UMEÅ UNIVERSITY MEDICAL DISSERTATIONS

New series No. 1505

________________________________________________

Cerebral blood flow and

intracranial pulsatility studied

with MRI

Measurement, physiological and pathophysiological aspects

Anders Wåhlin

Departments of Radiation Sciences and Clinical Neuroscience

Umeå University

Department of Biomedical Engineering and Informatics

Umeå University Hospital

Umeå, Sweden, 2012

ii

© Anders Wåhlin, 2012

ISBN: 978-91-7459-428-7 ISSN: 0346-6612

Printed by Print & Media

Umeå University, Sweden 2012

iii

Abstract

During each cardiac cycle pulsatile arterial blood inflates the vascular bed

of the brain, forcing cerebrospinal fluid (CSF) and venous blood out of the

cranium. Excessive arterial pulsatility may be part of a harmful mechanism

causing cognitive decline among elderly. Additionally, restricted venous

flow from the brain is suggested as the cause of multiple sclerosis.

Addressing hypotheses derived from these observations requires accurate

and reliable investigational methods. This work focused on assessing the

pulsatile waveform of cerebral arterial, venous and CSF flows. The overall

aim of this dissertation was to explore cerebral blood flow and intracranial

pulsatility using MRI, with respect to measurement, physiological and

pathophysiological aspects.

Two-dimensional phase contrast magnetic resonance imaging (2D PCMRI)

was used to assess the pulsatile waveforms of cerebral arterial, venous and

CSF flow. The repeatability was assessed in healthy young subjects. The 2D

PCMRI measurements of cerebral arterial, venous and CSF pulsatility were

generally repeatable but the pulsatility decreased systematically during the

investigation.

A method combining 2D PCMRI measurements with invasive CSF infusion

tests to determine the magnitude and distribution of compliance within the

craniospinal system was developed and applied in a group of healthy

elderly. The intracranial space contained approximately two thirds of the

total craniospinal compliance. The magnitude of craniospinal compliance

was less than suggested in previous studies.

iv

The vascular hypothesis for multiple sclerosis was tested. Venous drainage

in the internal jugular veins was compared between healthy controls and

multiple sclerosis patients using 2D PCMRI. For both groups, a great

variability in the internal jugular flow was observed but no pattern specific

to multiple sclerosis could be found.

Relationships between regional brain volumes and potential biomarkers of

intracranial cardiac-related pulsatile stress were assessed in healthy

elderly. The biomarkers were extracted from invasive CSF pressure

measurements as well as 2D PCMRI acquisitions. The volumes of temporal

cortex, frontal cortex and hippocampus were negatively related to the

magnitude of cardiac-related intracranial pulsatility.

Finally, a potentially improved workflow to assess the volume of arterial

pulsatility using time resolved, four-dimensional phase contrast MRI

measurements (4D PCMRI) was evaluated. The measurements showed

good agreement with 2D PCMRI acquisitions.

In conclusion, this work showed that 2D PCMRI is a feasible tool to study

the pulsatile waveforms of cerebral blood and CSF flow. Conventional

views regarding the magnitude and distribution of craniospinal compliance

was challenged, with important implications regarding the understanding

of how intracranial vascular pulsatility is absorbed. A first counterpoint to

previous near-uniform observations of obstructions in the internal jugular

veins in multiple sclerosis was provided. It was demonstrated that large

cardiac-related intracranial pulsatility were related to smaller volumes of

brain regions that are important in neurodegenerative diseases among

elderly. This represents a strong rationale to further investigate the role of

excessive intracranial pulsatility in cognitive impairment and dementia.

For that work, 4D PCMRI will facilitate an effective analysis of cerebral

blood flow and pulsatility.

v

Table of Contents

Original papers 1  1 Introduction 3  

1.1 Arterial pulsatility 5  1.2 Restricted venous flow 8  1.3 CSF dynamics 10  1.4 Structural magnetic resonance imaging 14  1.5 Flow sensitive magnetic resonance imaging 16  1.6 Quantification of CSF pulse pressure 21  

2 Aims 25  3 Materials and methods 27  

3.1 Subjects 27  3.2 Magnetic resonance imaging 28  3.3 Volume of arterial and CSF pulsatility 30  3.4 CSF pulse pressure 33  3.5 Mathematical model 34  3.6 Intracranial volume 35  3.7 Regional brain volumes 35  3.8 Statistics 37  

4 Results 39  4.1 Repeatability 39  4.2 Craniospinal compliance 41  4.3 Restricted venous flow 42  4.4 Intracranial pulsatility and brain volumes 43  4.5 Four-dimensional flow measurements 45  

5 Discussion 47  5.1 Measurement aspects 47  5.2 Physiological aspects 49  5.3 Pathophysiological aspects 51  

6 Conclusion 55  7 Acknowledgements 57  8 References 59  

1

Original papers

This dissertation is based on the following papers, which are referred to by

their Roman numerals in the text.

I. Wåhlin, A., Ambarki, K., Hauksson, J., Birgander, R., Malm, J., and

Eklund, A. (2012). Phase contrast MRI quantification of pulsatile

volumes of brain arteries, veins, and cerebrospinal fluids

compartments: Repeatability and physiological interactions. J Magn

Reson Imaging 35, 1055–1062. Reprinted with permission.

II. Wåhlin, A., Ambarki, K., Birgander, R., Alperin, N., Malm, J., and

Eklund, A. (2010). Assessment of craniospinal pressure-volume

indices. AJNR Am J Neuroradiol 31, 1645–1650. Reprinted with

permission.

III. Sundström, P., Wåhlin, A., Ambarki, K., Birgander, R., Eklund, A.,

and Malm, J. (2010). Venous and cerebrospinal fluid flow in

multiple sclerosis: a case-control study. Ann. Neurol. 68, 255–259.

Reprinted with permission.

IV. Wåhlin, A., Ambarki, K., Birgander, R., Malm, J., and Eklund, A. In

healthy elderly the volumes of several brain regions are related to

pulsatility in cerebral arteries and cerebrospinal fluid. In

Manuscript.

