by
for the degree of Doctor of Philosophy
at the University of Hong Kong.
August 2012
submitted by
Zhou Yuwen
at The University of Hong Kong
in August 2012
Benefiting from higher SNR as well as better spatial, temporal and spectral
resolution, magnetic resonance imaging (MRI) at high field has proved to be a
valuable neuroimaging modality which provides comprehensive evaluation of the
central nervous system non-invasively. The objectives of this doctoral work were
to develop MRI methodologies and to assess  the functional, metabolic and
structural alterations in rodent brains under normal and manipulated conditions.
Firstly, to improve the functional sensitivity and spatial precision, a novel
functional MRI (fMRI) method using balanced steady state free precession with
intravascular susceptibility contrast agent was proposed and its feasibility was
evaluated in rat visual system. This new approach was sensitized to cerebral blood
8/17/2019 FullText (39)
volume (CBV) changes. It provided comparable sensitivity to conventional CBV-
weighted fMRI using echo planar imaging but with no severe image distortion and
signal dropout. Robust negative responses during visual stimulation were
observed and activation patterns were in excellent agreement with known
neuroanatomy. As a promising alternative to conventional CBV-weighted fMRI,
it was particularly suited for fMRI investigation of animal models at high field.
Secondly, the relationship between anatomical connections and resting-state
fMRI connectivity was explored using a well-controlled animal model of corpus
callosotomy. Both complete and partial callosotomy resulted in significant loss of
interhemispheric connectivity in the cortical areas whose primary
interhemispheric connections via corpus callosum (CC) were severed. Partial
restoration of interhemispheric connectivity and increased intrahemispheric
connectivity were also observed. The experimental findings directly supported
that anatomical connections via CC play a primary and indispensable role in
resting-state connectivity, and that resting-state networks could be dynamically
reorganized or acquired directly or indirectly through the remaining anatomical
connections.
stimulation and endogenous modification, respectively. Significantly lower
hippocampal N-acetylaspartate (NAA) was observed in fear conditioning animals,
indicating reduced neuronal dysfunction and/or integrity, which contributed to the
trauma-related symptoms. Meanwhile, pregnant animals exhibited prominently
higher hippocampal NAA level, reflecting the increased density of neurons in this
8/17/2019 FullText (39)
region, which might facilitate supporting behaviors that involving learning and
memory. The 1H MRS detection of ongoing neurochemical changes induced by
fear conditioning and pregnancy, especially in the hippocampus, can shed light on
the mechanisms of learning and memory and the neurochemical underpinnings of
behavioral improvement in pregnant animals.
Lastly, manganese-enhanced MRI (MEMRI) was employed to investigate
the hypoxic-ischemic (HI) injury in the late phase and the neural response to
conditioned fear. Significantly higher enhancement in T1-weighted images was
found in the peri-lesional region 24 hours after manganese administration and it
colocalized with the increase in glial cell density in histological staining,
demonstrating the existence of reactive gliosis in the late phase after HI injury. In
fear conditioned animals, higher manganese uptake was observed in amygdala,
hippocampus, paraventricular nucleus of hypothalamus and cingulate cortex,
which were all highly-involved in the process of fear. These findings suggested
MEMRI approach were useful in investigation of post-injury cellular events and
functional reorganization as well as for in vivo mapping of neuronal activity.
(500 words)
I
DECLARATION
I declare that the thesis represents my own work, except where due
acknowledgement is made, and that is it has not been previously included in a
thesis, dissertation or report submitted to this University or to any other institution
for a degree, diploma or other qualifications.
Signed ……………………………………………………
II
ACKNOWLEDGEMENTS
It is an honor to express my sincere thanks to all of those who supported and
helped me in any way during my doctoral study. First and foremost, I owe my
deepest gratitude to my supervisor, Prof. Ed Wu, for his guidance, encouragement,
and support throughout the years. I truly admire him for being an inspiring mentor
and an insightful scientist with enthusiasm. He generously shares his immense
knowledge, skills and experiences, which have benefited me both professionally
and personally. I am also grateful for the time and availability on discussion and
consultation whenever I encounter problems. Talking with him always motivates
me, broadens my vision and refreshes my mind. It is my privilege to be his
student and my growth would not be so rewarding without him.
I am indebted to many of my colleagues from the Laboratory of Biomedical
Imaging and Signal Processing, who have supported and helped me over the years.
My gratitude goes to Dr. Kevin Chan, Dr. Matthew Cheung and Dr. Condon Lau
for the knowledgeable advice and inspiring discussions. Special thanks to Miss
Abby Ding and Miss Shujuan Fan for their technical contribution towards the
work that has been included in this thesis and for being supportive friends.
Deepest thanks to Dr. Kexia Cai, Dr. Kannie Chan, Dr. April Chow and Mr. Frank
Lee for their friendship, care and encouragement. Many thanks to Dr. Hua Guo,
Dr. Edward Hui, Dr. Jerry Cheung, Dr. Li Xiao, Dr. Zhongwei Qiao, Mr. Peng
Cao, Mr. Kyle Xing, Miss Darwin Gao, Miss Joe Cheng, Mr. Jevin Zhang, Mr.
8/17/2019 FullText (39)
III
Russell Chan, Mr. Patrick Gao, Mr. Leon Ho, Miss Anna Wang and Mr. Victor
Xie for their generous supports and assistance. I would also like to thank Prof.
Kwok-Fai So, Prof. Grainne McAlonan, Prof. Yong Hu, Dr. Yu-Xiang Liang and
Dr. Qi Li at The University of Hong Kong for their collaboration and helpful
discussions.
It is a pleasure to acknowledge The University of Hong Kong for offering
me a postgraduate studentship throughout my study (2008-2012). I am also
grateful to the conference grants that enabled me to participate in several
international conferences, especially the International Society for Magnetic
Resonance in Medicine Annual Meetings (2010-2012), which widened my
horizon and inspired my work.
I would also like to thank my fiancé, Alvin Yang and my dear friends for
their company and tremendous supports throughout my 4-year Ph.D. study.
Lastly, and most importantly, I would like to express my deepest gratitude
to my parents, Meizhen Chen and Renming Zhou. They bore me, raised me,
supported me, taught me, and loved me. To them I dedicate this thesis.
8/17/2019 FullText (39)
CHAPTER 1 INTRODUCTION .......................................................................... 1 
1.1.1  General Overview .................................................................................. 1 
1.1.4 MRI Techniques ...................................................................................... 4 
1.2 Functional Magnetic Resonance Imaging.......................................................... 5 
1.2.2 Resting-state functional MRI ................................................................... 7 
1.3 Magnetic Resonance Spectroscopy ................................................................... 8 
1.4 Manganese-enhance Magnetic Resonance Imaging .......................................... 9 
1.5 Thesis outline and contribution........................................................................ 10 
WITH INTRAVASCULAR SUSCEPTIBILITY CONTRAST AGENT ....... 16 
2.1 Introduction ...................................................................................................... 16 
8/17/2019 FullText (39)
CALLOSOTOMY ............................................................................................... 41 
3.4 Discussion ........................................................................................................ 59  
3.4.2 The Role of Callosal Connections in Interhemispheric Connectivity ... 62 
3.4.3 Post-callosotomy Plasticity of Interhemispheric and Intrahemispheric Connectivity ......................................................................................... 64 
3.4.4 Limitations and Future Work................................................................. 65 
ALTERED NEUROCHEMICAL PROFILES ................................................. 69 
4.1 Proton MRS Reveals N-acetylaspartate Reduction in Hippocampus and Cingulate Cortex after Fear Conditioning ............................................................. 69 
4.1.1 Introduction ............................................................................................ 69 
4.1.3 Results .................................................................................................... 75 
4.1.4 Discussion .............................................................................................. 79 
4.1.5 Conclusion ............................................................................................. 84 
4.2 Proton MRS Reveals Regional Metabolic Changes in Rat Brain during Pregnancy and Motherhood ................................................................................... 85 
4.2.1 Introduction ............................................................................................ 85 
CHAPTER 5 MANGANESE ENHANCED MAGNETIC RESONANCE
IMAGING OF MORPHOLOGICAL AND FUNCTIONAL BRAIN
CHANGES ............................................................................................................ 96 
5.1 MEMRI Study of Neonatal Hypoxic-ischemic Injury in the Late Stage ......... 96 
5.1.1 Introduction ............................................................................................ 96 
5.1.3 Results .................................................................................................. 100 
5.2 In Vivo Manganese-enhanced MRI of Conditioned Fear Response ............. 105 
5.2.1 Introduction .......................................................................................... 105 
5.2.2 Methods ............................................................................................... 106 
8/17/2019 FullText (39)
CHAPTER 6 GENERAL CONCLUSIONS AND FUTURE STUDIES ...... 115 
REFERENCES .................................................................................................. 119 
LIST OF FIGURES
Figure 1.1 A schematic diagram showing the source of BOLD signal in detecting neural activity. 6 
Figure 2.1 Typical images acquired with conventional T2-weighted spin-echo (a), conventional
T2*-weighted gradient-echo (b), bSSFP (c), GE-EPI (d) and SE-EPI (e) at the identical
slice location from a rat brain before and after intravenous injection of intravascular
susceptibility contrast agent MION (15 mg/kg). Note that all bSSFP, GE-EPI and SE-
EPI images were acquired with the same spatial resolution and matrix size (64×64) and
reconstructed without any post-processing. ................................................... .............. 25 
Figure 2.2 Post-MION bSSFP fMRI yielded good agreement between the activation patterns and
known neuroanatomy during unilateral 20-s block-design stimulations in a normal
adult SD rats (a). The activation map was computed by correlating the fMRI time-
course with the stimulation paradigm and overlaid on the post-MION bSSFP image.
