
[62] M. J. Rosner, S. D. Rosner, and A. H. Johnson, “Cerebral perfusion pressure:
management protocol and clinical results,” Journal of neurosurgery, vol. 83, no. 6, pp. 949–962,
1995.
[63] F. Silvio Taccone, S. Scolletta, F. Franchi, K. Donadello, and M. Oddo, “Brain perfusion in
sepsis,” Current vascular pharmacology, vol. 11, no. 2, pp. 170–186, 2013.
[64] O. B. Paulson, S. Strandgaard, and L. Edvinsson, “Cerebral autoregulation,” Cerebrovasc
Brain Metab Rev, vol. 2, no. 2, pp. 161–192, Jan. 1990.
[65] S. Strandgaard, J. Olesen, E. Skinhøj, and N. A. Lassen, “Autoregulation of Brain
Circulation in Severe Arterial Hypertension,” Br Med J, vol. 1, no. 5852, pp. 507–510, Mar. 1973,
doi: 10.1136/bmj.1.5852.507.
[66] K. L. Gould, K. Lipscomb, and C. Calvert, “Compensatory changes of the distal coronary
vascular bed during progressive coronary constriction.,” Circulation, vol. 51, no. 6, pp. 1085–
1094, 1975.
[67] M. J. Joyner and D. P. Casey, “Regulation of increased blood flow (hyperemia) to
muscles during exercise: a hierarchy of competing physiological needs,” Physiological reviews,
2015.
[68] A. M. Bush et al., “Determinants of resting cerebral blood flow in sickle cell disease,”
American journal of hematology, vol. 91, no. 9, pp. 912–917, 2016.
[69] M. T. Borzage et al., “Predictors of cerebral blood flow in patients with and without
anemia,” Journal of Applied Physiology, vol. 120, no. 8, pp. 976–981, 2016.

77
[70] M. B. Vestergaard et al., “Acute hypoxia increases the cerebral metabolic rate–a
magnetic resonance imaging study,” Journal of cerebral Blood flow & metabolism, vol. 36, no. 6,
pp. 1046–1058, 2016.
[71] M. A. Mintun, B. N. Lundstrom, A. Z. Snyder, A. G. Vlassenko, G. L. Shulman, and M. E.
Raichle, “Blood flow and oxygen delivery to human brain during functional activity: theoretical
modeling and experimental data,” Proceedings of the National Academy of Sciences, vol. 98, no.
12, pp. 6859–6864, 2001.
[72] N. U. Ko et al., “Cerebral blood flow changes after endovascular treatment of
cerebrovascular stenoses,” American journal of neuroradiology, vol. 26, no. 3, pp. 538–542,
2005.
[73] H. White and B. Venkatesh, “Cerebral perfusion pressure in neurotrauma: a review,”
Anesthesia & Analgesia, vol. 107, no. 3, pp. 979–988, 2008.
[74] M. Smith, “Cerebral perfusion pressure,” BJA: British Journal of Anaesthesia, vol. 115,
no. 4. Oxford University Press, pp. 488–490, 2015.
[75] R. Aaslid, K.-F. Lindegaard, W. Sorteberg, and H. Nornes, “Cerebral autoregulation
dynamics in humans.,” Stroke, vol. 20, no. 1, pp. 45–52, 1989.
[76] A. H. Van Beek, J. A. Claassen, M. G. O. Rikkert, and R. W. Jansen, “Cerebral
autoregulation: an overview of current concepts and methodology with special focus on the
elderly,” Journal of Cerebral Blood Flow & Metabolism, vol. 28, no. 6, pp. 1071–1085, 2008.
[77] “Anemia - StatPearls - NCBI Bookshelf.” Accessed: Jun. 19, 2023. [Online]. Available:
https://www-ncbi-nlm-nih-gov.libproxy1.usc.edu/books/NBK499994/
[78] Y. Kuwabara et al., “Cerebral blood flow and vasodilatory capacity in anemia secondary
to chronic renal failure,” Kidney International, vol. 61, no. 2, pp. 564–569, Feb. 2002, doi:
10.1046/j.1523-1755.2002.00142.x.
[79] “Sickle Cell Disease: Continuous Arterial Spin-labeling Perfusion MR Imaging in Children
| Radiology.” Accessed: Jun. 20, 2023. [Online]. Available:
https://pubs.rsna.org/doi/abs/10.1148/radiol.2272020903
[80] J. J. Strouse et al., “Inverse correlation between cerebral blood flow measured by
continuous arterial spin-labeling (CASL) MRI and neurocognitive function in children with sickle
cell anemia (SCA),” Blood, vol. 108, no. 1, pp. 379–381, 2006.

78
[81] F. J. Kirkham et al., “Perfusion magnetic resonance abnormalities in patients with sickle
cell disease,” Annals of neurology, vol. 49, no. 4, pp. 477–485, 2001.
[82] A. Gupta et al., “Cerebrovascular reserve and stroke risk in patients with carotid stenosis
or occlusion: a systematic review and meta-analysis,” Stroke, vol. 43, no. 11, pp. 2884–2891,
2012.
[83] W. G. Schreiber, F. Gückel, P. Stritzke, P. Schmiedek, A. Schwartz, and G. Brix, “Cerebral
blood flow and cerebrovascular reserve capacity: estimation by dynamic magnetic resonance
imaging,” Journal of Cerebral Blood Flow & Metabolism, vol. 18, no. 10, pp. 1143–1156, 1998.
[84] I. Prohovnik, A. Hurlet-Jensen, R. Adams, D. De Vivo, and S. G. Pavlakis, “Hemodynamic
etiology of elevated flow velocity and stroke in sickle-cell disease,” Journal of Cerebral Blood
Flow & Metabolism, vol. 29, no. 4, pp. 803–810, 2009.
[85] M. M. Dowling et al., “Acute silent cerebral ischemia and infarction during acute anemia
in children with and without sickle cell disease,” Blood, The Journal of the American Society of
Hematology, vol. 120, no. 19, pp. 3891–3897, 2012.
[86] Y. Chai et al., “White matter has impaired resting oxygen delivery in sickle cell patients,”
American journal of hematology, vol. 94, no. 4, pp. 467–474, 2019.
[87] J. Shen et al., “Anemia Increases Oxygen Extraction Fraction in Deep Brain Structures but
Not in the Cerebral Cortex,” Frontiers in Physiology, vol. 13, 2022.
[88] S. Choi et al., “Anemia predicts lower white matter volume and cognitive performance
in sickle and non-sickle cell anemia syndrome,” American journal of hematology, vol. 94, no. 10,
pp. 1055–1065, 2019.
[89] S. Choi et al., “Hemoglobin and mean platelet volume predicts diffuse T1-MRI white
matter volume decrease in sickle cell disease patients,” NeuroImage: Clinical, vol. 15, pp. 239–
246, 2017.
[90] M. E. Fields et al., “Regional oxygen extraction predicts border zone vulnerability to
stroke in sickle cell disease,” Neurology, vol. 90, no. 13, pp. e1134–e1142, 2018.
[91] G. P. Rodgers, C. M. Clark, S. M. Larson, S. I. Rapoport, A. W. Nienhuis, and A. N.
Schechter, “Brain glucose metabolism in neurologically normal patients with sickle cell disease:
regional alterations,” Archives of neurology, vol. 45, no. 1, pp. 78–82, 1988.