V. Wåhlin, A., Ambarki, K., Birgander, R., Wieben, O., Johnson, KM.,

Malm, J., and Eklund, A. Measuring pulsatile flow in cerebral

arteries using four-dimensional phase contrast magnetic resonance

imaging. In Manuscript.

2

3

1 Introduction

The adult brain is contained in a closed compartment constituted by the

cranium and dura mater (Mokri, 2001). Inside this space cerebrospinal fluid

(CSF) surrounds the entire central nervous system (Magendie, 1842). During

each cardiac cycle arterial pulsatility causes a systolic increase in cerebral

arterial blood volume (Greitz et al., 1992), something that can be observed

during neurosurgery as a cyclic pulsation of the brain. To compensate for

this volume increase, venous blood is forced out of the cranium and CSF is

pushed down to the spinal canal (Enzmann and Pelc, 1991; Greitz et al.,

1992; Enzmann and Pelc, 1993).

Recently, there has been increased attention on the pulsatility of cerebral

arteries and CSF. Among elderly, an increased pulsatility in the arteries

supplying the brain has been linked to cognitive impairment in several

domains, including memory (Mitchell et al., 2011). It has been hypothesized

that an age-related increase in pulsatility in the cerebral microcirculation is

harmful for the brain (O'Rourke and Hashimoto, 2007). The CSF pulse

pressure (Anile et al., 2010; Eide and Brean, 2010) and cardiac-related flow

pulsatility have also been linked to the outcome following shunt surgery in

subjects with communicating hydrocephalus (Bradley et al., 1996).

Moreover, although multiple sclerosis (MS) is commonly regarded as an

inflammatory disorder caused by a complicated interaction between

environmental and genetic risk factors (Compston and Coles, 2002),

restricted venous flow (Zamboni et al., 2009b) and altered CSF flow

(Zamboni et al., 2009c) were recently suggested to be parts of the disease

pathophysiology. This controversial theory is appealing as restricted venous

outflow potentially can be treated (Zamboni et al., 2009a).

4

Addressing hypotheses derived from these observations requires reliable

investigational methods. It is also necessary to explore the normal

physiological mechanisms associated with cerebral arterial, venous and CSF

dynamics.

In this dissertation a biomechanical perspective on the interaction between

cerebral arteries, veins and CSF was applied. With a theoretical model as a

starting point, we probed the system from several angles while maintaining a

critical evaluation of measurement reliability. Focus was on assessing the

pulsatile waveforms of arterial and venous blood and CSF with phase

contrast magnetic resonance imaging (PCMRI). These measurements were

also combined with invasive CSF dynamic investigations, quantifying the

magnitude of the CSF pulse pressure occurring in response to the vascular

pulsations.

5

1.1 Arterial pulsatility

The left and right internal carotid artery (ICA) and vertebral artery (VA),

respectively, are four arteries that supply the brain (Figure 1). Closer to the

brain, the VA merge to form the basilar artery (BA) (Purves et al., 2011).

Figure 1. Arterial supply and venous drainage of the brain. ICA=internal

carotid artery, VA=vertebral artery, BA=basilar artery, MCA=middle

cerebral artery, ACA=anterior cerebral artery, SSS=superior sagittal

sinus, IJV=internal jugular vein. The figure is based on a 4D PCMRI

acquisition from Paper V.

6

At the circle of Willis, a hub connecting the major cerebral arteries, the ICA

is divided into the anterior and middle cerebral artery (ACA and MCA,

respectively). Here, the BA is also divided to form the posterior cerebral

arteries. Blood from the ICA supplies a large part of the cerebrum. The VA

and the BA supply the brain stem, cerebellum and posterior parts of the

cerebrum (Hall and Guyton, 2010).

In young subjects with well functioning arterial systems, the arterial

distensibility, or compliance, cushions the arterial pulse. This is sometimes

referred to as the Windkessel effect (Westerhof et al., 2009). This absorption

reduces the systolic pressure peak and smoothens the pulsatility of blood

flow that reaches the capillaries (O'Rourke and Hashimoto, 2007). In

general, for a higher cardiac stroke volume the compensating mechanisms of

the arteries have to absorb a greater volume. The compliance of the arterial

tree together with the cardiac stroke volume ultimately determine the

magnitude of arterial pulse pressure (Hall and Guyton, 2010).

About 30 million heartbeats per year induce arterial fatigue and dilatation in

the later stages of life (O'Rourke and Hashimoto, 2007; Bullitt et al., 2010).

In parallel with atherosclerotic changes the artery is also hardened and less

compliant, and a gradual decline in the Windkessel function is observed

(Avolio et al., 1983). These interrelated changes are most prominent in the

aorta and major proximal arteries and less prominent in the distal regions of

the circulation. Consequently, a stiffer arterial system may result in

increased pulsatility and cyclic distention of distal branches of the arterial

system (Boutouyrie et al., 1992).

For most organs, the small arteries, arterioles and capillaries complete the

transition of pulsatile blood flow into a steady flow. However, the high

perfusion rate of the brain makes arterial pulsatility reach deeply towards

the smallest vessels (O'Rourke and Hashimoto, 2007; Mitchell, 2008).

Research regarding large arteries and their age-related stiffening have

7

promoted questions on what effects this might have on tissues and organs of

the body (Mitchell, 2008). If aortic stiffness impedes appropriate cushioning

of the arterial pulse, the pulsatile stress will extend deeper into the arterial

tree. In this event the sites with the most dilated arterioles, e.g. the brain,

may be at risk to damage caused by the increase in arterial pulsatility.

Research rationale

Recently, a large-scale evaluation focusing on elderly subjects demonstrated

that cognitive performance in several domains, including memory, is

negatively correlated with the amplitude of pulsatility in arteries supplying

the brain (Mitchell et al., 2011). Based on these observations, it is relevant to

evaluate if excessive pulsatility in cerebral arteries catalyze the aging of the

brain, and possibly even pushes a subject into a degenerative dementia. It is

essential to clarify if any specialized brain region is particularly susceptible

to excessive pulsatile stress. Studies evaluating associations between

regional brain volumes and intracranial vascular pulsatility may provide a

first indication of which brain regions that may be harmed by such

processes.