Only the voxels with cross-correlation coefficient (cc) < -0.35 were color coded. For
each activation cluster, the time-courses from the single voxel with the strongest cc
value (as marked by the yellow crosses in the small inset) (b) and all activated voxels
within the ROI (as defined by the green lines in the small inset) (c) were plotted in
mean ± SD. These time-courses were computed by averaging across all blocks, trials
and animals studied (n = 9). The strongest activation was observed in the superficial
layers of contralateral SC. Note that bSSFP images were acquired in a 64×64 matrix
and reconstructed to 128×128 by zero padding. The shaded area indicates the
stimulation period. ....................................................................................................... 26 
Figure 2.3 Typical activation patterns observed by post-MION bSSFP (a), GE-EPI (b) and SE-
EPI (c) fMRI methods from an animal. The imaging slice was located at Bregma -
7.2mm as shown in the coronal T2W image (d). For each method, the time-course
8/17/2019 FullText (39)
IX
depicts the raw (colored) and low-pass filtered (black) signal changes in the voxel with
the strongest cc value in the SC (indicated by the green squares in the activation maps).
 ..................................................................................................................................... 28 
Figure 2.4  Average post-MION bSSFP (a), GE-EPI (b) and SE-EPI (c) time-courses in all
activated voxels (cc < -0.35) within the entire slice in the animals studied (n = 5). The
shaded areas indicate the stimulation periods. ......................................................... .... 31 
Figure 2.5 Average post-MION bSSFP signal time-course from the entire brain during the 4-min
hypercapnia challenge using 5% CO2 inhalation in the animals studied (n = 3).
Measurement ROIs were shown in the small inset, and covered all cortical and
subcortical regions except those severely darkened by MION because of large blood
vessels. The shaded area identifies the period of hypercapnia challenge. .................... 32 
Figure 2.6 Apparent tissue longitudinal relaxation rate R1 (=1/ T1) (a), transverse relaxation rate
R2 (=1/T2) (b) and bSSFP signal (c) as a function of MION concentration in different
brain regions of one animal. ........................................................ ................................. 33 
Figure 3.1 Representative T2-weighted (T2W) images and fractional anisotropy (FA) maps from
the animals with complete (a), anterior partial (b) and posterior partial (c) corpus
callosotomy and sham surgery (d). The transected part of the corpus callosum (CC) is
indicated in red color in the sagittal planes (left panel) and by yellow arrows in the
T2W images. The blue lines indicate the corresponding locations of T2W and FA slices
in the right panel. Distance to Bregma for each slice is given at the bottom. .............. 49 
Figure 3.2  Typical resting-state connectivity maps from individual animals with complete (a),
anterior partial (b) and posterior partial (c) callosotomy and sham surgery (d) at post-
surgery day 7. Independent component analysis (ICA) was performed. Spatial ICA
maps of independent components were scaled to z-scores (z>1.5) and overlaid on the
8/17/2019 FullText (39)
X
EPI images. The color bars display z-scores with a higher z-score (yellow)
representing a stronger correlation between the time course of that voxel and the mean
time course of this component. The CC is organized in a rostrocaudal topographical
manner with anterior fibers connecting frontal areas of the two hemispheres and
posterior fibers connecting caudal cortical structures. The components shown in this
figure correspond to five brain areas ranging from the anterior to posterior part of the
brain. They are caudate putamen (CPu), secondary somatosensory cortex (S2), primary
somatosensory cortex (S1), auditory cortex (AC) and visual cortex (VC), respectively.
 ..................................................................................................................................... 51 
Figure 3.3 Localization of the seeds and corresponding regions of interest (ROIs) for seed-based
cross-correlation analysis (SBA). Five brain areas were selected based on the ICA
results. Seeds and ROIs were overlaid on EPI images with slices located from Bregma
1.2mm to -7.2mm (slice 1-9). Color code: CPu (red), S2 (yellow), S1 (green), AC
(purple) and VC (blue). ..................................................... ........................................... 53 
Figure 3.4 Typical histograms showing the distribution of numbers of voxels within the ROIs (as
illustrated in Figure 3.3) across all correlation coefficient (cc) values. The data was
from the SBA results of one sham animal at post-surgery day 7. For each brain area,
the results of the seeds in both left (in red) and right (in blue) hemispheres and the
ROIs ipsilateral (a) and contralateral (b) to the seeds are presented here. ................... 54 
Figure 3.5 Scatter plots of the mean value µ of the cc distribution in S2 and AC ROIs ipsilateral (a)
and contralateral (b) to the seeds for all animals at post-surgery day 7. ...................... 54 
Figure 3.6  Typical resting-state connectivity maps from individual animals with complete (a),
anterior partial (b) and posterior partial (c) callosotomy and sham surgery (d) at post-
surgery day 28. Spatial ICA maps of independent components were scaled to z-scores
8/17/2019 FullText (39)
XI
(z>1.5) and overlaid on the EPI images. The color bars display z-scores with a higher
z-score (yellow) representing a stronger correlation between the time course of that
voxel and the mean time course of this component. The ICA components shown in this
figure correspond to secondary somatosensory cortex and auditory cortex. ................ 57 
Figure 4.1 Freezing responses measured during the initial 6 minutes (pre-shock, free exploring)
and the following 6 minutes (fear conditioning) of training session as well as
contextual and cued test sessions, respectively. One-way ANOVA followed by
Bonferroni multiple comparison post-test was performed with * P < 0.05, ** P < 0.01,
*** P < 0.001. Data were presented as mean ± standard deviation. ............................ 76 
Figure 4.2 Typical localization of voxel of interest (VOI) in the hippocampus (a), cingulate cortex
(b) and thalamus (c) (solid-line boxes) on coronal and axial slices of T2-weighted
images for proton magnetic resonance spectroscopy measurements (L-left; R-right; A-
anterior; P-posterior). The size of VOI in the left hippocampus, the cingulate cortex
and the left thalamus was 1.2×2.5×1.6 mm3, 1.2×1.5×2.5 mm3 and 2×2×2 mm3,
respectively. .................................................. ........................................................... .... 77 
Figure 4.3 Representative raw spectra (black) along with QUEST fitting (red) of the VOIs in the
hippocampus, cingulate cortex and thalamus, respectively. The spectra were acquired
from the same mouse before fear conditioning. Residuals of QUEST quantitation are
shown in the top entry. Abbreviations: NAA, N-acetylaspartate; Glu, glutamate; Cr,
creatine; Cho, choline; Tau, taurine; m-Ins, myo-inositol. .......................................... 78 
Figure 4.4 A schematic diagram summarizing the timeline of experimental procedures. 1H MRS
measurements were performed on pregnant primiparous rats (N=15) at 3 days before
mating (Baseline), gestational day 17 (G17), lactation day 7 (L7) and post-weaning
day 7 (PW7). Age-matched nulliparous female rats (N=9) which served as non-
8/17/2019 FullText (39)
XII
pregnant controls were examined at the same timepoints with the pregnant rats. ....... 88 
Figure 4.5 Typical in vivo 1H MRS spectra of voxel of interest (VOI) in the hippocampus (a) and
thalamus (b) (solid-line boxes) on coronal slices of T2-weighted images for 1H MRS
measurement (L-left; R-right) from the same rat. ........................................................ 91 
Figure 4.6 Comparisons of metabolite to Cr ratios between the pregnant and non-pregnant rats in
the hippocampus (left column) and thalamus (right column) at each timepoint. Mean ±
SD presented. Unpaired t-tests were performed with * P<0.05, ** P <0.01. .............. 92 
Figure 5.1  A schematic diagram showing the timeline of experimental procedures. Animals in
Group 1 (N=6) were subjected to hypoxic-ischemic (HI) insult at postnatal day 7 while
animals from Group 2 (N=6) were served as controls. Manganese-enhanced MRI
(MEMRI) measurements were performed prior to Mn2+ administration and at 1, 7 days
after the injection. ........................................................................................................ 99 
Figure 5.2 Representative T1W images of Group 1 (HI) and Group 2 (normal control) before and
after Mn2+ administration. The yellow lines indicate the manually defined regions of
interest that were used for comparisons of signal intensity. ....................................... 100 
Figure 5.3  Percentage change maps (from pre-injection) at day 1 and day 7 computed from
coregistered images, directly illustrating the significant SI increase in peri-lesional
region at day 1 (white arrows) and relatively slow clearance in contra-lesional
thalamus (black arrow). ............................................................................................. 101 
Figure 5.4 Mean ± SD illustrates SI changes in different regions after Mn2+ administration in
peri-lesional area: ipsi-lateral thalamus (top left), ipsi-lateral cingulate cortex (top right)
and contra-lateral thalamus (bottom). All the post-injection SI was normalized by the
pre-injection SI within the same animal, respectively, before calculating mean ± SD.