79
[92] S. Herold et al., “Measurement of regional cerebral blood flow, blood volume and
oxygen metabolism in patients with sickle cell disease using positron emission tomography.,”
Stroke, vol. 17, no. 4, pp. 692–698, 1986.
[93] A. Bush et al., “Calibration of T2 oximetry MRI for subjects with sickle cell disease,”
Magnetic Resonance in Medicine, vol. 86, no. 2, pp. 1019–1028, 2021, doi:
10.1002/mrm.28757.
[94] C. P. Derdeyn et al., “Variability of cerebral blood volume and oxygen extraction: stages
of cerebral haemodynamic impairment revisited,” Brain, vol. 125, no. 3, pp. 595–607, 2002.
[95] W. J. Powers, “Cerebral hemodynamics in ischemic cerebrovascular disease,” Annals of
Neurology: Official Journal of the American Neurological Association and the Child Neurology
Society, vol. 29, no. 3, pp. 231–240, 1991.
[96] C. P. Derdeyn, R. L. Grubb, and W. J. Powers, “Cerebral hemodynamic impairment:
Methods of measurement and association with stroke risk,” Neurology, vol. 53, no. 2, pp. 251–
251, Jul. 1999, doi: 10.1212/WNL.53.2.251.
[97] C. P. Derdeyn, T. O. Videen, R. L. Grubb, and W. J. Powers, “Comparison of PET oxygen
extraction fraction methods for the prediction of stroke risk,” Journal of Nuclear Medicine, vol.
42, no. 8, pp. 1195–1197, 2001.
[98] D. Jiang et al., “Brain oxygen extraction is differentially altered by Alzheimer’s and
vascular diseases,” Journal of Magnetic Resonance Imaging, vol. 52, no. 6, pp. 1829–1837,
2020.
[99] B. P. Thomas et al., “Reduced global brain metabolism but maintained vascular function
in amnestic mild cognitive impairment,” Journal of Cerebral Blood Flow & Metabolism, vol. 37,
no. 4, pp. 1508–1516, 2017.
[100] R. L. Grubb Jr et al., “Importance of hemodynamic factors in the prognosis of
symptomatic carotid occlusion,” Jama, vol. 280, no. 12, pp. 1055–1060, 1998.
[101] A. Stadlbauer et al., “MR imaging–derived oxygen metabolism and neovascularization
characterization for grading and IDH gene mutation detection of gliomas,” Radiology, vol. 283,
no. 3, pp. 799–809, 2017.
[102] A. Jubran, “Pulse oximetry,” Critical care, vol. 3, pp. 1–7, 1999.
[103] J. E. Sinex, “Pulse oximetry: principles and limitations,” The American journal of
emergency medicine, vol. 17, no. 1, pp. 59–66, 1999.

80
[104] I. Yoshiya, Y. Shimada, and K. Tanaka, “Spectrophotometric monitoring of arterial
oxygen saturation in the fingertip,” Medical and Biological Engineering and Computing, vol. 18,
pp. 27–32, 1980.
[105] K. R. Thulborn, J. C. Waterton, P. M. Matthews, and G. K. Radda, “Oxygenation
dependence of the transverse relaxation time of water protons in whole blood at high field,”
Biochimica et Biophysica Acta (BBA)-General Subjects, vol. 714, no. 2, pp. 265–270, 1982.
[106] P. C. Van Zijl et al., “Quantitative assessment of blood flow, blood volume and blood
oxygenation effects in functional magnetic resonance imaging,” Nature medicine, vol. 4, pp.
159–167, 1998.
[107] W. Li and P. C. van Zijl, “Quantitative theory for the transverse relaxation time of blood
water,” NMR in Biomedicine, vol. 33, no. 5, p. e4207, 2020.
[108] G. A. Wright, B. S. Hu, and A. Macovski, “Estimating oxygen saturation of blood in vivo
with MR imaging at 1.5 T,” Journal of Magnetic Resonance Imaging, vol. 1, no. 3, pp. 275–283,
1991.
[109] X. Golay et al., “Measurement of tissue oxygen extraction ratios from venous blood T2:
increased precision and validation of principle,” Magnetic Resonance in Medicine: An Official
Journal of the International Society for Magnetic Resonance in Medicine, vol. 46, no. 2, pp. 282–
291, 2001.
[110] H. Lu, F. Xu, K. Grgac, P. Liu, Q. Qin, and P. Van Zijl, “Calibration and validation of TRUST
MRI for the estimation of cerebral blood oxygenation,” Magnetic resonance in medicine, vol.
67, no. 1, pp. 42–49, 2012.
[111] H. Lu and Y. Ge, “Quantitative evaluation of oxygenation in venous vessels using T2-
relaxation-under-spin-tagging MRI,” Magnetic Resonance in Medicine: An Official Journal of the
International Society for Magnetic Resonance in Medicine, vol. 60, no. 2, pp. 357–363, 2008.
[112] O. W. Schalm, N. C. Jain, E. J. Carroll, and others, Veterinary hematology. Lea & Febiger.,
1975.
[113] G. Benga and T. Borza, “Diffusional water permeability of mammalian red blood cells,”
Comparative Biochemistry and Physiology Part B: Biochemistry and Molecular Biology, vol. 112,
no. 4, pp. 653–659, 1995.
[114] D. Jiang et al., “Validation of T2-based oxygen extraction fraction measurement with
15O positron emission tomography,” Magnetic resonance in medicine, vol. 85, no. 1, pp. 290–
297, 2021.