8

1.2 Restricted venous flow

In contrast to the arterial system, the cerebral venous drainage displays large

individual variations (Doepp et al., 2004) and is easily influenced by external

disturbances such as body position (Valdueza et al., 2000). Through

bridging veins and deep cerebral veins the blood is transmitted to the venous

sinuses, formed by the dura mater (Figure 1). The sinuses transport blood to

the neck veins. In supine position, the major venous pathways of the neck

are the left and right internal jugular vein (IJV) (Valdueza et al., 2000;

Alperin et al., 2005). Commonly, the right IJV carries a dominant part of the

total IJV outflow but individual variations of this distribution are large

(Stoquart-ElSankari et al., 2009). In addition to the left and right IJV, extra-

jugular venous pathways contribute to the venous outflow. These include

vertebral veins, epidural veins and deep neck veins. The large capacity of

these veins indicates that they could take over the entire venous drainage of

the brain if needed (Doepp et al., 2004).

Despite the extensive capacity of collateral pathways, a controversial

hypothesis suggests that the development of MS is a manifestation of

restricted venous flow. MS lesions often develop around a central vein (Tan

et al., 2000), a rationale for suspecting venous involvement in the etiology of

MS. Recently, the pattern of impaired cerebral venous outflow termed

chronic cerebrospinal venous insufficiency (CCSVI) has been linked to MS

(Singh and Zamboni, 2009; Zamboni et al., 2009b). The theory describes

compromised venous outflow as a cause of accumulation of iron in the

central nervous system, which may initiate the development of MS lesions.

The IJV flow is reported as a central component in the hypothesis (Zamboni

et al., 2009b). Stenoses or malformations have been suggested to be the

primary cause of the restricted venous outflow. The restricted venous flow is

also reported to bring about an abnormal CSF flow in the cerebral aqueduct

(Zamboni et al., 2009c). Potentially, venous congestions of the CCSVI type

9

can be treated, improving the symptoms of MS patients (Zamboni et al.,

2009a). Therefore, if proven accurate, this may represent a major

therapeutic option for MS patients.

Research rationale

Because CCSVI potentially can be treated (Zamboni et al., 2009a), centers

around the world are now offering endovascular procedures, i.e. angioplasty

and stenting, to treat CCSVI. This has resulted in “medical tourism” where

MS patients themselves seek help abroad. The scientific process needs to

catch up with this potentially hasty development and the CCSVI theory

needs to be evaluated by accurate, reliable and objective methods. The link

between venous obstructions and MS was observed using ultrasound

methods (Zamboni et al., 2009b). Because venous flow can also be

quantified using PCMRI (Stoquart-ElSankari et al., 2009), it can be expected

that the observations also will appear using this modality. Therefore, studies

clarifying the PCMRI appearance of venous hemodynamics in MS are

urgently needed.

10

1.3 CSF dynamics

The brain and spinal cord are surrounded by CSF (Figure 2). CSF is a clear

fluid (99 % water) that is produced within the choroid plexus of the cerebral

ventricles (Redzic and Segal, 2004) at a rate of approximately 7 µl/s

(Ekstedt, 1978). In healthy elderly, the typical intracranial, ventricular and

spinal CSF volumes have been measured to 195 ml (Tsunoda et al., 2002), 37

ml (Ambarki et al., 2010), and 85 ml (Edsbagge et al., 2011), respectively.

CSF outflow is commonly thought to occur in the arachnoid villi located in

the superior sagittal sinus (Figure 2), although at physiological pressure

levels the significance of cerebral capillary fluid exchange may be more

significant than previously recognized (Bulat and Klarica, 2011). The CSF

production rate, outflow resistance and also venous pressure determine the

CSF resting pressure (Davson et al., 1970; 1973).

The CSF has a role in clearing metabolic waste as well as providing

mechanical support and buoyancy to the central nervous system (Irani,

2009). Furthermore, it provides the craniospinal system with the ability to

compensate for a volume expansion. This function of the system is activated

during every cardiac cycle, when CSF is redistributed upon the systolic

increase in arterial blood volume (Enzmann and Pelc, 1991).

In the short time window of a cardiac cycle, the CSF outflow mechanisms are

ineffective for dampening the CSF pulse pressure generated by the arteries.

Instead the increased CSF pressure compresses the craniospinal venous bed,

forcing venous blood out from the system (Greitz et al., 1992). This

mechanism is a manifestation of craniospinal compliance. In addition,

elasticity of dura mater, especially in the spinal compartment, is generally

acknowledged to contribute to the overall capacity of craniospinal

compliance (Martins et al., 1972; Tain et al., 2011).

11

Figure 2. Overview of the CSF system surrounding the entire central

nervous system. CSF is produced within the ventricles, circulates and

escapes to blood in the venous sinuses.

12

To describe craniospinal compliance, mathematical models have been

developed. The relationship between pressure and volume of the

craniospinal system can be approximated by a simple exponential function

(Marmarou et al., 1975):

P = PrekV (1)

where P is the CSF pressure, Pr is the baseline pressure, V is the volume

increase and k is the elastance coefficient. This model has been extended

with a constant term to better match experimental observations (Avezaat

and van Eijndhoven, 1986):

P = P1ekV +P0 (2)

where P0 and P1 are pressure constants with a sum that equals Pr. P0 has been

linked to venous pressure (Löfgren et al., 1973; Avezaat and van Eijndhoven,

1986) and hydrostatic pressure gradients within the system (Raabe et al.,

1999). Compliance is defined as the volume change per unit change in

pressure (dV/dP), and for the craniospinal system the pressure dependent

compliance becomes:

C(P) = 1k(P −P0 )

(3)

Usually, the term pressure volume index (PVI) is used instead of k. The

mathematical relationship between PVI and k is:

PVI = 10.4343k

(4)

The PVI is therefore a factor describing the overall compensating capacity of

the craniospinal system.