8/17/2019 FullText (39)
XIII
T-tests between Group 1: HI (N=6) and Group 2: normal control (N=6), *P<0.05, ** P 
<0.01 and ++ P=0.09. ................................................................................................ 102 
Figure 5.5 GFAP staining, illustrating the hyper cellular density of astrocytes in the peri-lesional
region in thalamus (middle column) and cingulate cortex (right column) as compared
to the contra-lesional hemisphere (left column). ...................................................... .. 103 
Figure 5.6  Schematic diagrams showing the timeline of experimental procedures (a), the fear
conditioning paradigm (b) and setup (c). ................................................................... 108 
Figure 5.7  Freezing response and locomotor activity measured during the initial habituation
period and the following fear conditioning (FC) period. Significantly increased
freezing duration and decreased locomotor activity (expressed as total distance
traveled in cm) confirmed that the mice acquired associative learning with the electric
footshock. Two-tailed Student’s t-tests were performed between the habituation period
and fear conditioning period: * P<0.05 ** P <0.005. ................................................ 109 
Figure 5.8 Averaged post-Mn T1W images from Group 1 (FC) and Group 2 (control) with the
ratio maps showing the percentage signal differences between them. Enhanced Mn-
uptake was observed in amygdala (yellow arrows), hippocampus (red arrows),
paraventricular nucleus of hypothalamus (green arrows) and cingulate cortex (purple
arrows) in FC animals. ...................................................... ......................................... 110 
Figure 5.9  T1W signal intensity changes (Mean ± SD) before and after Mn injection were
compared between the two groups using ROIs covering amygdala (Amyg),
hippocampus (Hip), paraventricular nucleus of hypothalamus (PVH), cingulate cortex
(Cg) and the entire brain. Two-tailed t-test was performed with * P < 0.05, ** P <0.01
 ................................................................................................................................... 111 
H MRS ...................................................................... 8 
Table 2.1 Comparison of post-MION bSSFP, GE-EPI and SE-EPI fMRI methods as determined
from activated voxels (cc < -0.35) within the entire slice in all five animals studied. ... 30 
Table 3.1 Mean value µ and standard deviation σ of the cc values within the ROIs in hemispheres
ipsilateral and contralateral to the seeds at post-surgery day 7. The results are presented
in mean ± standard deviation. Statistical comparisons between different groups were
performed using one-way ANOVA. ↓ and ↑ denote decrease and increase, respectively,
with the significance level indicated by †, †† (P<0.05, P<0.01 compared to anterior
partial transection, respectively), §, §§ (P<0.05, P<0.01 compared to posterior partial
transection, respectively) and *, **, *** (P<0.05, P<0.01, P<0.005 compared to sham
control, respectively). .................................................................................................... 55 
Table 3.2 Mean value µ and standard deviation σ of the cc values within the ROIs in hemispheres
ipsilateral and contralateral to the seeds at post-surgery day 28. The results are
presented in mean ± standard deviation. Statistical comparisons between different
groups was performed using one-way ANOVA. ↓ and ↑ denote decrease and increase,
respectively, with the significance level indicated by †, †† (P<0.05, P<0.01 compared
to anterior partial transection, respectively), §, §§, §§§ (P<0.05, P<0.01, P<0.005
compared to posterior partial transection, respectively) and *, **, *** (P<0.05, P<0.01,
P<0.005 compared to sham control, respectively). .................................................... .... 58 
Table 4.1 Metabolite to Cr ratios and corresponding Cramér-Rao lower bounds (CRLBs) at 1 day
before, 1 day and 1 week after the fear conditioning experiment in the hippocampus,
cingulate cortex and thalamus, respectively. Data from all animals studied ( N   = 12)
8/17/2019 FullText (39)
8/17/2019 FullText (39)
1H MRS proton magnetic resonance spectroscopy
2D two-dimensional
3D three-dimensional
bSSFP balanced steady state free precession
CBV cerebral blood volume
CO2  carbon dioxide
CPu caudate putamen
FOV field of view
GE gradient echo
m-Ins myo-inositol
MRI magnetic resonance imaging
RARE rapid acquisition with relaxation enhancement
8/17/2019 FullText (39)
SI signal intensity
SpO2 oxygen saturation
T2*W T2*-weighted
tSNR temporal signal to noise ratio
US unconditioned stimulus
VC visual cortex
δ  duration of diffusion gradient pulse
  diffusion time
Magnetic resonance imaging (MRI) is one of the most significant
developments in medical imaging in the twentieth century. Although the physical
phenomenon of nuclear magnetic resonance (NMR) has been known since the
early 1940s (1, 2), it has not been applied to the field of medical imaging until
1973 when Paul C. Lauterbur made the first NMR image by introducing gradients
in the magnetic field (3). In 1974 Peter Mansfield presented the mathematical
theory for rapid imaging and image reconstruction, which were found much
needed in clinical practice. Since then, though only developed within decades,
MRI has become one of the most clinically used diagnosis and assessment tools,
largely due to its non-invasive nature and its sensitivity to a broad range of tissue
properties. In addition to routine clinical diagnosis, MRI is widely used for in vivo 
biomedical research. Due to its excellent capability of soft-tissue characterization,
MRI offers great promises in neurological imaging, especially in understanding
the brain, both its structure and its functions. Today, MRI has evolved into an
extremely versatile modality for evaluating many different parameters of
anatomical, physiological and metabolic interest.
In parallel to the rapid growth of MR techniques, there is a drive towards
8/17/2019 FullText (39)
2
high-field MRI. The principal advantage of MRI at high field is the increase in
signal to noise ratio, contrast to noise ratio and spectral resolution. Such increases
have been largely demonstrated experimentally, which lead to significant
improvement in both diagnostic imaging and biomedical research.
1.1.2 Basic Principles of MRI
The proton possesses a property called spin which can be considered as a
small magnetic field. Under normal circumstances these tiny magnets are
randomly distributed in space and thus the net magnetic vector is zero. When an
external magnetic field B0  is present, the spin nuclei align parallel or antiparallel
to the external field B0. The parallel orientation is the low energy state while
antiparallel orientation is the high-energy state.  The number of spins in the low
energy state excesses the number in the high-energy state. A net magnetization is
produced by the difference between the two populations and thus the tissue placed
in the magnetic field becomes magnetized.  Nuclei do not line up with the
magnetic field but wobble or precess around the direction of the external field.
The Larmor equation gives the relationship between the strength of a magnetic
field, B0, and the precessional frequency, ω0, of an individual spin:
00  Bγ  =  
The hydrogen nucleus contains one proton and possesses a significant
8/17/2019 FullText (39)
magnetic moment. Hydrogen is extremely abundant in biological tissues because
hydrogen nuclei exist in water and fat molecules. The tissue (hydrogen) will be
magnetized in a large external magnetic field, preparing it for the MR imaging
process.
When a radiofrequency (RF) energy pulse is oriented perpendicular to B0 at
the Larmor frequency, the individual spins begin to precess in phase, as will the
net magnetization vector. Magnetic resonance occurs when some of the spins in
the lower energy state absorb energy from the RF field and make a transition into
the higher energy state. This has the effect of tipping the net magnetization away
from B0 at the flip angle and toward the transverse plane.  As the transverse
magnetization precesses through the receiver coil, a current or a signal is induced
in the coil. When the RF pulse is discontinued, the signal in the coil begins at a
given amplitude (determined by the amount of magnetization precessing in the
transverse plane and the precessional frequency) and fades away rapidly. This
initial signal is referred to as the free induction decay or FID. Spatial information
was resolved by applying slice selection, frequency encoding and phase encoding.
Fourier transform is then performed to convert the detected signals and
reconstruct the image in the spatial domain.
1.1.3 Image Contrast in MRI
In MRI, image contrast is determined by the imaging protocols and the tissue
properties, including proton density (PD), longitudinal relaxation time (T1) and
transverse relaxation time (T2). PD refers to the amount of protons that could
8/17/2019 FullText (39)
4
contribute to the MR signal within a voxel. T1 is defined as the time required for
the longitudinal magnetization to return to the equilibrium state after a RF pulse.
T2 reflects the time required for the transverse magnetization to decay to 1/e of its
original amplitude after disturbed by a RF pulse. However, in the presence of the
inhomogeneity in the external magnetic field, the time required for the transverse
magnetization to decay is named as effective transverse relaxation time (T2*),
which is usually smaller than T2.
The image contrast of MRI also depends on the selecting of pulse sequences
and imaging parameters. Variations in the value of TR (repetition time), TE (echo
time) and flip angle can affect the image contrast characteristics. For example, if
short values of TR and TE are used, images will exhibit T1  contrast while long
values of TR and TE result in a T2-dependent contrast. Middle TR values and
middle TE values are common for density weighted contrast.
One may also manipulates the MR image contrast by the use of exogenous
contrast agents, such as paramagnetic ions (e.g. Mn2+) and ferromagnetic particles
(e.g. iron oxide). In Chapter 2 and 5, these contrast agents have been used to
enhance the contrast and improve the sensitivity of detection.