81
[115] P. Liu, F. Xu, and H. Lu, “Test–retest reproducibility of a rapid method to measure brain
oxygen metabolism,” Magnetic resonance in medicine, vol. 69, no. 3, pp. 675–681, 2013.
[116] P. Liu et al., “Multisite evaluations of a T 2-relaxation-under-spin-tagging (TRUST) MRI
technique to measure brain oxygenation,” Magnetic resonance in medicine, vol. 75, no. 2, pp.
680–687, 2016.
[117] D. Jiang, P. Liu, Y. Li, D. Mao, C. Xu, and H. Lu, “Cross-vendor harmonization of T2-
relaxation-under-spin-tagging (TRUST) MRI for the assessment of cerebral venous oxygenation,”
Magnetic Resonance in Medicine, vol. 80, no. 3, pp. 1125–1131, 2018, doi:
10.1002/mrm.27080.
[118] F. Xu, P. Liu, J. J. Pekar, and H. Lu, “Does acute caffeine ingestion alter brain metabolism
in young adults?,” Neuroimage, vol. 110, pp. 39–47, 2015.
[119] F. Xu et al., “The influence of carbon dioxide on brain activity and metabolism in
conscious humans,” Journal of Cerebral Blood Flow & Metabolism, vol. 31, no. 1, pp. 58–67,
2011.
[120] F. Xu, P. Liu, J. M. Pascual, G. Xiao, and H. Lu, “Effect of hypoxia and hyperoxia on
cerebral blood flow, blood oxygenation, and oxidative metabolism,” Journal of Cerebral Blood
Flow & Metabolism, vol. 32, no. 10, pp. 1909–1918, 2012.
[121] D. Jiang et al., “Quantitative validation of MRI mapping of cerebral venous oxygenation
with direct blood sampling: A graded-O2 study in piglets,” Magnetic resonance in medicine, vol.
86, no. 3, pp. 1445–1453, 2021.
[122] A. Bush et al., “Empirical model of human blood transverse relaxation at 3 T improves
MRI T2 oximetry,” Magnetic resonance in medicine, vol. 77, no. 6, pp. 2364–2371, 2017.
[123] N. Nakano, J. Otsuka, and A. Tasaki, “Paramagnetic anisotropy measurements on a
single crystal of deoxyhemoglobin,” Biochimica et Biophysica Acta (BBA)-Protein Structure, vol.
278, no. 2, pp. 355–371, 1972.
[124] W. M. Spees, D. A. Yablonskiy, M. C. Oswood, and J. J. Ackerman, “Water proton MR
properties of human blood at 1.5 Tesla: Magnetic susceptibility, T1, T2, T, and non-Lorentzian
signal behavior,” Magnetic Resonance in Medicine: An Official Journal of the International
Society for Magnetic Resonance in Medicine, vol. 45, no. 4, pp. 533–542, 2001.
[125] J. M. Oja, J. S. Gillen, R. A. Kauppinen, M. Kraut, and P. C. Van Zijl, “Determination of
oxygen extraction ratios by magnetic resonance imaging,” Journal of Cerebral Blood Flow &
Metabolism, vol. 19, no. 12, pp. 1289–1295, 1999.

82
[126] R. M. Weisskoff and S. Kiihne, “MRI susceptometry: image-based measurement of
absolute susceptibility of MR contrast agents and human blood,” Magnetic resonance in
medicine, vol. 24, no. 2, pp. 375–383, 1992.
[127] C. Eldeniz et al., “Sickle hemoglobin vs normal hemoglobin: any changes in
susceptibility,” in Proceedings of the 25th Annual Meeting of International Society for Magnetic
Resonance in Medicine, 2017, pp. 22–27.
[128] C. Li et al., “Accuracy of the cylinder approximation for susceptometric measurement of
intravascular oxygen saturation,” Magnetic resonance in medicine, vol. 67, no. 3, pp. 808–813,
2012.
[129] F. W. Wehrli, A. P. Fan, Z. B. Rodgers, E. K. Englund, and M. C. Langham, “Susceptibilitybased time-resolved whole-organ and regional tissue oximetry,” NMR in biomedicine, vol. 30,
no. 4, p. e3495, 2017.
[130] E. M. Haacke et al., “In vivo measurement of blood oxygen saturation using magnetic
resonance imaging: a direct validation of the blood oxygen level-dependent concept in
functional brain imaging,” Human brain mapping, vol. 5, no. 5, pp. 341–346, 1997.
[131] V. Jain, M. C. Langham, and F. W. Wehrli, “MRI estimation of global brain oxygen
consumption rate,” Journal of Cerebral Blood Flow & Metabolism, vol. 30, no. 9, pp. 1598–1607,
2010.
[132] M. C. Langham, J. F. Magland, C. L. Epstein, T. F. Floyd, and F. W. Wehrli, “Accuracy and
precision of MR blood oximetry based on the long paramagnetic cylinder approximation of
large vessels,” Magnetic Resonance in Medicine: An Official Journal of the International Society
for Magnetic Resonance in Medicine, vol. 62, no. 2, pp. 333–340, 2009.
[133] R. Bhagwandien, M. Moerland, C. Bakker, R. Beersma, and J. Lagendijk, “Numerical
analysis of the magnetic field for arbitrary magnetic susceptibility distributions in 3D,” Magnetic
resonance imaging, vol. 12, no. 1, pp. 101–107, 1994.
[134] S. Barhoum et al., “Method for rapid MRI quantification of global cerebral metabolic
rate of oxygen,” Journal of Cerebral Blood Flow & Metabolism, vol. 35, no. 10, pp. 1616–1622,
2015.
[135] Z. B. Rodgers, V. Jain, E. K. Englund, M. C. Langham, and F. W. Wehrli, “High temporal
resolution MRI quantification of global cerebral metabolic rate of oxygen consumption in
response to apneic challenge,” Journal of Cerebral Blood Flow & Metabolism, vol. 33, no. 10, pp.
1514–1522, 2013.