13

Research rationale

The concept of compliance and PVI also applies to arterial pulsations

(Avezaat et al., 1979). A reduced PVI corresponds to a reduced ability of the

craniospinal system to accommodate arterial pulsatility. A dysfunction of

this type has been suggested to be present in Alzheimer’s disease and

vascular dementia (Bateman et al., 2008). Moreover, the amplitude of the

CSF pulse pressure has been shown to be important in predicting shunt

outcome in idiopathic normal pressure hydrocephalus (Eide and Brean,

2006). Together these observations suggest that increases in arterial

pulsations (often associated with increasing age) and impaired pulsation

dampening may be involved in several degenerative neurological disorders

among the elderly.

For this reason, it is important to develop accurate methods to measure PVI

and remedy the gap in basic physiological knowledge regarding the

magnitude and distribution of craniospinal compliance among healthy, non-

demented elderly.

14

1.4 Structural magnetic resonance imaging

Magnetic resonance imaging (MRI) is based upon the interactions between

nuclear spins, a fundamental property of all elementary particles, and

external magnetic fields. Briefly, a powerful external magnetic field aligns all

spins in the body. An external excitation radiofrequency pulse rapidly “tips”

the spins in a prescribed volume (or slice). In their excited state the spins

induce detectable signals in receiver coils, integrated within the system

(Bloch, 1946; Purcell et al., 1946).

Spatial information is encoded to the spins by position-dependent phase and

frequency modulation achieved by adding small magnetic field gradients to

the static field. The detected signals can therefore be reconstructed into an

image (Lauterbur, 1973).

The duration of the induced signal depends on the unique relaxation

properties of the tissue. The difference in relaxation properties between

tissues can be exploited to generate contrast in the images. The relaxation

properties can be described by a longitudinal and transversal relaxation time

(T1 and T2, respectively) and those two parameters differ between white

matter, gray matter and CSF (Figure 3) (Bernstein et al., 2004).

The T1 and T2 contrast can be used to depict brain structures and also to

delineate the intracranial compartment. Although the image intensity alone

under many circumstances is not specific to a certain brain structure, e.g. the

hippocampus and amygdala are inseparable by intensity alone (Fischl,

2012), the spatial information and surrounding pixels can be weighed in to

the decision whether a pixel belongs to the brain structure. This

methodology is being exploited in highly automated structural processing

tools that are capable of providing an accurate segmentation of the entire

brain. In this dissertation Freesurfer was employed. Freesurfer is a set of

tools for automated structural processing that are based on probabilistic

15

models created from manually labeled brains (Fischl et al., 2002; 2004;

Desikan et al., 2006). Freesurfer is capable of utilizing spatial relationships,

e.g. the hippocampus is always behind and below the amygdala and never

above and in front of amygdala (Fischl, 2012), in the automatic labeling of

brain regions.

Figure 3. MRI appearance at different locations and weightings. In the T1

image CSF is black. In the T2 image CSF is white.

16

1.5 Flow sensitive magnetic resonance imaging

There are several techniques for measuring and visualizing motion using

MRI. Typically, dynamic information such as flow or diffusion is acquired

simultaneously as the spatial information. PCMRI is useful for flow rate and

velocity quantification purposes (Bernstein et al., 2004).

PCMRI employs velocity encoding in a way analogous to spatial encoding

(Moran, 1982). Because the angular frequency of a spin is proportional to the

external magnetic field, a spin that travels in a magnetic field gradient (the

velocity encoding gradient) will accumulate a different phase as compared to

a stationary spin. Therefore, a bipolar gradient toggle rewinds stationary

spins but induces a phase shift in moving spins (Figure 4).

Figure 4. Velocity encoding principle of PCMRI. (a) The spin moves in the

gradient direction during a bipolar gradient toggle, G+ and G-. (b) The

phase evolution over time for the spin. Note the resulting phase shift.

17

Phase wrapping occurs at shifts of π and -π. To avoid this, the magnitude of

the velocity encoding gradient is prospectively adjusted according to the

maximum velocity expected in the vessel of interest. The velocity encoding

parameter corresponds to the velocity where phase wrapping occurs

(Bernstein et al., 2004).

The first moment of a gradient field determines the velocity sensitivity of the

acquisition. In two-dimensional (2D) PCMRI, two acquisitions with different

first moments are required to quantify flow in one direction. This way, a

subtraction or complex conjugate multiplication between the two

acquisitions allows removal of unwanted phase shifts that do not depend on

velocity. The relationship between velocity and phase shift can be described

as:

φ =v •γΔ

M1 (5)

where φ is the phase shift, v is the velocity, γ is the Larmour frequency

constant and ΔM1 the difference in first moments of the velocity encoding

gradients.

When quantifying flow in all tree spatial directions, a series of three or four

acquisitions together with a reference scan can be utilized to shorten the

scan time (Wigström et al., 1996; Johnson and Markl, 2010). The velocity

can be obtained from solving:

φ = γAv (6)

where A is the encoding matrix containing the differences in first moments

in analogy to equation 5. Commonly, time resolved quantification of flow in

all three directions is referred to as four-dimensional (4D) PCMRI.

18

Figure 5. 2D PCMRI images at three time points during the cardiac cycle.

(a1-3) Blood flow at cervical level in the systolic, intermediate and diastolic

cardiac phase, respectively. (b1-3) Spinal CSF flow. Black and white

represents flow from and towards the cranium, respectively. (c1-3)

Aqueduct CSF flow. Black and white represents flow towards and from the

fourth ventricle, respectively.

19

The PCMRI sequence can be used to quantify arterial, venous and CSF flow

waveforms of the cardiac cycle (Figure 5 and 6) (Nayler et al., 1986; Greitz et

al., 1992; Enzmann and Pelc, 1993). To achieve this, an ECG or a peripheral

pulse is recorded by equipment integrated with the MRI system, allowing the

sequence to be prospectively or retrospectively gated (Bernstein et al.,

2004). In prospective gating, the data acquisition is initiated by a trigger

signal. This technique is robust even in mild arrhythmia but a small fraction

of the cardiac cycle, where the sequence waits for the next trigger, is not

acquired. In retrospective gating, continuously acquired data is sorted into

an average cardiac cycle. This technique provides the flow waveform of the

complete cardiac cycle, although some temporal smoothing may occur as a

result of heart rate variability (Lotz et al., 2002).