1.1.4 MRI Techniques
Since its introduction, MRI has become one of the fastest expanding and
most active modality in the imaging field. There are a large number of MRI
8/17/2019 FullText (39)
5
techniques, each of which provides a particular type of information about the
subject tissues. According to the tissues properties they sensitized to, the MRI
techniques for brain imaging can be categorized into three main types, including
functional, metabolic and anatomical MRI. Functional MRI measures blood
oxygenation level dependent (BOLD) signal, cerebral blood flow and cerebral
blood volume, providing information of neuronal activity and tissue perfusion.
Metabolic MRI mainly refers to magnetic resonance spectroscopy which provides
information about chemical composition of the tissues and changes in chemical
composition. Anatomical MRI includes basic T1-, T2-, PD- and T2*-weighted
MRI scans as well as diffusion MRI, which measures the diffusion characteristics
of the water molecules in brain tissues and the connectivity of white matter axons
in the central nervous system. MRI can add valuable information by multi-
parametric and longitudinal assessments of the functional, metabolic and
structural changes in the central nervous system (CNS).
1.2 Functional Magnetic Resonance Imaging
1.2.1 Blood oxygenation level dependent contrast
Functional magnetic resonance imaging (fMRI) can be used to map the
neuronal activity by using blood oxygenation level dependent (BOLD) contrast
(4). The BOLD mechanism refers to the phenomenon that increases in neural
activity cause changes in the MR signal via T2* changes (Figure 1.1). Increased
neural activity leads to an increased demand for oxygen. The vascular system
8/17/2019 FullText (39)
overcompensates for this demand, resulting in the larger amount of
oxyhemoglobin relative to deoxyhemoglobin. Deoxyhemoglobin [dHb], as an
endogenous paramagnetic contrast, attenuates the MR signal. BOLD fMRI takes
advantage of the magnetic susceptibility difference of oxyhemoglobin and
deoxyhemoglobin and measures the neuronal responses indirectly via
hemodynamic changes. It has shown better correlation with synaptic and
postsynaptic activity than the spiking activity (5).
Figure 1.1 A schematic diagram showing the source of BOLD signal in detecting neural
activity.
Based on the rapid echo planar imaging (EPI) sequence, fMRI provides a
good compromise between spatial and temporal resolution. However, severe
image distortion and signal dropout at air/tissue interface due to the susceptibility
and field inhomogeneities deteriorates the quality of EPI images (6). The other
major limiting factor in EPI fMRI is the constraint on achievable spatial resolution.
At high field, EPI spatial resolution is limited intrinsically by T2* that can be short
and vary across image due to field inhomogeneity (7, 8). Furthermore, several
8/17/2019 FullText (39)
physiological factors can also adversely affect EPI-based fMRI, including motion
artifacts and noises from cardiac and respiratory pulsations (9). These limitations
must be addressed to achieve accurate and high resolution mapping of brain
activities.
1.2.2 Resting-state functional MRI
Since its introduction of BOLD contrast (10), BOLD fMRI has been widely
applied to study the functions and connectivity of the CNS due to its
noninvasiveness, large field-of-view and 3D imaging capabilities. The majority of
fMRI studies have focused on studying neuronal activities associated with stimuli
or tasks. It is not until recently that investigating the resting brain by fMRI
became of immense interest. The motivations mainly arise from two aspects. First,
most of the brain’s energy is consumed at rest by spontaneous neuronal activity
(20% of body’s energy) while the task-related increases in energy metabolism are
usually small (<5%) (11). Second, low-frequency fluctuations (LFFs) (<0.1 Hz) of
resting-state fMRI (rsfMRI) signals were found to be coherent among brain areas
with similar functions and known anatomical interconnections (12, 13). Therefore,
efforts have been made to examine the coherence in LFFs, or resting-state
connectivity, providing not only new insights into the functional organization of
the brain (14-16), but also a better understanding of brain functional plasticity
during disease, aging and learning (17-19). Despite the wide interest in mapping
resting-state connectivity, the underlying physiological mechanisms remain to be
fully understood, and thus hinders the interpretation of rsfMRI data.
8/17/2019 FullText (39)
Magnetic resonance spectroscopy (MRS) is one of the few available
techniques that can assess the neurochemistry non-invasively. Though MRS can
be performed using a variety of nuclei such as carbon (13C), nitrogen (15N),
fluorine (19F), and sodium (23Na), the nuclei phosphorus (31P) and hydrogen (1H),
which have high concentrations in vivo, are more frequently used. Proton (1H)
MRS has become popular due to the high natural abundance of protons and their
high absolute sensitivity to magnetic manipulation and better spectral resolution
(20). In the brain, the principle molecules that can be detected by 1H MRS are
listed in Table 1.1 with chemical shifts and physiological significance (21).
Table 1.1 Principal metabolites observed in 1H MRS
Metabolites Chemical
shift (ppm)
Physiological significance
m-Ins Myo-inositol 3.6 Glial marker and cerebral osmolyte Tau Taurine 3.4 Pediatric tumors Cho Choline 3.2 Cell membrane metabolism marker Cr PCr
Creatine, Phosphocreatine
Glx GABA, Glutamate, Glutamine
2.1-2.5 Intracellular neurotransmitter marker
Ala Alanine 1.5 Meningioma, Abscess
Lac Lactate 1.3 Anaerobic glycolysis Lip Free lipids 0.9, 1.4 Tissue necrosis
The current impetus for higher field strengths benefits 1H MRS in terms of
better SNR and increased spectral dispersion, favoring the detection of many
overlapping resonances. Chapter 4 has demonstrated the usage of 1H MRS at 7T
8/17/2019 FullText (39)
to monitor longitudinal neurochemical alterations.
1.4 Manganese-enhance Magnetic Resonance Imaging
As previously mentioned, tissue T1, T2 or T2* can be manipulated by the use
of MR contrast agents, and thus differentiates the targeted cells from surrounding
tissues. Paramagnetic ion Mn2+  is one of the most widely used positive MRI
contrast agents. After the administration of manganese, the MRI signal intensity is
altered due to more changes in T1 than T2, thus providing increased signal in T1-
weighted MRI. Based on the increasing number of applications, manganese-
enhance MRI (MEMRI) has proved to be a valuable tool for molecular imaging
(22). There are three major types of applications using MEMRI. First, systemic
administration, such as intraperitoneal, intravenous or subcutaneous injection of
Mn2+ ,   can be used for enhanced visualization of the cerebral neuroarchitecture,
providing unique contrast in specific areas of the brain (23-27). Second, Mn2+ has
shown the capability of tracing neuronal pathways in an anterograde manner when
injected to specific brain regions. Therefore, MEMRI has been increasingly used
to map neuronal tracts in the living brain (28-30). Third, due to the factor that
Mn2+  can enter synaptically activated neurons through voltage-gated calcium
channels, MEMRI has been also used to map brain activities (23, 31-33). This
technique is also referred to as activation-induced MEMRI (AIM-MRI) (34, 35).
Comparing to fMRI, MEMRI maps the neuronal activations independently of the
surrogate hemodynamic responses. MEMRI also provides higher SNR and spatial
8/17/2019 FullText (39)
10
resolution. However, the major limitation of using Mn2+  is its cellular toxicity.
Therefore, it is critical to achieve desired contrast with as low a dose as possible.
1.5 Thesis outline and contribution
This thesis focuses on the development and applications of several MRI
methods, for in vivo assessments of the function, metabolism and structure of the
rodent brains at 7 Tesla. Briefly, a new fMRI technique was introduced and
demonstrated on rat brain in Chapter 2. The mechanism of resting-state fMRI was
explored in Chapter 3. Chapter 4 used proton MRS to investigate the
neurochemical alterations in mouse brain elicited by exogenous stimulation and
endogenous modification, respectively. In Chapter 5, manganese-enhanced MRI
was applied to investigate the cellular alterations after brain injury and neural
responses to conditioned fear. The organization of this thesis is described as
follows:
In Chapter 2, a new CBV-weighted fMRI technique using distortion-free
balanced steady-state free precession (bSSFP) sequence was proposed and its
feasibility was investigated in rat brain at 7 Tesla. After intravascular
susceptibility contrast agent administration (MION at 15 mg/kg), unilateral visual
stimulation was presented using a block-design paradigm. With TR/TE = 3.8/1.9
ms and α=18o, bSSFP fMRI was performed and compared with the conventional
CBV-weighted fMRI using post-MION GE- and SE-EPI. The results showed that
post-MION bSSFP fMRI provides comparable sensitivity but with no severe
8/17/2019 FullText (39)
image distortion and signal dropout. Robust negative responses were observed
during stimulation and activation patterns were in excellent agreement with
known neuroanatomy. Furthermore, the post-MION bSSFP signal was observed
to decrease significantly during hypercapnia challenge, indicating its sensitivity to
CBV changes. These findings demonstrated that post-MION bSSFP fMRI is a
promising alternative to conventional CBV-weighted fMRI. This technique is
particularly suited for fMRI investigation of animal models at high field. This
work was published in Magnetic Resonance in Medicine 68(1):65-73.