83
[136] Z. B. Rodgers, E. K. Englund, M. C. Langham, J. F. Magland, and F. W. Wehrli, “Rapid T2-
and susceptometry-based CMRO2 quantification with interleaved TRUST (iTRUST),”
Neuroimage, vol. 106, pp. 441–450, 2015.
[137] M. C. Langham, J. F. Magland, T. F. Floyd, and F. W. Wehrli, “Retrospective correction for
induced magnetic field inhomogeneity in measurements of large-vessel hemoglobin oxygen
saturation by MR susceptometry,” Magnetic Resonance in Medicine: An Official Journal of the
International Society for Magnetic Resonance in Medicine, vol. 61, no. 3, pp. 626–633, 2009.
[138] X. Miao, K. S. Nayak, and J. C. Wood, “In vivo validation of T2-and susceptibility-based
SvO2 measurements with jugular vein catheterization under hypoxia and hypercapnia,”
Magnetic resonance in medicine, vol. 82, no. 6, pp. 2188–2198, 2019.
[139] Y. Wang and T. Liu, “Quantitative susceptibility mapping (QSM): decoding MRI data for a
tissue magnetic biomarker,” Magnetic resonance in medicine, vol. 73, no. 1, pp. 82–101, 2015.
[140] E. M. Haacke, S. Liu, S. Buch, W. Zheng, D. Wu, and Y. Ye, “Quantitative susceptibility
mapping: current status and future directions,” Magnetic resonance imaging, vol. 33, no. 1, pp.
1–25, 2015.
[141] F. Schweser, K. Sommer, A. Deistung, and J. R. Reichenbach, “Quantitative susceptibility
mapping for investigating subtle susceptibility variations in the human brain,” Neuroimage, vol.
62, no. 3, pp. 2083–2100, 2012.
[142] C. Langkammer et al., “Quantitative susceptibility mapping in multiple sclerosis,”
Radiology, vol. 267, no. 2, pp. 551–559, 2013.
[143] C. Langkammer et al., “Quantitative susceptibility mapping in Parkinson’s disease,” PLoS
One, vol. 11, no. 9, p. e0162460, 2016.
[144] J. Acosta-Cabronero, G. B. Williams, A. Cardenas-Blanco, R. J. Arnold, V. Lupson, and P. J.
Nestor, “In vivo quantitative susceptibility mapping (QSM) in Alzheimer’s disease,” PloS one,
vol. 8, no. 11, p. e81093, 2013.
[145] C. Langkammer et al., “Quantitative susceptibility mapping (QSM) as a means to
measure brain iron? A post mortem validation study,” Neuroimage, vol. 62, no. 3, pp. 1593–
1599, 2012.
[146] J. Zhang, T. Liu, A. Gupta, P. Spincemaille, T. D. Nguyen, and Y. Wang, “Quantitative
mapping of cerebral metabolic rate of oxygen (CMRO2) using quantitative susceptibility
mapping (QSM),” Magnetic resonance in medicine, vol. 74, no. 4, pp. 945–952, 2015.

84
[147] J. Zhang, D. Zhou, T. D. Nguyen, P. Spincemaille, A. Gupta, and Y. Wang, “Cerebral
metabolic rate of oxygen (CMRO2) mapping with hyperventilation challenge using quantitative
susceptibility mapping (QSM),” Magnetic resonance in medicine, vol. 77, no. 5, pp. 1762–1773,
2017.
[148] P. P. Ruetten, J. H. Gillard, and M. J. Graves, “Introduction to quantitative susceptibility
mapping and susceptibility weighted imaging,” The British Journal of Radiology, vol. 92, no.
1101, p. 20181016, 2019.
[149] T. Liu et al., “A novel background field removal method for MRI using projection onto
dipole fields,” NMR in Biomedicine, vol. 24, no. 9, pp. 1129–1136, 2011.
[150] F. Schweser, A. Deistung, B. W. Lehr, and J. R. Reichenbach, “Quantitative imaging of
intrinsic magnetic tissue properties using MRI signal phase: an approach to in vivo brain iron
metabolism?,” Neuroimage, vol. 54, no. 4, pp. 2789–2807, 2011.
[151] H. Sun and A. H. Wilman, “Background field removal using spherical mean value filtering
and Tikhonov regularization,” Magnetic resonance in medicine, vol. 71, no. 3, pp. 1151–1157,
2014.
[152] D. Zhou, T. Liu, P. Spincemaille, and Y. Wang, “Background field removal by solving the
Laplacian boundary value problem,” NMR in Biomedicine, vol. 27, no. 3, pp. 312–319, 2014.
[153] T. Liu, P. Spincemaille, L. De Rochefort, B. Kressler, and Y. Wang, “Calculation of
susceptibility through multiple orientation sampling (COSMOS): a method for conditioning the
inverse problem from measured magnetic field map to susceptibility source image in MRI,”
Magnetic Resonance in Medicine: An Official Journal of the International Society for Magnetic
Resonance in Medicine, vol. 61, no. 1, pp. 196–204, 2009.
[154] S. Wharton and R. Bowtell, “Whole-brain susceptibility mapping at high field: a
comparison of multiple-and single-orientation methods,” Neuroimage, vol. 53, no. 2, pp. 515–
525, 2010.
[155] S. Wharton, A. Schäfer, and R. Bowtell, “Susceptibility mapping in the human brain using
threshold-based k-space division,” Magnetic resonance in medicine, vol. 63, no. 5, pp. 1292–
1304, 2010.
[156] B. Bilgic, A. Pfefferbaum, T. Rohlfing, E. V. Sullivan, and E. Adalsteinsson, “MRI estimates
of brain iron concentration in normal aging using quantitative susceptibility mapping,”
Neuroimage, vol. 59, no. 3, pp. 2625–2635, 2012.