Figure 6. Blood and CSF flow waveforms. In these acquisitions 32 image

frames represent the cardiac cycle. Positive values represent flow in the

caudal-cranial direction.

20

Research rationale

There are many factors that potentially degrade the quality of the PCMRI

quantification. Partial volume effects (Tang et al., 1995) cause flow rate

overestimations. Misalignment errors influence the observed velocity (Wolf

et al., 1993). Moreover, physiological variability adds to the uncertainties

associated with a measurement.

Net flow rate quantifications have been thoroughly validated, both for 2D

(Spilt et al., 2002) and 4D PCMRI (Gu et al., 2005; Chang et al., 2011).

However, less is known about the repeatability when quantifying the volume

of cerebral arterial, venous and CSF pulsatility. Therefore, to meet with the

increased interest in intracranial pulsatile hemodynamics and CSF flow it is

necessary to evaluate the stability of PCMRI measurements.

21

1.6 Quantification of CSF pulse pressure

The CSF system dynamics can be investigated invasively by inserting two

needles in the spinal lumbar subarachnoid space (Eklund et al., 2007). One

needle is used to infuse artificial CSF and the other is used to determine the

pressure response. Lumbar infusion tests are commonly used to investigate

the CSF outflow resistance in cases of suspected normal pressure

hydrocephalus (Andersson et al., 2005; Sundström et al., 2010; Malm et al.,

2012), but the technique is also useful for investigating compliance related

parameters of the craniospinal system (Juniewicz et al., 2005). Various

infusion patterns are available and they can be divided into two categories,

those regulating the CSF pressure to predetermined levels (Ekstedt, 1977)

and those employing a predetermined rate of infusion (Hussey et al., 1970;

Katzman and Hussey, 1970).

Figure 7. CSF pressure during a constant pressure infusion test. The CSF

pressure is regulated to six levels (each given its own color) above baseline

pressure (yellow).

22

The constant pressure regulated infusion pattern (Figure 7) is designed for

estimations of CSF outflow resistance and does not provide estimations of

PVI (Andersson et al., 2005; 2010). However, in the pressure signals

recorded from a lumbar infusion test it is evident that physiological activity

such as arterial cardiac-related pulsatility influences the CSF pressure.

During an infusion test it can be observed how the CSF pulse pressure,

caused by arterial pulsatility, increases linearly as the average pressure in the

system is raised (Figure 8) (Avezaat et al., 1979; Avezaat and van

Eijndhoven, 1986). This is a manifestation of a gradual decrease in

compliance.

Figure 8. CSF pulse pressure versus average CSF pressure. The colors

correspond to different levels of regulated pressure.

23

Research rationale

Although the relationship between CSF pressure and pulse pressure is

related to the compliance properties of the craniospinal system, i.e. PVI, it is

also related to the volume of cerebral arterial pulsatility, which cannot be

estimated solely from CSF pressure measurements. Interestingly, the volume

of arterial pulsatility can be estimated from PCMRI measurements.

Therefore, bringing together data from CSF infusion measurements and

PCMRI acquisitions can potentially provide estimations of the craniospinal

PVI.

24

25

2 Aims

The overall aim of this dissertation was to explore cerebral blood flow and

intracranial pulsatility using MRI, with respect to measurement,

physiological and pathophysiological aspects.

The specific aims for Papers I-V were to:

I. Assess 2D PCMRI measurement repeatability for quantifications of the

pulsatile waveforms of cerebral arterial, venous and CSF flow.

II. Combine lumbar CSF infusion tests with 2D PCMRI flow measurements

to assess PVI and its distribution between the intracranial and spinal

compartments.

III. Test the most vital part of the CCSVI hypothesis by comparing IJV flow

as well as CSF dynamics between healthy and MS subjects.

IV. Assess relationships between regional brain volumes and arterial and

CSF pulse pressure and the volume of cerebral arterial and CSF pulsatility.

V. Evaluate 4D PCMRI for assessing pulsatility of large cerebral arteries.

26

27

3 Materials and methods

3.1 Subjects

The regional ethical review board approved all separate studies. Informed

consent was obtained from all participants. The subject groups for all studies

are listed in Table 1. All healthy young (HY) subjects were normal as

determined from anatomical MRI scans. The healthy elderly (HE) passed a

neurological examination, had a mini mental state examination score of at

least 28 and did not suffer from advanced vascular disease (previous stroke,

myocardial infarction, diabetes). The MS group had the relapsing-remitting

disease course with a median disease duration of five years.

Table 1. Overview of subjects enrolled in the five studies. HY=healthy

young, HE=healthy elderly, MS=multiple sclerosis.

Paper No. of subjects Male / female Age (mean ± SD)

I 20 HYa 12/8 33±8

II 37 HEb 15/22 71±6

III 20 HYa, 21 MS 12/8, 8/13 33±8, 34±10

IV 37 HEb 13/24 71±6

V 10 HY 7/3 36±9

aSame subjects. b34 subjects were shared.

28

3.2 Magnetic resonance imaging

For Paper I-IV MRI measurements were performed on a 3T Achieva scanner

(Philips Healthcare, Best, the Netherlands) using an 8-channel head coil.

Routine anatomical sequences were performed to confirm a healthy status. A

3D T1 weighted turbo field echo sagittal scan with good gray/white matter

contrast was used for automated regional brain volume quantification (echo

time 5.0 ms, repetition time 10 ms, acquisition resolution 0.5 x 0.6 x 1 mm3).

An axial T2 weighted turbo spin-echo (echo time 80 ms, repetition time

3000 ms, acquisition resolution 0.5 x 0.5 x 3.3 mm3) was used for manual

delineation of the intracranial space.