In Chapter 3, an experimental model of corpus callosotomy was employed to
investigate the relationship between anatomical connections and resting-state
fMRI connectivity. Complete, anterior and posterior mid-sagittal corpus callosum
(CC) transections were performed in normal adult Sprague-Dawley rats. Resting-
state fMRI was performed in these animals and sham controls at post-surgery day
7 and day 28. Five resting-state networks, including caudate putamen (CPu),
secondary somatosensory cortex (S2), primary somatosensory cortex (S1),
auditory cortex (AC) and visual cortex (VC) were examined using both
independent component analysis and seed-based analysis. Complete callosotomy
resulted in loss of interhemispheric connectivity in all cortical areas examined,
including S2, S1, AC and VC, at day 7 and day 28. Partial callosotomy led to
significantly decreased interhemispheric connectivity at day 7 in the cortical areas
whose primary interhemispheric connections via CC were severed, namely S2 and
S1 after anterior transection and S1, AC and VC after posterior transection. At
8/17/2019 FullText (39)
12
day 28, some of these connectivity reductions were restored. In addition,
intrahemispheric connectivity was found to generally increase in areas where
interhemispheric connectivity reductions sustained at day 28 after partial and
complete callosotomy. These experimental findings directly support that
anatomical connections via CC play a primary and indispensable role in resting-
state connectivity, and that resting-state networks can be dynamically reorganized
or acquired directly or indirectly through the remaining anatomical connections.
This work provides direct evidence of the relationship between anatomical and
functional connectivity, contributing to a better understanding of the biological
mechanisms of rsfMRI. This work was collaborated with Prof. KF So and Dr.
Yuxiang Liang in Department of Anatomy in The University of Hong Kong who
prepared the animal model and performed the surgery.
In Chapter 4, proton MRS was employed to monitor the longitudinal
metabolic alterations in animal brains elicited by exogenous stimulation and
endogenous modification, respectively. First, longitudinal neurochemical changes
underlying fear conditioning was characterized by 1H MRS, aiming to contribute
towards a clear understanding of the neurobiological mechanisms of fear learning
and memory. The fear conditioning in rodents provides a valuable translational
tool to investigate the neural basis of learning and memory and potentially the
neurobiology of post-traumatic stress disorder (PTSD). Neurobiological changes
induced by fear conditioning have largely been examined ex vivo  while
progressive ‘real-time’ changes in vivo remain underexplored. Single voxel proton
magnetic resonance spectroscopy of the hippocampus, cingulate cortex and
8/17/2019 FullText (39)
13
thalamus of adult male C57BL/6N mice (N=12) was performed at 1 day before, 1
day and 1 week after, fear conditioning training using a 7T scanner. N-
acetylaspartate (NAA), a marker for neuronal integrity and viability, significantly
decreased in the hippocampus at 1 day and 1 week post-conditioning. Significant
NAA reduction was also observed in the cingulate cortex at 1 day post-
conditioning. These findings of hippocampal NAA decrease indicate reduced
neuronal dysfunction and/or neuronal integrity, contributing to the trauma-related
PTSD-like symptoms. The neurochemical changes characterized by 1H MRS can
shed light on the biochemical mechanisms of learning and memory. Moreover,
such information can potentially facilitate prompt intervention for patients with
psychiatric disorders. This work was collaborated with Prof. GM McAlonan and
Dr. Qi Li in Department of Psychiatry in The University of Hong Kong who
provided the experimental setup for fear conditioning. This work has been
submitted to Psychiatry Research: Neuroimaging and it is under revision at the
time this thesis is submitted. Second, longitudinal 1H MRS during pregnancy and
motherhood was performed to evaluate the regional metabolic changes in the
hippocampus and thalamus of maternal brain. Pregnant primiparous rats (N=15)
were studied at 3 days before mating, gestational day 17, lactation day 7 and post-
weaning day 7. Age-matched nulliparous female rats (N=9) served as non-
pregnant controls. Single voxel 1H MRS of the hippocampus and thalamus was
performed using a 7T scanner. Significantly higher NAA level observed in the
hippocampus of late pregnant rats level of hippocampal NAA of pregnant rats
indicates the increased density of neurons in this region, facilitating supporting
8/17/2019 FullText (39)
14
behaviors that involving learning and memory. Reduced level of choline in the
maternal brain reflects high fetal demands. The raised lactate level in thalamus
might be related to the hyperventilation of pregnancy. The results of this study
provide neurochemical evidence of the behavioral changes associated with
pregnancy.
In Chapter 5, manganese-enhanced MRI was applied to investigate the
cellular alterations after neonatal hypoxic-ischemic injury and after fear
conditioning training, respectively. First, in vivo  MEMRI was employed to
investigate the hypoxic-ischemic injury in the late phase. Mn2+  induced signal
changes were examined using SPM coregistration and ROI analysis. T1W images
SI increase was detected in the peri-lesional region 24 hours after Mn2+
administration and it colocalized with the increase in glial cell density in GFAP
staining, demonstrating the existence of reactive gliosis in the late phase after H-I
injury. The results suggest such MEMRI approach may be useful in investigation
of post-injury cellular events and functional reorganization. Second, to study the
neurocircuitry behind this paradigm, in vivo MEMRI was employed to investigate
the neural response after subjection to fear conditioning in mice. Fear
conditioning is a widely used procedure to study the neural basis of learning and
memory. Compared to controls, fear conditioned animals exhibited higher Mn
uptake in amygdala, hippocampus, paraventricular nucleus of hypothalamus and
cingulate cortex, which are all highly-involved in the process of fear. The results
provide insights to neurocircuitry involved in fear-conditioning and consolidate
8/17/2019 FullText (39)
15
the capability of MEMRI as an in vivo probe for mapping neural activity.
In Chapter 6, potential applications and expansion of the investigated MRI
methods for future studies were discussed.
8/17/2019 FullText (39)
FREE PRECESSION FMRI WITH INTRAVASCULAR SUSCEPTIBILITY
CONTRAST AGENT
2.1 Introduction
Since the introduction of blood oxygenation level dependent (BOLD)
contrast, fMRI has assumed an invaluable role in mapping brain functions due to
its noninvasiveness, large field-of-view and 3D imaging capabilities compared
with other imaging modalities (10). fMRI is often conducted with echo planar
imaging (EPI), which goes through the entire frequency space from one single
shot, providing a good compromise between spatial and temporal resolution. With
the availability of high field scanners, fMRI has also been expanded from humans
to small animal models for various neuroscience applications (36). High field
improves fMRI by increasing signal to noise ratio (SNR) and sensitivity (37).
However, it also increases the susceptibility and field inhomogeneities that give
rise to severe image distortion and signal dropout at air/tissue interface in EPI
images (6). The other major limiting factor in EPI fMRI is the constraint on
achievable spatial resolution. At high field, EPI spatial resolution is limited
intrinsically by T2* that can be short and vary across image due to field
inhomogeneity (7, 8). Furthermore, several physiological factors can also
adversely affect EPI-based fMRI, including motion artifacts and noises from
8/17/2019 FullText (39)
17
cardiac and respiratory pulsations (9). These limitations must be addressed to
achieve accurate and high resolution mapping of brain activities.
Balanced steady-state free precession (bSSFP) imaging has been proposed as
a promising alternative to EPI for fMRI (38, 39). bSSFP is a technique that uses
fully balanced gradients in each repetition time. The image contrast of bSSFP is
determined by T2 /T1 (40). Its short repetition time and readout duration, together
with high signal efficiency, allow the fast, distortion-free and high resolution
imaging that is highly desirable for functional imaging. The original bSSFP fMRI
studies using the steep magnitude/phase transition in the bSSFP off-resonance
profile (38, 39, 41) were limited by the need for multi-frequency acquisitions to
find the narrow range transition band. Later, it was demonstrated that the
functional contrast can be achieved by utilizing the relatively large flat portion of
the bSSFP profile (42-45). The contrast mechanism of this “pass-band bSSFP” is
similar to that of conventional spin-echo BOLD but is less sensitive to motion and
physiological noises (40, 44, 46).
Signal contrast-to-noise ratio (CNR) measured during activation is another
key factor in fMRI. Using an intravascular susceptibility contrast agent of long
blood half-life such as monocrystalline iron oxide nanoparticles (MIONs) or other
ultrasmall superparamagnetic iron oxides (USPIOs), larger CNR and more robust
signal changes can be achieved (47-49). The negative post-contrast fMRI signal
changes primarily reflect the increase in cerebral blood volume (CBV) that
reduces the apparent T2  and T2* significantly (50). Signals from large blood
8/17/2019 FullText (39)
vessels are effectively suppressed in CBV-weighted images because of the
relatively high intravascular concentration of contrast agent (51). However, the
strong susceptibility contrast agent causes further image distortion and severe
signal loss in highly vascularized regions, which worsen at high field.
In this study, we investigated the feasibility of bSSFP imaging in
combination with intravascular susceptibility contrast agent MION as a new fMRI
approach. Brain responses to unilateral visual stimulation were examined at 7T
using this approach. The results were compared with those obtained using post-
MION GE- and SE-EPI fMRI. To support our hypothesis that this approach is
primarily sensitized to CBV changes, we qualitatively examined the post-MION
bSSFP signal changes during hypercapnia. Apparent tissue T1  and T2  changes
were also measured at varying MION concentrations to assess the post-MION
bSSFP signal properties.