85
[157] T. Liu et al., “Morphology enabled dipole inversion (MEDI) from a single-angle
acquisition: comparison with COSMOS in human brain imaging,” Magnetic resonance in
medicine, vol. 66, no. 3, pp. 777–783, 2011.
[158] E. M. Haacke, J. Tang, J. Neelavalli, and Y. C. N. Cheng, “Susceptibility mapping as a
means to visualize veins and quantify oxygen saturation,” J Magn Reson Imaging, vol. 32, no. 3,
pp. 663–676, Sep. 2010, doi: 10.1002/jmri.22276.
[159] C.-Y. Hsieh, Y.-C. N. Cheng, J. Neelavalli, E. M. Haacke, and R. J. Stafford, “An improved
method for susceptibility and radius quantification of cylindrical objects from MRI,” Magnetic
resonance imaging, vol. 33, no. 4, pp. 420–436, 2015.
[160] C.-Y. Hsieh, Y.-C. N. Cheng, H. Xie, E. M. Haacke, and J. Neelavalli, “Susceptibility and size
quantification of small human veins from an MRI method,” Magnetic resonance imaging, vol.
33, no. 10, pp. 1191–1204, 2015.
[161] Y.-C. N. Cheng, C.-Y. Hsieh, J. Neelavalli, and E. M. Haacke, “Quantifying effective
magnetic moments of narrow cylindrical objects in MRI,” Physics in Medicine & Biology, vol. 54,
no. 22, p. 7025, 2009.
[162] Y.-C. N. Cheng, C.-Y. Hsieh, J. Neelavalli, Q. Liu, M. S. Dawood, and E. M. Haacke, “A
complex sum method of quantifying susceptibilities in cylindrical objects: the first step toward
quantitative diagnosis of small objects in MRI,” Magnetic resonance imaging, vol. 25, no. 8, pp.
1171–1180, 2007.
[163] D. A. Yablonskiy and E. M. Haacke, “Theory of NMR signal behavior in magnetically
inhomogeneous tissues: the static dephasing regime,” Magnetic resonance in medicine, vol. 32,
no. 6, pp. 749–763, 1994.
[164] D. A. Yablonskiy, “Quantitation of intrinsic magnetic susceptibility-related effects in a
tissue matrix. Phantom study,” Magnetic resonance in medicine, vol. 39, no. 3, pp. 417–428,
1998.
[165] H. An and W. Lin, “Impact of intravascular signal on quantitative measures of cerebral
oxygen extraction and blood volume under normo-and hypercapnic conditions using an
asymmetric spin echo approach,” Magnetic Resonance in Medicine: An Official Journal of the
International Society for Magnetic Resonance in Medicine, vol. 50, no. 4, pp. 708–716, 2003.
[166] H. An and W. Lin, “Cerebral oxygen extraction fraction and cerebral venous blood
volume measurements using MRI: effects of magnetic field variation,” Magnetic Resonance in
Medicine: An Official Journal of the International Society for Magnetic Resonance in Medicine,
vol. 47, no. 5, pp. 958–966, 2002.

86
[167] H. An and W. Lin, “Quantitative measurements of cerebral blood oxygen saturation
using magnetic resonance imaging,” Journal of Cerebral Blood Flow & Metabolism, vol. 20, no.
8, pp. 1225–1236, 2000.
[168] C. Wang et al., “Simultaneous dynamic R2 , and measurement using periodic π pulse
shifting multiecho asymmetric spin echo sequence moving estimation strategy: A feasibility
study for lower extremity muscle,” Magnetic Resonance in Medicine, vol. 77, no. 2, pp. 766–
773, 2017, doi: 10.1002/mrm.26126.
[169] C. P. Elder, R. N. Cook, M. A. Chance, E. A. Copenhaver, and B. M. Damon, “Image-based
calculation of perfusion and oxyhemoglobin saturation in skeletal muscle during submaximal
isometric contractions,” Magnetic Resonance in Medicine, vol. 64, no. 3, pp. 852–861, 2010.
[170] X. Zhang et al., “Feasibility of noninvasive quantitative measurements of intrarenal R2′
in humans using an asymmetric spin echo echo planar imaging sequence,” NMR in Biomedicine,
vol. 26, no. 1, pp. 91–97, 2013.
[171] K. P. Guilliams et al., “Red cell exchange transfusions lower cerebral blood flow and
oxygen extraction fraction in pediatric sickle cell anemia,” Blood, The Journal of the American
Society of Hematology, vol. 131, no. 9, pp. 1012–1021, 2018.
[172] H. An, Q. Liu, Y. Chen, and W. Lin, “Evaluation of MR-derived cerebral oxygen metabolic
index in experimental hyperoxic hypercapnia, hypoxia, and ischemia,” Stroke, vol. 40, no. 6, pp.
2165–2172, 2009.
[173] M. E. Fields et al., “Functional connectivity decreases with metabolic stress in sickle cell
disease,” Annals of neurology, vol. 88, no. 5, pp. 995–1008, 2020.
[174] V. Kiselev and S. Posse, “Analytical model of susceptibility-induced MR signal dephasing:
effect of diffusion in a microvascular network,” Magnetic Resonance in Medicine: An Official
Journal of the International Society for Magnetic Resonance in Medicine, vol. 41, no. 3, pp. 499–
509, 1999.
[175] A. J. Stone, N. C. Holland, A. J. Berman, and N. P. Blockley, “Simulations of the effect of
diffusion on asymmetric spin echo based quantitative BOLD: An investigation of the origin of
deoxygenated blood volume overestimation,” Neuroimage, vol. 201, p. 116035, 2019.
[176] X. He and D. A. Yablonskiy, “Quantitative BOLD: mapping of human cerebral
deoxygenated blood volume and oxygen extraction fraction: default state,” Magnetic
Resonance in Medicine: An Official Journal of the International Society for Magnetic Resonance
in Medicine, vol. 57, no. 1, pp. 115–126, 2007.

87
[177] X. He, M. Zhu, and D. A. Yablonskiy, “Validation of oxygen extraction fraction
measurement by qBOLD technique,” Magnetic Resonance in Medicine: An Official Journal of the
International Society for Magnetic Resonance in Medicine, vol. 60, no. 4, pp. 882–888, 2008.
[178] C. Z. Mooney, Monte carlo simulation. Sage, 1997.
[179] J. Geweke, “Monte Carlo simulation and numerical integration,” Handbook of
computational economics, vol. 1, pp. 731–800, 1996.
[180] J. L. Boxerman, L. M. Hamberg, B. R. Rosen, and R. M. Weisskoff, “MR contrast due to
intravascular magnetic susceptibility perturbations,” Magnetic resonance in medicine, vol. 34,
no. 4, pp. 555–566, 1995.
[181] M. Cercignani and D. C. Alexander, “Optimal acquisition schemes for in vivo quantitative
magnetization transfer MRI,” Magnetic Resonance in Medicine: An Official Journal of the
International Society for Magnetic Resonance in Medicine, vol. 56, no. 4, pp. 803–810, 2006.
[182] S. Ramani, D. S. Weller, J.-F. Nielsen, and J. A. Fessler, “Non-Cartesian MRI
reconstruction with automatic regularization via Monte-Carlo SURE,” IEEE transactions on
medical imaging, vol. 32, no. 8, pp. 1411–1422, 2013.
[183] A. K. Liu, J. W. Belliveau, and A. M. Dale, “Spatiotemporal imaging of human brain
activity using functional MRI constrained magnetoencephalography data: Monte Carlo
simulations,” Proceedings of the National Academy of Sciences, vol. 95, no. 15, pp. 8945–8950,
1998.
[184] M. G. Hall and D. C. Alexander, “Convergence and parameter choice for Monte-Carlo
simulations of diffusion MRI,” IEEE transactions on medical imaging, vol. 28, no. 9, pp. 1354–
1364, 2009.
[185] S. Sakadžić et al., “Large arteriolar component of oxygen delivery implies a safe margin
of oxygen supply to cerebral tissue,” Nature communications, vol. 5, no. 1, p. 5734, 2014.
[186] M. Sharan, M. Jones, R. Koehler, R. Traystman, and A. Popel, “A compartmental model
for oxygen transport in brain microcirculation,” Annals of biomedical engineering, vol. 17, pp.
13–38, 1989.
[187] W. C. Wang et al., “Multicenter comparison of magnetic resonance imaging and
transcranial Doppler ultrasonography in the evaluation of the central nervous system in
children with sickle cell disease,” Journal of pediatric hematology/oncology, vol. 22, no. 4, pp.
335–339, 2000.