For Paper I-IV three 2D PCMRI protocols were used to assess blood and CSF

flow pulsatility. The tree protocols measured:

1. ICA and VA flow (level between second and third cervical vertebrae)

2. Spinal CSF flow (level between second and third cervical vertebrae)

3. Aqueduct CSF flow

The scanning parameters for sequences 1 / 2 / 3 were: encoding velocity 70 /

7 / 20 cm s-1, flip angle 15 / 10 / 10 degrees, repetition time 9.6 / 16 / 15 ms,

echo time 5.8 / 11 / 10 ms, voxel size 0.9 x 0.9 x 6 / 1.2 x 1.2 x 5 / 1.2 x 1.2 x 5

mm3 and two signal averages.

In Paper I five repeated measurements of the three sequences were

performed in the 20 HY subjects. An aqueduct measurement with increased

resolution (0.7 x 0.7 x 0.7 mm3) was also added to evaluate potential

influence of partial volume effects. For paper II and IV the first two 2D

PCMRI protocols were utilized (34 HE were shared between the studies).

29

In Paper III, 21 MS subjects underwent the first and last 2D PCMRI

protocols, and the values were subsequently compared with the first

repetition of the same sequences performed on the HY group of Paper I.

In Paper V measurements on 10 HY subjects were performed on a 3T

Discovery MR 750 scanner (General Electric, Waukesha, WI, USA) with a

32-channel head coil. Measurements started with a 3D time of flight

acquisition for vessel localization. Secondly, a 4D PCMRI measurement was

made using a phase contrast vastly undersampled projection imaging

(PCVIPR) sequence (Frydrychowicz et al., 2010), covering the entire

intracranial space. The resolution was 0.7 x 0.7 x 0.7 mm, velocity encoding

110 cm s-1, repetition time/echo time 6.5 / 2.7 ms, flip angle 8 degrees.

Besides velocity images for 20 time positions of the cardiac cycle, a time

averaged complex difference image was reconstructed to provide anatomical

localization of the vascular system.

After the 4D PCMRI acquisition, five 2D PCMRI scans were performed.

These measured flow in:

1. ICA and BA

2. Right MCA (M1 segment)

3. Left MCA (M1 segment)

4. Right ACA (A1 segment)

5. Left ACA (A1 segment)

The encoding velocity for protocol 1 / 2 / 3 / 4 / 5 was 80 / 110 / 110 / 90 /

90 cm s-1. The additional 2D PCMRI parameters were: repetition time/echo

time 7.6-8.5 / 4.1-5.0 ms, 15 degrees flip angle, resolution 0.5 × 0.5 x 3.0

mm3, six views per segment and two signal averages. Thirty-two phases were

retrospectively reconstructed.

30

3.3 Volume of arterial and CSF pulsatility

For all 2D PCMRI acquisitions, arterial, venous and CSF lumen

segmentations were performed manually using the free software ImageJ

(National Institute of Health, Bethesda, MD)(Paper I-IV) and Segment

(http://segment.heiberg.se)(Paper V). All measurements were systematically

reviewed and corrected for phase aliasing if needed. The total arterial flow

waveform was calculated from the sum of ICA and VA flow. Total internal

jugular blood flow was calculated from the sum of left and right IJV flow

(Paper III). The venous waveforms were also normalized by the total arterial

inflow (Paper III). An IJV reflux was registered if, during any time of the

cardiac cycle, the flow was in the caudo-cranial direction.

For the 4D PCMRI segmentations in Paper V, in-house software was

developed (Figure 9). The software enabled the user to rotate, zoom and

visualize flow in the 4D PCMRI data. The segmentation was initiated by

selecting a reference pixel in the vessel of interest (a mouse click in a

maximum intensity projection of the complex difference image). The

software included neighboring pixels with values of at least 18 % of the

maximum of the complex difference image. An additional requirement was

used to limit the segmentation to the branch of interest:

ri •v0v0

≤l2

(7)

where is a vector from the reference pixel to the i:th pixel, is the vessel

direction estimated by the velocity direction (temporally averaged) in the

reference pixel and l is desired segment length. The segment length l was

fixed at five pixel widths (similar to the slice thickness of the 2D PCMRI

scans). The value 18 % was selected from a pre-calibration series indicating

that this threshold provided complete vessel coverage.

riv0

31

In this dissertation flow pulsatility was described in terms of volume. This

concept is natural considering that cerebral arterial, venous and CSF

volumes interact as a result of the limited intracranial volume. The volume of

pulsatility (∆V) was calculated using an identical procedure for arteries,

veins and CSF (Figure 10) (Paper I-V). After subtraction of the net flow rate,

cumulative integration was used to estimate the volume variation associated

with the remaining waveform. ∆V was defined as the difference between the

maximum and minimum of the integrated waveform.

Figure 9. A maximum intensity projection of the complex difference

image. Flow rates from the colored arterial segmentations were compared

with the 2D PCMRI data.

32

Figure 10. Arterial waveforms from a 4D PCMRI measurement. (a) The

flow rate during the cardiac cycle for right ICA, MCA and ACA as well as

the basilar artery. (b) Cumulative integral of the ICA waveform after

subtracting the net flow rate. The volume (∆V) of ICA pulsatility was

defined as the maximum minus minimum of this waveform.

33

3.4 CSF pulse pressure

Lumbar CSF pressure and pulse pressure were measured with the subject in

a supine position (Paper II and IV). A special bed, exposing the region for

needle insertion, enabled access to the lumbar space. The highly automated

investigational method has previously been described in detail (Andersson et

al., 2005; Malm et al., 2011). After 10-15 minutes of rest, with the needles in

place, the CSF pressure was sampled for 5 minutes at 100 Hz.

After the resting pressure measurement the CSF pressure was regulated to

six predetermined pressure levels using infusion of artificial CSF. In a few

exceptions, pressure elevation was performed using a constant rate of

artificial CSF infusion.

Figure 11. The ∆P for each regulated pressure level (colored dots) was

defined as the median CSF pulse pressure (gray dots). The colors

correspond to the pressure levels in Figure 8. The line represents the

relationship between ∆P and P determined from a linear regression.

34

In the post-processing step the slow pressure variations, not related to the

cardiac cycle, were eliminated using a high-pass filter. After this operation,

pressure pulses were estimated by analyzing the difference between

maximum and minimum pressure in successive 1.5 second time intervals.