2.2.1 Animal Procedure
All experiments were approved by the Institutional Animal Care and Use
Committee. Normal adult Sprague–Dawley rats (220~250g) were initially
anesthetized with 3% isoflurane and then maintained with 1.5% isoflurane during
the left femoral vein catheterization surgery. Animals were then placed on a
plastic cradle and maintained with 1% isoflurane during imaging. To minimize
8/17/2019 FullText (39)
19
head motion, the head was fixed with a tooth bar and plastic screws were inserted
into the ear canals. Animals were kept warm by a circulating water pad at 37oC.
Respiration rate, heart rate, oxygenation saturation and rectal temperature were
continuously monitored and maintained within normal physiological ranges (52,
53).
Before fMRI experiments, MION (MGH Center of Molecular Imaging
Research, MA) with approximately 4-hr blood half-life (54, 55) was intravenously
injected via the femoral vein catheter at a dosage of 15 Fe mg/kg. This dosage was
shown to be sufficient to overwhelm the positive BOLD contribution during
functional activations (56, 57). The fMRI scans were initiated 15 min after the
injection, which was substantially longer than the recirculation time (on order of
seconds) in adult rats, to ensure steady-state MION distribution in the vasculature.
Using unilateral visual stimulation, nine rats were examined by post-MION
bSSFP fMRI. Among them, five were also examined by conventional CBV-
weighed fMRI using post-MION GE- and SE-EPI methods for comparison. For
visual stimulation, an optical fiber with one end connected to a green light-
emitting diode (LED) was placed 5 mm in front of the left eye. The LED was
flashed at 1 Hz with 5% duty cycle. One trial of the simulation paradigm
consisted of four blocks of 40-s rest and 20-s stimulation. Stimuli were
synchronized with the scanner under computerized control using LabVIEW v8.0
(National Instruments Corporation, Austin, TX). All animals were allowed to rest
for 10 minutes between stimulation trials. Two to five trials were recorded for
8/17/2019 FullText (39)
The effect of hypercapnia challenge on post-MION bSSFP signals was
examined in three animals. The gas conditioning paradigm consisted of 2-min
baseline followed by 4-min 5% CO2 exposure and 4-min recovery. In addition, T1 
and T2 mapping and bSSFP imaging were performed on one animal to document
the longitudinal and transverse relaxation rates (R1  = 1/T1  and R2  = 1/T2) and
bSSFP signal as a function of accumulated MION dosage ranging from 0 mg/kg
(before injection) to 30 mg/kg in 5 mg/kg increment.
2.2.2 MRI Protocols
All MRI experiments were performed on a 7 T Bruker scanner with a
maximum gradient of 360 mT/m (70/16 PharmaScan, Bruker Biospin GmbH,
Germany) using a 72 mm birdcage transmit-only RF coil and an actively
decoupled receive-only quadrature surface coil. Scout T2-weighted (T2W) images
were first acquired in three planes with a rapid acquisition with relaxation
enhancement (RARE) sequence to guide the positioning of the subsequent fMRI
slice at the standard coronal orientation covering Bregma -6.7 to -7.7 mm. To
avoid the narrow transition band and minimize banding artifacts in bSSFP, local
shimming was performed with a FieldMap based procedure prior to fMRI scans
(58).
All bSSFP fMRI images were acquired with alternating RF pulse (β  = π),
8/17/2019 FullText (39)
21
TR/TE = 3.8/1.9 ms, FOV = 32×32 mm2, acquisition matrix = 64×64 (zero-filled
to 128×128 during reconstruction unless when compared with EPI results), slice
thickness = 1 mm, number of slices = 1, NEX = 4 and temporal resolution of 1 s.
A relatively small flip angle α of 18o was estimated from the apparent T1 and T2 
values measured at 15 mg/kg MION by 1 / 
1 /  cos
 to provide maximum
flat pass-band region in the bSSFP profile (59). For comparison, post-MION
single-shot SE-EPI fMRI with TR/TE = 1000/21 ms and GE-EPI fMRI with
TR/TE = 1000/18 ms and α = 30o were performed with identical slice orientation,
spatial geometry and visual stimulation paradigm. The total acquisition time for
bSSFP, GE-EPI and SE-EPI is identical, which is 5 minutes each. For anatomical
referencing, a 2D RARE T2W image was acquired at the same slice location with
TR/TE = 4200/36 ms, acquisition matrix = 256×256, echo train length = 4 and
NEX = 2. To depict brain vasculature, a high resolution 2D T2*-weighted (T2*W)
image was also acquired using a FLASH sequence with TR/TE = 250/10 ms,
acquisition matrix = 512×512, α = 15o and NEX = 2.
For T1 mapping, a saturation recovery method, using rapid acquisition with
relaxation enhancement with variable repetition time (RAREVTR) sequence (60,
61), was employed with TR = 60, 120, 240, 480, 960, 1920, 3840 and 7680 ms
and TE = 7.5 ms. T2 mapping was performed using a Carr–Purcell–Meiboom–Gill
(CPMG) imaging sequence (62, 63) with TR = 4000 ms, 12 echoes, first TE of 7.5
ms and echo spacing of 7.5 ms.
8/17/2019 FullText (39)
2.2.3 Data Analysis
All fMRI data was realigned to the mean image of the time series using the
2D rigid-body transformation with AIR v5.2.5 (Roger Woods, UCLA, USA). The
first 5 images of each fMRI trial were discarded to eliminate possible non-
equilibrium effects in dynamic bSSFP or EPI series. A linear detrending with
least-square estimation was performed on the time-course of each voxel to
eliminate the baseline drift caused by physiological noises and system instability.
Cross-correlation coefficient (cc) activation maps were generated by calculating
the correlation between the measured time-course and the box-car function
representing the stimulation paradigm on a voxel-by-voxel basis using the
STIMULATE software package (Center for Magnetic Resonance Research,
University of Minnesota). To identify areas showing strong activation, cc
threshold of -0.35 and cluster of 2 voxels were applied.
bSSFP signal time-courses were collected from every activated voxel within
three regions of interest (ROIs) covering the contralateral (right) superior
colliculus (SC), the contralateral visual cortex (VC) and ipsilateral (left) VC.
After temporal low-pass filtering (<0.1 Hz), the time-courses were transformed to
signal percentage changes by normalizing to the baseline signal (mean of first 40
time points). Within each ROI, the time-courses for the voxel with the strongest
cc value as well as all activated voxels were computed by averaging all blocks,
trials and nine animals studied. They were presented as mean ± standard deviation
(SD).
For comparison of post-MION bSSFP, GE-EPI and SE-EPI fMRI methods,
the single voxel with the strongest cc value for each method was selected. Their
raw time-courses were normalized and plotted together with the temporally low-
pass filtered ones. Temporal SNR (tSNR) and CNR were measured for all the raw
time-courses. tSNR was calculated by σ  
 µ  =tSNR , where  µ  was the mean of the
non-stimuli-related time-courses and σ   its SD. CNR was calculated by
σ  
S  CNR
  = , where  S  was the average signal change during activation (44). The
time-courses averaged over all visually activated voxels (cc<-0.35) within the
entire brain region were also plotted for each method by averaging the five
animals studied (mean ± SD).
For hypercapnia data, the bSSFP time-courses were computed in all three
animals from an ROI covering the entire brain except the regions that were
severely darkened by MION due to large blood vessels. For T1 and T2 mapping,
T1-weighted and T2W signals in each voxel were fitted with the monoexponential
relaxation and decay functions, respectively, using a nonlinear least square
algorithm provided by ParaVision 5.03 (Bruker Biospin GmbH, Germany). ROI
measurements were performed using ImageJ v1.40g (Wayne Rasband, NIH,
USA).
2.3 Results
Figure 2.1 shows that bSSFP images exhibited better spatial conformity to
8/17/2019 FullText (39)
24
anatomical T2W images than GE- and SE-EPI ones. The severe image distortion
seen in EPI images was not observed in bSSFP images. Note that all images were
acquired from the same animal at identical slice location during the same scan
session. bSSFP, GE-EPI and SE-EPI images had the same spatial resolution (500
× 500 µm2). Before MION injection, certain brain regions (e.g. those close to the
ear canal) in EPI images suffered from signal dropout and image distortion, which
were absent in bSSFP images within the locally shimmed region (Figure 2.1 left
column). After 15 mg/kg MION injection, the T2W signal was seen to decrease in
the area between cortical and subcortical regions that contain mainly large blood
vessels (arrow in Figure 2.1a right). Such intravascular MION susceptibility effect
became more pronounced in the T2*W image as both blood vessels and their
surrounding tissue were dramatically darkened (arrow in Figure 2.1b right). More
importantly, severe signal loss and distortion occurred in GE-EPI (Figure 2.1d
right) and SE-EPI (Figure 2.1e right) after MION injection. In contrast, post-
MION bSSFP image exhibited less signal loss and no apparent distortion (Figure
2.1c right) and was in good agreement with post-MION T2W images.
8/17/2019 FullText (39)
Figure 2.1  Typical images acquired with conventional T2-weighted spin-echo (a),
conventional T2*-weighted gradient-echo (b), bSSFP (c), GE-EPI (d) and SE-EPI (e) at
the identical slice location from a rat brain before and after intravenous injection of
intravascular susceptibility contrast agent MION (15 mg/kg). Note that all bSSFP, GE-
EPI and SE-EPI images were acquired with the same spatial resolution and matrix size
(64×64) and reconstructed without any post-processing. 