88
[188] C. Muroi, N. Khan, D. Bellut, M. Fujioka, and Y. Yonekawa, “Extracranial–intracranial
bypass in atherosclerotic cerebrovascular disease: report of a single centre experience,” British
journal of neurosurgery, vol. 25, no. 3, pp. 357–362, 2011.
[189] M. Tanaka et al., “Chronic middle cerebral artery occlusion: a hemodynamic and
metabolic study with positron-emission tomography,” American journal of neuroradiology, vol.
29, no. 10, pp. 1841–1846, 2008.
[190] L. C. Jordan et al., “Non-invasive imaging of oxygen extraction fraction in adults with
sickle cell anaemia,” Brain, vol. 139, no. 3, pp. 738–750, 2016.
[191] M. R. Juttukonda et al., “Preliminary evidence for cerebral capillary shunting in adults
with sickle cell anemia,” Journal of Cerebral Blood Flow & Metabolism, vol. 39, no. 6, pp. 1099–
1110, 2019.
[192] M. N. Diringer, V. Aiyagari, A. R. Zazulia, T. O. Videen, and W. J. Powers, “Effect of
hyperoxia on cerebral metabolic rate for oxygen measured using positron emission tomography
in patients with acute severe head injury,” Journal of neurosurgery, vol. 106, no. 4, pp. 526–
529, 2007.
[193] M. Mintun, M. Raichle, W. Martin, and P. Herscovitch, “Brain oxygen utilization
measured with O-15 radiotracers and positron emission tomography.,” Journal of nuclear
medicine: official publication, Society of Nuclear Medicine, vol. 25, no. 2, pp. 177–187, 1984.
[194] M. Jenkinson, C. F. Beckmann, T. E. Behrens, M. W. Woolrich, and S. M. Smith, “Fsl,”
Neuroimage, vol. 62, no. 2, pp. 782–790, 2012.
[195] H. Lu, C. Clingman, X. Golay, and P. C. Van Zijl, “Determining the longitudinal relaxation
time (T1) of blood at 3.0 Tesla,” Magnetic Resonance in Medicine: an Official Journal of the
International Society for Magnetic Resonance in Medicine, vol. 52, no. 3, pp. 679–682, 2004.
[196] L. Václav\uu et al., “In vivo T1 of blood measurements in children with sickle cell disease
improve cerebral blood flow quantification from arterial spin-labeling MRI,” American Journal
of Neuroradiology, vol. 37, no. 9, pp. 1727–1732, 2016.
[197] A. M. Bush, T. D. Coates, and J. C. Wood, “Diminished cerebral oxygen extraction and
metabolic rate in sickle cell disease using T2 relaxation under spin tagging MRI,” Magnetic
resonance in medicine, vol. 80, no. 1, pp. 294–303, 2018.
[198] W. Li et al., “Quantification of whole-brain oxygenation extraction fraction and cerebral
metabolic rate of oxygen consumption in adults with sickle cell anemia using individual T2-

89
based oxygenation calibrations,” Magnetic resonance in medicine, vol. 83, no. 3, pp. 1066–
1080, 2020.
[199] K. Wasserman and B. J. Whipp, “Exercise physiology in health and disease,” American
Review of Respiratory Disease, vol. 112, no. 2, pp. 219–249, 1975.
[200] P. L. Croal, J. Leung, C. L. Phillips, M. G. Serafin, and A. Kassner, “Quantification of
pathophysiological alterations in venous oxygen saturation: a comparison of global MR
susceptometry techniques,” Magnetic Resonance Imaging, vol. 58, pp. 18–23, 2019.
[201] L. Afzali-Hashemi et al., “Impairment of cerebrovascular hemodynamics in patients with
severe and milder forms of sickle cell disease,” Frontiers in physiology, vol. 12, p. 430, 2021.
[202] A. Bush et al., “Pseudo continuous arterial spin labeling quantification in anemic
subjects with hyperemic cerebral blood flow,” Magnetic resonance imaging, vol. 47, pp. 137–
146, 2018.
[203] J. C. Wood, “Brain O2 reserve in sickle cell disease,” Blood, The Journal of the American
Society of Hematology, vol. 133, no. 22, pp. 2356–2358, 2019.
[204] P. D. Kosinski et al., “The severity of anaemia depletes cerebrovascular dilatory reserve
in children with sickle cell disease: a quantitative magnetic resonance imaging study,” British
Journal of Haematology, vol. 176, no. 2, pp. 280–287, 2017.
[205] R. Harb, C. Whiteus, C. Freitas, and J. Grutzendler, “In vivo imaging of cerebral
microvascular plasticity from birth to death,” Journal of Cerebral Blood Flow & Metabolism, vol.
33, no. 1, pp. 146–156, 2013.
[206] J. Duffin, “Fail-safe aspects of oxygen supply,” The Journal of Physiology, vol. 598, no. 21,
pp. 4859–4867, 2020.
[207] S. Herold et al., “Measurement of regional cerebral blood flow, blood volume and
oxygen metabolism in patients with sickle cell disease using positron emission tomography.,”
Stroke, vol. 17, no. 4, pp. 692–698, 1986.
[208] V. Jain, O. Abdulmalik, K. J. Propert, and F. W. Wehrli, “Investigating the magnetic
susceptibility properties of fresh human blood for noninvasive oxygen saturation
quantification,” Magnetic resonance in medicine, vol. 68, no. 3, pp. 863–867, 2012.
[209] P. Liu et al., “T1 and T2 values of human neonatal blood at 3 Tesla: dependence on
hematocrit, oxygenation, and temperature,” Magnetic resonance in medicine, vol. 75, no. 4, pp.
1730–1735, 2016.