For each regulated pressure level, a ∆P value was calculated as the median

value of all pulse pressures. Resting pressure ∆P was used for Paper IV. In

Paper II a linear regression was used to estimate the relationship between

∆P and P, as a step in calculating PVI (Figure 11). In Paper IV, baseline ∆P

could not be assessed for three subjects because of needle obstruction. In

addition to these measurements a standard automated blood pressure

monitor, Omron M5-I (Omron Health Care, Kyoto, Japan) was used to

assess left upper arm mean arterial pressure and arterial pulse pressure.

3.5 Mathematical model

In Paper II, a way to calculate PVI from PCMRI and CSF infusion data was

presented. Here, the craniospinal model (equation 2) was adapted to suit the

combination of measurements. The response in pressure following an

increase in arterial volume can be written as:

P +∆ P = P1ek (V+∆V ) +P0 (8)

Where ∆V is the cyclic increase in intracranial arterial blood volume and ∆P

is the CSF pulse pressure. This can be rewritten to:

(9)

where the left hand is the mathematical expression of the slope between CSF

∆P and P that can be determined from an infusion test (as described in the

previous section and Figure 11), with P0 as the intersection between the

regression line and the pressure axis.

∆ PP −P0

= ek∆V −1

35

Consequently, using both a CSF infusion measurement and a measurement

of ∆V, the elastance coefficient can be calculated as:

k = ln( ∆ PP −P0

+1) /∆V (10)

To comply with common nomenclature, the elastance coefficient was

translated into PVI according to equation 4.

3.6 Intracranial volume

The post-processing software QBrain (version 2.0; Medis Medical Imaging

Systems, Leiden, the Netherlands) was used to segment the intracranial

volume of each subject in Paper IV. This was performed by manual

delineation in T2 weighted images using a threshold based drawing tool

(Ambarki et al., 2010; 2011).

3.7 Regional brain volumes

In Paper IV cortical and subcortical structures were segmented using

FreeSurfer 5.0 (available online at http://surfer.nmr.mgh.harvard.edu). The

software utilizes probabilistic models treating relationships between

characteristic magnetic resonance appearances and spatial orientations

(Figure 12) (Fischl et al., 2002; 2004; Desikan et al., 2006).

An experienced neuroradiologist verified the accuracy of the segmentations

of the 13 brain regions extracted for the analysis (Figure 12). Paper IV

originally included 38 subjects, but one subject had too severe atrophy for

the automated segmentation, which ended with unacceptable errors. This

subject was excluded from all analyses.

36

Figure 12. Result from brain segmentation (Paper IV). (a) Cortical

subdivisions. (b) Subcortical structures. Besides these structures the insula,

cerebral white matter and cerebellum white matter were included in the

analyses.

37

3.8 Statistics

For Paper I the within-subject and between-subject standard deviations were

calculated (SDintra and SDinter, respectively). SDintra describes the variation

over the five repetitions and is thus a measure of repeatability. To put this

value in proportion to the population distribution, an intraclass correlation

coefficient (ICC) was calculated as SDinter2 /(SDintra

2 +SDinter2 ) . An ICC above

0.80-0.85 is generally regarded as sufficient repeatability for isolated

measurement. Differences between the results from the two aqueduct

measurements with different resolution were assessed using the Wilcoxon

signed rank test. The Pearson correlation coefficient was used to assess the

relationships between pulsatility measures.

Analysis of variance for repeated measurements was used to assess

systematic differences between the repetitions performed in Paper I. The

association between total arterial ∆V and the sum of cervical CSF and IJV ∆V

was investigated using the Pearson correlation coefficient.

Group differences were assessed by 2-sided t-tests or Mann-Whitney tests

where appropriate (Paper II-III). The chi-square test was used for

comparing prevalence of IJV reflux between HY and MS groups (Paper III).

In Paper IV, multiple linear regression was used to assess relationships

between regional brain volumes and CSF pulse pressure, arterial pulse

pressure, the volume of spinal CSF pulsatility, the total volume of VA

pulsatility (left plus right artery) and the total volume of ICA pulsatility (left

plus right artery). The partial correlation coefficient was used to describe the

strength of identified relationships.

In Paper V, differences between 2D and 4D PCMRI derived measures were

investigated using 2-sided paired samples t-tests. In all studies, a p-

value<0.05 was regarded as significant.

38

39

4 Results

4.1 Repeatability

Results from repeated 2D PCMRI measurements in Paper I are displayed in

Table 2. For all measures, the ICC was above 0.85, except for the volume of

right and left ICA pulsatility (ICC=0.80 and ICC=0.84, respectively) and net

flow in the aqueduct (ICC=0.41). Estimations of the volume of aqueduct CSF

pulsatility were lower for the high-resolution sequence than for the low-

resolution sequence (0.03 ml versus 0.05 ml, p<0.001). The difference

between the sequences decreased for larger sizes of the cerebral aqueduct, as

determined from the high-resolution sequence (p=0.031).

There was a significant decrease in the volume of total arterial pulsatility and

net flow rate between repetitions (p=0.017 and p=0.024, respectively).

Intra-subject variations in the volume of total arterial pulsatility correlated

with changes in heart rate (r=-0.51, p<0.001). When removing linear trends

between measurements, i.e. taking away the systematic decrease in

pulsatility that occurred between repetitions on group level, the repeatability

for the volume of both left and right ICA pulsatility increased to ICC>0.85.

A second analysis was aimed towards assessing the relationship between

pulsatility in arteries, veins and CSF. Analyzing the flow into and out of the

cranium showed that the volume of total arterial pulsatility correlated with

the volume of summed spinal CSF and IJV pulsatility (r=0.75, p<0.001).

This relationship is visualized in Figure 13.

40

Table 2. Results from repeated ∆V measurements. Blood and spinal CSF

flows were quantified at cervical C2-C3 level.