8/17/2019 FullText (39)
Figure 2.2  Post-MION bSSFP fMRI yielded good agreement between the activation patterns and known neuroanatomy during unilateral 20-s block-design stimulations in a
normal adult SD rats (a). The activation map was computed by correlating the fMRI
time-course with the stimulation paradigm and overlaid on the post-MION bSSFP image.
Only the voxels with cross-correlation coefficient (cc) < -0.35 were color coded. For each
activation cluster, the time-courses from the single voxel with the strongest cc value (as
marked by the yellow crosses in the small inset) (b) and all activated voxels within the
8/17/2019 FullText (39)
27
ROI (as defined by the green lines in the small inset) (c) were plotted in mean ± SD.
These time-courses were computed by averaging across all blocks, trials and animals
studied (n = 9). The strongest activation was observed in the superficial layers of
contralateral SC. Note that bSSFP images were acquired in a 64×64 matrix and reconstructed to 128×128 by zero padding. The shaded area indicates the stimulation
period.
With post-MION bSSFP fMRI, unilateral visual stimulation produced robust
activations in contralateral SC and bilateral VC (Figure 2.2a). Without any spatial
co-registration, the activation map derived from bSSFP data yielded excellent
agreement between the activation patterns and known neuroanatomy such as the
superficial layers of contralateral SC, monocular area of primary VC (V1M), and
secondary VC (V2) of the contralateral cortical region, and binocular area (V1B)
of ipsilateral primary VC. Within each clustered region, the average time-course
from the single voxel with the strongest cc value (Figure 2.2b) and that from all
activated voxels (Figure 2.3c) showed robust bSSFP signal decrease in all animals
(n = 9) with small SDs. Contralateral SC, especially its superficial layers which
receives direct inputs from the retina (29, 64, 65), exhibited the highest percentage
signal changes. Contralateral and ipsilateral cortical activations exhibited similar
changes but contralateral activations were more extensive, covering the entire V1
and neighboring part of V2.
8/17/2019 FullText (39)
28
Figure 2.3 Typical activation patterns observed by post-MION bSSFP (a), GE-EPI (b)
and SE-EPI (c) fMRI methods from an animal. The imaging slice was located at Bregma
-7.2mm as shown in the coronal T2W image (d). For each method, the time-course
depicts the raw (colored) and low-pass filtered (black) signal changes in the voxel with
the strongest cc value in the SC (indicated by the green squares in the activation maps).
8/17/2019 FullText (39)
29
Figure 2.3 compares the typical post-MION bSSFP, GE- and SE-EPI fMRI
results obtained in one animal. All showed strong activations in SC and bilateral
VC (Figure 2.3 left column). For each method, the voxel with the strongest cc
value in the SC region was selected (as marked by green squares in Figure 2.3a-c).
Its raw time-course and filtered one were plotted for comparison (Figure 2.3 right).
Robust signal decrease in response to stimulation was observed by all three
methods from all five animals studied. The percentage bSSFP signal change was
between those for SE-EPI and GE-EPI. The tSNRs were 42.8 ± 4.7, 32.4 ± 3.4
and 53.6 ± 6.2 for bSSFP, GE- and SE-EPI methods, respectively, while CNRs
were 1.6 ± 0.3, 2.2 ± 0.5, 1.4 ± 0.4.
Table 2.1 summarizes the strongest voxel-wise cc value, mean cc value of all
activated voxels and number of activated voxels within the entire brain region in
all five animals studied. There was no significant difference in the strongest
voxel-wise cc value between bSSFP and GE- or SE-EPI fMRI results. The mean
cc value of bSSFP method was significantly weaker than that of GE-EPI (P  <
0.001), but not significantly different from that of SE-EPI method. Note that
bSSFP yielded more activated voxels than GE- and SE-EPI methods (P < 0.01).
Figure 2.4 depicts the mean time-courses of all activated voxels (cc value < -0.35)
for post-MION bSSFP, GE-EPI, and SE-EPI methods by averaging the data from
all animals (n = 5). Robust signal decrease in response to four 20-s stimulations
was observed for all methods. Although the mean bSSFP signal percentage
change was smaller than that of EPI methods, bSSFP yielded smaller SD.
8/17/2019 FullText (39)
30
Table 2.1  Comparison of post-MION bSSFP, GE-EPI and SE-EPI fMRI methods as
determined from activated voxels (cc < -0.35) within the entire slice in all five animals
studied.
bSSFP -0.69 ± 0.05 *  -0.46 ± 0.02 †  29 ± 7 § 
GE-EPI  -0.73 ± 0.06  -0.51 ± 0.06  22 ± 9 
SE-EPI  -0.64 ± 0.04  -0.49 ± 0.06  20 ± 11 
†
  Significantly weaker than that of GE-EPI method (P  < 0.001) but not significantly
different from that of SE-EPI method.
§ Significantly more than those by GE- and SE-EPI methods (P < 0.01).
Statistical analysis was performed using two-tailed paired Student’s t-test with P < 0.05
considered as significant.
8/17/2019 FullText (39)
31
Figure 2.4 Average post-MION bSSFP (a), GE-EPI (b) and SE-EPI (c) time-courses in
all activated voxels (cc < -0.35) within the entire slice in the animals studied (n = 5). The
shaded areas indicate the stimulation periods.
Figure 2.5 shows that the average post-MION bSSFP signal decreased
significantly in the entire brain in response to the 4-min 5% CO2 challenge in all
three animals studied. Such global signal reduction was observed in each
8/17/2019 FullText (39)
32
individual animal. Figure 2.6a shows that R1  increased substantially at small
MION concentration (≤  5 mg/kg) in various brain regions. However, such R1 
increase reached a plateau at high concentration (≥ 10 mg/kg), which was similar
to previous studies (66, 67). On the other hand, a strong and approximately linear
dependence (e.g., R2  = 0.98 for the whole brain ROI) of R2  on MION
concentration was observed (Figure 2.6b) in agreement with the earlier reports (68,
69). Figure 2.6c shows that bSSFP signal decreased monotonically with
increasing MION concentration.
Figure 2.5 Average post-MION bSSFP signal time-course from the entire brain during
the 4-min hypercapnia challenge using 5% CO2 inhalation in the animals studied (n = 3).
Measurement ROIs were shown in the small inset, and covered all cortical and
subcortical regions except those severely darkened by MION because of large blood
vessels. The shaded area identifies the period of hypercapnia challenge.
8/17/2019 FullText (39)
33
Figure 2.6  Apparent tissue longitudinal relaxation rate R1  (=1/   T1) (a), transverse
relaxation rate R2 (=1/T2) (b) and bSSFP signal (c) as a function of MION concentration
in different brain regions of one animal.
8/17/2019 FullText (39)
The present study demonstrated the robustness of post-MION bSSFP fMRI
approach for detecting brain activity. The negative signal change mainly arises
from its CBV sensitivity although the detailed mechanism is complex and remains
to be elucidated. Conventional CBV-weighted fMRI relies on the increased
susceptibility effect of intravascular contrast agent caused by local CBV increase
upon activation, which manifests as the apparent R2* or R2 increase in GE- or SE-
EPI. At steady-state MION concentration, the relationship between relative CBV
change and R2* or R2  change is approximately linear (68). Consequently, the
conventional CBV-weighted fMRI is capable of quantitatively estimating the
relative CBV increases upon stimulation. In contrast, the bSSFP signal in
biological tissues is determined by multiple factors such as T1, T2, diffusion,
actual sequence parameters and susceptibility contributions. For simplicity, it is
often regarded to provide the T2 /T1- or R1 /R2-weighted contrast (59). With the
high intravascular MION dosage used in the current study, bSSFP signal from
intravascular spins becomes negligible and extravascular spins are subject to the
strong microscopically inhomogeneous magnetic fields around capillary vessels.
Given the small capillary vessel size (on order of few microns), short TR (of 3.8
ms) and high intravascular MION concentration (of 15 mg/kg), the mechanistic
effect of intravascular MION on bSSFP signal will likely be dominated by the
transition regime (between diffusion narrowing and static dephasing regimes) (46).
Therefore, the susceptibility-induced bSSFP signal decrease during activation
8/17/2019 FullText (39)
could be primarily attributed to the apparent R2 increase.
In this study, the linear dependency of apparent R2  increase on MION
concentration was observed (Figure 2.6b) although the actual R2  measurement
could be TE and sequence dependent. Such apparent R2  increase was a direct
result of the microscopic susceptibility effect, arising from diffusion decay in the
randomized magnetic susceptibility field created by intravascular contrast agent.
During functional activation, local CBV increase will augment the local
intravascular susceptibility effect and cause further bSSFP signal decrease. On the
other hand, one cannot completely ignore the MION T1  shortening effect on the
post-MION bSSFP fMRI contrast although the effect diminishes at high MION
concentration. Our experimental observation of the apparent tissue R1  saturation
at high MION concentration (Figure 2.6a) was consistent with the previous animal
studies in the brain and kidney (70, 71). It was also in good agreement with the
two-compartment model analysis (66, 72, 73) where the effect of intravascular T1 
contrast agent on apparent tissue R1  is ultimately limited by the
intravascular/extravascular water exchange rate and blood volume fraction. Thus,
the MION T1  effect on bSSFP signal changes during activation was likely
minimal at the high concentration used in the present study. In addition, the
bSSFP signal was observed to generally decrease with MION concentration
(Figure 2.6c), supporting the hypothesis that post-MION bSSFP fMRI is likely
dominated by the apparent R2 increase caused by CBV increase.