90
[210] W. Li et al., “A method for estimating and removing streaking artifacts in quantitative
susceptibility mapping,” Neuroimage, vol. 108, pp. 111–122, 2015.
[211] L. C. Krishnamurthy, P. Liu, Y. Ge, and H. Lu, “Vessel-specific quantification of blood
oxygenation with T2-relaxation-under-phase-contrast MRI,” Magnetic resonance in medicine,
vol. 71, no. 3, pp. 978–989, 2014.
[212] Y. He et al., “Regional coherence changes in the early stages of Alzheimer’s disease: a
combined structural and resting-state functional MRI study,” Neuroimage, vol. 35, no. 2, pp.
488–500, 2007.
[213] M. R. Juttukonda et al., “Reduced oxygen extraction efficiency in sickle cell anemia
patients with evidence of cerebral capillary shunting,” Journal of Cerebral Blood Flow &
Metabolism, vol. 41, no. 3, pp. 546–560, 2021.
[214] L. Václav\uu et al., “Cerebral oxygen metabolism in adults with sickle cell disease,”
American journal of hematology, vol. 95, no. 4, pp. 401–412, 2020.
[215] R. Boushel, H. Langberg, J. Olesen, J. Gonzales-Alonzo, J. Bülow, and M. Kjaer,
“Monitoring tissue oxygen availability with near infrared spectroscopy (NIRS) in health and
disease,” Scandinavian journal of medicine & science in sports, vol. 11, no. 4, pp. 213–222,
2001.
[216] A. Villringer, J. Planck, C. Hock, L. Schleinkofer, and U. Dirnagl, “Near infrared
spectroscopy (NIRS): a new tool to study hemodynamic changes during activation of brain
function in human adults,” Neuroscience letters, vol. 154, no. 1–2, pp. 101–104, 1993.
[217] S. N. Jespersen and L. Østergaard, “The roles of cerebral blood flow, capillary transit
time heterogeneity, and oxygen tension in brain oxygenation and metabolism,” Journal of
cerebral blood flow & metabolism, vol. 32, no. 2, pp. 264–277, 2012.
[218] S. Sakadžić et al., “Two-photon high-resolution measurement of partial pressure of
oxygen in cerebral vasculature and tissue,” Nature methods, vol. 7, no. 9, pp. 755–759, 2010.
[219] P. N. Ainslie et al., “Stability of cerebral metabolism and substrate availability in humans
during hypoxia and hyperoxia,” Clinical Science, vol. 126, no. 9, pp. 661–670, 2014.
[220] J. A. Fisher and D. J. Mikulis, “Cerebrovascular reactivity: Purpose, optimizing methods,
and limitations to interpretation–a personal 20-year odyssey of (re) searching,” Frontiers in
Physiology, vol. 12, p. 629651, 2021.

91
[221] K. P. A. Baas, B. F. Coolen, E. T. Petersen, B. J. Biemond, G. J. Strijkers, and A. J.
Nederveen, “Comparative Analysis of Blood T2 Values Measured by T2 -TRIR and TRUST,” J
Magn Reson Imaging, Jan. 2022, doi: 10.1002/jmri.28066.
[222] V. E. Griffeth and R. B. Buxton, “A theoretical framework for estimating cerebral oxygen
metabolism changes using the calibrated-BOLD method: modeling the effects of blood volume
distribution, hematocrit, oxygen extraction fraction, and tissue signal properties on the BOLD
signal,” Neuroimage, vol. 58, no. 1, pp. 198–212, 2011.
[223] A. G. Tsai, P. C. Johnson, and M. Intaglietta, “Oxygen gradients in the microcirculation,”
Physiological reviews, vol. 83, no. 3, pp. 933–963, 2003.
[224] V. Jain, M. C. Langham, T. F. Floyd, G. Jain, J. F. Magland, and F. W. Wehrli, “Rapid
magnetic resonance measurement of global cerebral metabolic rate of oxygen consumption in
humans during rest and hypercapnia,” Journal of Cerebral Blood Flow & Metabolism, vol. 31,
no. 7, pp. 1504–1512, 2011.
[225] J. Helenius et al., “Cerebral hemodynamics in a healthy population measured by
dynamic susceptibility contrast MR imaging,” Acta radiologica, vol. 44, no. 5, pp. 538–546,
2003.
[226] H. Herzog et al., “Quantitation of regional cerebral blood flow with 15O-butanol and
positron emission tomography in humans,” Journal of Cerebral Blood Flow & Metabolism, vol.
16, no. 4, pp. 645–649, 1996.
[227] C. Eldeniz et al., “Bulk volume susceptibility difference between deoxyhemoglobin and
oxyhemoglobin for HbA and HbS: A comparative study,” Magnetic resonance in medicine, vol.
85, no. 6, pp. 3383–3393, 2021.
[228] K. M. Schmainda et al., “Dynamic-susceptibility contrast agent MRI measures of relative
cerebral blood volume predict response to bevacizumab in recurrent high-grade glioma,”
Neuro-oncology, vol. 16, no. 6, pp. 880–888, 2014.
[229] P. Schellinger et al., “Evidence-based guideline: The role of diffusion and perfusion MRI
for the diagnosis of acute ischemic stroke [RETIRED]: Report of the Therapeutics and
Technology Assessment Subcommittee of the American Academy of Neurology,” Neurology,
vol. 75, no. 2, pp. 177–185, 2010.
[230] G. Quan et al., “Role of dynamic susceptibility contrast perfusion MRI in glioma
progression evaluation,” Journal of Oncology, vol. 2021, 2021.

92
[231] S. Gahramanov et al., “Potential for differentiation of pseudoprogression from true
tumor progression with dynamic susceptibility-weighted contrast-enhanced magnetic
resonance imaging using ferumoxytol vs. gadoteridol: a pilot study,” International Journal of
Radiation Oncology* Biology* Physics, vol. 79, no. 2, pp. 514–523, 2011.
[232] D. J. Wang et al., “The value of arterial spin-labeled perfusion imaging in acute ischemic
stroke: comparison with dynamic susceptibility contrast-enhanced MRI,” Stroke, vol. 43, no. 4,
pp. 1018–1024, 2012.
[233] S. F. Eskildsen et al., “Increased cortical capillary transit time heterogeneity in
Alzheimer’s disease: a DSC-MRI perfusion study,” Neurobiology of aging, vol. 50, pp. 107–118,
2017.
[234] T. Reiter et al., “Minimizing risk of nephrogenic systemic fibrosis in cardiovascular
magnetic resonance,” Journal of Cardiovascular Magnetic Resonance, vol. 14, no. 1, pp. 1–11,
2012.
[235] T. Christen, D. S. Bolar, and G. Zaharchuk, “Imaging brain oxygenation with MRI using
blood oxygenation approaches: methods, validation, and clinical applications,” American
journal of neuroradiology, vol. 34, no. 6, pp. 1113–1123, 2013.
[236] M. E. MacDonald, A. J. Berman, E. L. Mazerolle, R. J. Williams, and G. B. Pike, “Modeling
hyperoxia-induced BOLD signal dynamics to estimate cerebral blood flow, volume and mean
transit time,” Neuroimage, vol. 178, pp. 461–474, 2018.
[237] C. Vu et al., “Quantitative perfusion mapping with induced transient hypoxia using BOLD
MRI,” Magnetic resonance in medicine, vol. 85, no. 1, pp. 168–181, 2021.
[238] S. Ogawa, T.-M. Lee, A. R. Kay, and D. W. Tank, “Brain magnetic resonance imaging with
contrast dependent on blood oxygenation.,” proceedings of the National Academy of Sciences,
vol. 87, no. 24, pp. 9868–9872, 1990.
[239] J. Poublanc et al., “Perfusion MRI using endogenous deoxyhemoglobin as a contrast
agent: Preliminary data,” Magnetic resonance in medicine, vol. 86, no. 6, pp. 3012–3021, 2021.
[240] E. S. Sayin et al., “Investigations of hypoxia-induced deoxyhemoglobin as a contrast
agent for cerebral perfusion imaging,” Human Brain Mapping, vol. 44, no. 3, pp. 1019–1029,
2023.
[241] J. B. Schulman et al., “DSC MRI in the human brain using deoxyhemoglobin and
gadolinium—Simulations and validations at 3T,” Frontiers in Neuroimaging, vol. 2, p. 1048652,
2023.

93
[242] A. D. Harris et al., “Cerebral blood flow response to acute hypoxic hypoxia,” NMR in
Biomedicine, vol. 26, no. 12, pp. 1844–1852, 2013.
[243] B. Kjølby, L. Østergaard, and V. Kiselev, “Theoretical model of intravascular
paramagnetic tracers effect on tissue relaxation,” Magnetic Resonance in Medicine: An Official
Journal of the International Society for Magnetic Resonance in Medicine, vol. 56, no. 1, pp. 187–
197, 2006.
[244] C. Vu et al., “Sinusoidal CO2 respiratory challenge for concurrent perfusion and
cerebrovascular reactivity MRI,” Frontiers in Physiology, vol. 14, p. 1102983, 2023.
[245] C. Vu, J. Shen, S. Choi, K. Baas, and J. Wood, “Sinusoidal oxygen respiratory paradigm for
perfusion MRI”.

Abstract (if available)

Abstract Sickle cell disease (SCD) manifests as a hereditary disorder characterized by anomalous hemoglobin that undergoes polymerization in response to deoxygenation, resulting in the formation of inflexible, sickle-shaped red blood cells. The recurring sickling of red blood cells contributes to the development of anemia and vasculopathy, with the most severe repercussions observed in cerebral tissues. Notably, silent cerebral infarction (SCI) emerges as a prevalent and progressive complication in SCD, presenting substantial associations with heightened stroke risks and neurocognitive impairments. Despite these associations, the precise prediction and underlying pathogenesis of SCI remain enigmatic. This thesis endeavors to establish a cerebral oxygenation marker for the prospective prediction of SCI and concurrently explores the correlation between brain oxygen extraction fraction (OEF), SCD, and SCI. In addition, Monte Carlo simulation is applied to demonstrate the details regarding microvascular architecture and blood oxygenation. The logical framework of this research aims to enhance our comprehension of the intricate interplay between these factors within the context of sickle cell disease.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Susceptibility-weighted MRI for the evaluation of brain oxygenation and brain iron in sickle cell disease

PDF

Quantitative MRI for oxygenation imaging in cerebrovascular diseases

PDF

Assessing cerebral oxygen supply and demand in sickle cell anemia using MRI

PDF

Estimation of cognitive brain activity in sickle cell disease using functional near-infrared spectroscopy and dynamic systems modeling

PDF

Photoplethysmogram-based biomarker for assessing risk of vaso-occlusive crisis in sickle cell disease: machine learning approaches

PDF

Role of the autonomic nervous system in vasoconstriction responses to mental stress in sickle cell disease: a bioengineering perspective

PDF

Modeling of cardiovascular autonomic control in sickle cell disease

PDF

Cerebrovascular disease of white matter in patients with chronic anemia syndrome

PDF

Clinical significance of oxyhemoglobin dissociation curves on the interpretation of pulse oximetry in Sickle Cell disease

PDF

A novel role for hypoxia-inducible factor-1alpha (HIF-1alpha) in the regulation of inflammatory chemokines and leukotriene expression in sickle cell disease

PDF

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

PDF

Essays in health economics

PDF

Engineering a post-infarct human myocardium on a chip to reveal oxygen-dependent crosstalk in cardiac fibroblasts and myocytes

PDF

The role of miRNA and its regulation in pulmonary hypertension in sickle cell disease

PDF

Insights from a computational approach to natural killer cell activation

PDF

Anemia and white matter: a diffusion MRI analysis of the human brain

PDF

Cerebral oxygen saturation and cerebrovascular instability in newborn infants with congenital heart disease compared to healthy controls

PDF

Development of a quantitative proteomic molecular imaging platform using multiplexed nanoscale Raman scattering contrast agents

PDF

Mechanistic investigation of pro-angiogenic signaling and endothelial sprouting mediated by FGF and VEGF

PDF

Oxygen therapy for the treatment of retinal ischemia

Asset Metadata

Creator Shen, Jian (author)

Core Title Quantitative MRI for the measurement of cerebral oxygen extraction fraction in sickle cell disease

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Biomedical Engineering

Degree Conferral Date 2024-05

Publication Date 01/25/2024

Defense Date 12/18/2023

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

Tag magnetic resonance imaging,Monte Carlo simulation,OAI-PMH Harvest,oxygen extraction fraction,sickle cell disease

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Wood, John (committee chair), Khoo, Michael (committee member), Wang, Danny (committee member)

Creator Email brian.huaian@gmail.com,jianshen@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113810137

Unique identifier UC113810137

Identifier etd-ShenJian-12625.pdf (filename)

Legacy Identifier etd-ShenJian-12625

Document Type Dissertation

Format theses (aat)

Rights Shen, Jian

Internet Media Type application/pdf

Type texts

Source 20240125-usctheses-batch-1122 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name University of Southern California Digital Library

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

Repository Email cisadmin@lib.usc.edu