Hypercapnia challenge is known to induce vasodilation and CBV increase
8/17/2019 FullText (39)
36
(74). It is commonly employed to examine the brain hemodynamic regulation in
the absence of neuronal activation (50, 75). In the present study, continuous
physiological recordings showed that SpO2  and respiration rate increased while
heart rate slightly decreased upon inhaling CO2  (data not shown), which was in
agreement with the previous hypercapnia study (76). More importantly, the robust
post-MION bSSFP signal decrease during hypercapnia challenge (Figure 2.5)
further supported our hypothesis that post-MION bSSFP provides a CBV
sensitive fMRI approach. Approximately 35% CBV changes were reported
previously based on the post-MION SE T2W signal decrease during similar
hypercapnia challenge (74). However, the post-MION bSSFP signal decrease
during hypercapnia observed in the present study may not allow the direct
quantitation of such relative CBV increase because post-MION bSSFP signal is
not related to R2  in a purely exponential manner. In this regard, it remains
imperative to formulate a comprehensive description in order to understand post-
MION bSSFP signal change during functional activation.
Post-MION bSSFP fMRI provides better functional mapping quality than
conventional post-MION EPI-based fMRI. As shown in Figure 2.1, post-MION
EPI images suffered from signal dropout and severe distortion, dramatically
limiting their true spatial resolution and fidelity when mapping brain activities. In
contrast, post-MION bSSFP images showed good agreement with high resolution
anatomical images. The high image fidelity provided by post-MION bSSFP not
only alleviates difficulties in post-processing, but also facilitates accurate
8/17/2019 FullText (39)
37
localization of brain activities. As seen in Figure 2.3, the activation patterns in
both SE- and GE-EPI results were displaced to varying extents due to the
inhomogeneity and susceptibility induced image distortion. For example, the
ipsilateral cortical activation locations were shifted approximately by one to two
voxels when compared to those determined by bSSFP. Such region specific
distortion can affect the fMRI interpretation and is difficult to correct by post-
processing. This issue together with poor spatial resolution, strong susceptibility
and physiological noises, have severely hampered the conventional EPI-based
CBV-weighted fMRI study of small animals at high field. However, these
limitations can be mitigated by using the post-MION bSSFP fMRI demonstrated
in the present study. Rapid refocusing and short readout duration in bSSFP
sequence provide high signal efficiency with no image distortion (45) while the
CBV sensitivity incurred by intravascular susceptibility contrast agent minimizes
the contamination from remote site activations and physiological noises.
Robust tSNR and CNR enable post-MION bSSFP fMRI to capture
activations with sensitivity comparable to EPI methods. On the first order, bSSFP-
based fMRI contrast can be considered to be similar to that of post-MION SE-EPI,
which is mainly sensitive to changes in microvascular blood volume at capillary
level (44). Our experimental findings showed that post-MION bSSFP activation
pattern was more similar to that of SE-EPI (Figure 2.3 left). However, the
percentage change of bSSFP signal as determined from single voxel time-courses
was higher than that of SE-EPI during visual stimulation (Figure 2.3 right). This
8/17/2019 FullText (39)
38
might partly arise from the differences between bSSFP and SE-EPI signal
properties such as the prominent role of stimulated echo pathways in bSSFP
signal formation. Post-MION GE-EPI fMRI exhibited much larger percentage
signal change since it is sensitized to the relatively large R2* change during
activation, which is less influenced by vessel size and more sensitive to total CBV
change. Lastly, post-MION bSSFP fMRI yielded more activated voxels than EPI
methods (Table 2.1), likely a consequence of less signal loss and image distortion.
Several limitations exist for the fMRI approach proposed in this study. First,
use of intravascular susceptibility limits its application in humans. Nevertheless,
given that certain USPIOs such as MION have long blood half-life and induce no
apparent physiological and functional perturbations, this new approach can greatly
facilitate the fMRI study in animals such as rodents and monkeys. Second, the
present study demonstrated the post-MION bSSFP fMRI with single slice
acquisition. However, fast 3D bSSFP can be achieved for fMRI as shown in a
recent study using various 3D k-space trajectories (45). Such high resolution 3D
acquisition can be readily adopted to post-MION fMRI. Doing so may prolong the
acquisition time and affect temporal resolution, but it gains full brain coverage
and SNR. Third, the post-MION bSSFP fMRI requires good shimming to avoid
the banding artifacts in bSSFP image particularly at high field. Moreover, this
fMRI approach is largely based on the large flat portion of the off-resonance
profile that is sensitive to shimming. In this study, localized shimming was
carefully performed within the single image slab prior to fMRI acquisition to
8/17/2019 FullText (39)
minimize the banding artifacts and avoid the narrow transition band.
In this study, the distortion-free technique was demonstrated in the visual
area which was severely distorted in conventional EPI-based fMRI methods. It not
only enables precise localization of the activations to neuroanatomy in living
subjects, but also promotes better understandings of neuronal pathways of
different sensory systems. The visual region is not strongly affected by
susceptibility problems that causing signal dropouts, therefore the benefit in this
aspect from the new fMRI technique was not significant. However, it can be used
to investigate the brain regions that are vulnerable to the susceptibility in EPI
images. For example, the deep brain structures located in the posterior portion of
the brain, the auditory cortex and the olfactory bulbs which are close to air
cavities and suffer from significant signal loss in EPI images. With this regard,
post-MION bSSFP fMRI can be employed to examine the responses under
various types of functional manipulation such as auditory or pain stimulation.
2.5 Conclusion
contrast agent provides sensitivity comparable to conventional CBV-weighted
EPI-based fMRI but with no severe image distortion and signal dropout. Robust
negative responses during visual stimulation were observed and activation
patterns were in excellent agreement with known neuroanatomy. In addition, post-
MION bSSFP signal was observed to decrease significantly during hypercapnia
8/17/2019 FullText (39)
challenge, indicating its sensitivity to CBV changes. These findings demonstrated
that post-MION bSSFP fMRI is a promising alternative to conventional CBV-
weighted fMRI. It is particularly suited for fMRI investigation of animal models
at high field.
8/17/2019 FullText (39)
STATE FMRI CONNECTIVITY AFTER
COMPLETE AND PARTIAL CORPUS
Since the introduction of blood oxygenation level-dependent (BOLD)
contrast (10), functional MRI (fMRI) has offered a powerful approach to study
brain functions due to its noninvasiveness, large field-of-view and 3D imaging
capabilities compared with other imaging modalities. The majority of fMRI
studies have focused on examining the changes in neuronal activities associated
with stimuli or tasks. It is not until recently that studying the resting brain by
fMRI became of immense interest. The motivations mainly arise from two
aspects. First, most of the brain’s energy is consumed at rest by spontaneous
neuronal activity (20% of body’s energy) while the task-related increases in
energy metabolism are usually small (<5%) (11). Second, low-frequency
fluctuations (LFFs) (<0.1 Hz) of resting-state fMRI (rsfMRI) signals were found
to be coherent among brain areas with similar functions and known anatomical
interconnections (12, 13). Therefore, efforts have been made to examine the
coherence in LFFs, or resting-state connectivity, providing not only new insights
into the functional organization of the brain (14-16), but also a better
understanding of brain functional plasticity during disease, aging and learning
8/17/2019 FullText (39)
physiological mechanisms underlying coherent LFFs remain to be fully explored.
The interpretation of rsfMRI data is therefore hindered. Given the similarity
between the spatial organization of resting-state networks (RSNs) and
neuroanatomy, one common hypothesis is that resting-state connectivity is based
on anatomical connections. Anatomically, the hemispheres are interconnected by
axonal projections through midline commissural structures such as the corpus
callosum (CC), anterior commissure and posterior commissure. The largest
among them is the CC, which connects most areas of the cerebral cortex to
contralateral homologous areas that share similar functions (77, 78). Considering
the primary role of CC in interhemispheric communication, the relationship
between callosal connections and resting-state connectivity naturally is an issue of
interest. Previously, human studies have demonstrated the effects of the absence
of callosal connections on resting-state connectivity. Two rsfMRI studies on
callosal agenesis (79) and complete corpus callosotomy (in a single patient) (80)
showed significantly diminished and complete loss of interhemispheric
connectivity, respectively. These results support anatomical connections as key
constraints on resting-state connectivity. However, two other studies reported
predominately bilateral RSNs in a patient after complete transection of forebrain
commissures (81) and in patients with congenital callosal agenesis (82). These
findings favor another possibility that resting-state connectivity emerges flexibly
8/17/2019 FullText (39)
43
and is not limited by direct anatomical connections. While the interpretations of
the above studies seem to be contradictory, it may be due to the limited number of
subjects, large difference in subject age and diversity in remaining anatomical
connections. Therefore the role of CC in resting-state connectivity is still open t