∆V (ml) Net flow (ml/s)

Location mean±SDinter SDintra mean±SDinter SDintra

ICA (R) 0.56±0.14 0.07 4.7±1.2 0.33

ICA (L) 0.57±0.15 0.07 4.7±1.0 0.32

VA (R) 0.20±0.08 0.03 1.4±0.7 0.17

VA (L) 0.30±0.10 0.03 2.2±0.6 0.22

Total arterial

1.60±0.35 0.15 12.9±2.5 0.77

IJV (R) 0.64±0.39 0.11 5.6±2.4 0.63

IJV (L) 0.36±0.23 0.08 3.6±2.0 0.43

Total IJV 0.97±0.49 0.16 9.2±2.3 0.70

Aqueduct CSF

0.045± 0.023

0.005 0.0055± 0.003

0.004

Spinal CSF

0.77±0.23 0.05 - -

41

Figure 13. The volume of total arterial pulsatility versus the volume of

summed spinal CSF and total IJV pulsatility. The line represents equality.

4.2 Craniospinal compliance

Paper II provided values on the magnitude and distribution of craniospinal

compliance among healthy elderly. The volume of total cerebral arterial

pulsatility (ICA+VA) measured at cervical level was 1.98±0.43 ml. This was

significantly larger than in the younger group of Paper I (p<0.001). The

volume of spinal CSF pulsatility was 0.68±0.23 ml. The craniospinal PVI

calculated from combined 2D PCMRI and pressure data was 11.8±9.0 ml.

The volume of spinal CSF pulsatility was 35 % of the volume of total arterial

pulsatility. For comparison the corresponding value for the younger group of

Paper I was 49 %. This difference was significant (p<0.001).

42

4.3 Restricted venous flow

The balance between left and right IJV net flow rate for the MS and control

groups of Paper III are visualized in Figure 14. No clear difference between

the groups could be observed. There were no differences in total, left or right

IJV net flow (p=0.38, p=0.51 and p=0.38, respectively). Five MS subjects

and five controls displayed an IJV reflux during the cardiac cycle (p=0.93).

There were no significant differences regarding aqueduct CSF net flow or

volume of pulsatility (p=0.75 and p=0.14, respectively).

Figure 14. Normalized blood flow of the left and right internal jugular

veins. Encircled subjects indicate those with IJV reflux.

43

4.4 Intracranial pulsatility and brain volumes

In healthy elderly, the volumes of several gray matter structures were

negatively related to CSF pulse pressure and the volume of ICA and CSF

pulsatility (Figure 15 and Table 3). The strongest relationships concerned the

temporal cortex and hippocampus. In agreement with these observations,

the ventricular volume increased for larger pulsatility. The volume of VA

pulsatility and arterial pulse pressure were not correlated to the volume of

any brain region.

Figure 15. Partial correlation coefficients between regional brain volumes

and pulsatility variables. Correlation values are stacked on each other for

visibility. All significant relationships are also presented in Table 3.

GM=gray matter, WM=white matter. *Significant at α=0.05, **Significant

at α=0.01

44

Internal relationships between the measures of pulsatility were also

assessed. The volume of ICA pulsatility was related to the volume of spinal

CSF pulsatility as well as CSF pulse pressure, but only after adjusting for net

ICA flow (p=0.01 and p=0.05, respectively). The volume of spinal CSF

pulsatility and CSF pulse pressure were not significantly correlated with each

other.

Table 3. Partial correlations between regional brain volumes and

pulsatility measures. ∆V ICA was calculated from the sum of the left and

right ICA waveforms.

Brain region ∆V ICAa ∆V CSFb ∆P CSFc

Frontal cortex

p=0.404

p=0.017 r=-0.41

p=0.008 r=-0.48

Temporal cortex

p=0.003 r=-0.51

p=0.006 r=-0.47

p=0.028 r=-0.40

Cerebral White Matter

p=0.062 p=0.043 r=-0.35 p=0.698

Hippocampus p=0.003 r=-0.50

p=0.040 r=-0.36

p=0.679

Ventricular System

p=0.011 r=0.44 p=0.110 p=0.008

r=0.47

aAdjusted for net ICA flow, age, sex, intracranial volume and mean arterial

pressure, n=37. bAdjusted for age, sex, intracranial volume and mean

arterial pressure, n=37. cAdjusted for age, sex, intracranial volume and

mean arterial pressure, n=34. r=partial correlation coefficient.

45

4.5 Four-dimensional flow measurements

The comparison in Paper V showed that 2D and 4D PCMRI produced highly

correlated results for the volume of arterial pulsatility as well as for net flow

(r=0.86 and r=0.95, respectively, n=69 vessels). These values improved

further (r=0.93 and r=0.97, respectively) when excluding measurements

with more than a 5 % variation in heart rate between the 4D and 2D

acquisitions (n=31 vessels). The correlation between all vessel segments is

illustrated in Figure 16.

Significant differences between 4D and 2D PCMRI were found for ICA and

MCA net flow (p=0.004 and p<0.001, respectively) as well as for the volume

of MCA pulsatility (p=0.006), n=69 (Table 4 and 5). However, these

differences were attenuated when excluding measurements with large

variations in heart rate (n=31 vessels).

Table 4. Comparison between 2D and 4D PCMRI measurements of

arterial ∆V. Values are given in ml.

vessel n ∆V 4D

mean±SD

∆V 2D

mean±SD

2D-4D

mean±SD p-value

BA 10 0.20±0.08 0.21±0.08 0.01±0.05 0.504

ICA 20 0.35±0.10 0.36±0.10 0.01±0.06 0.656

MCA 20 0.23±0.05 0.20±0.05 -0.03±0.04 0.006

ACA 19 0.14±0.08 0.12±0.08 -0.02±0.05 0.102

46

Figure 16. 2D versus 4D PCMRI derived ∆V for major cerebral arteries.

The line represents equality. The correlation coefficient for the entire

ensemble of vessels was r=0.86.

Table 5. Comparison between 2D and 4D PCMRI measurements of

arterial net flow. Values are given in ml/s.

vessel n Net flow 4D

mean±SD

Net flow 2D

mean±SD

2D-4D

mean±SD p-value

BA 10 2.15±0.58 2.18±0.49 0.03±0.34 0.787

ICA 20 3.72±0.70 4.02±0.62 0.29±0.40 0.004

MCA 20 2.54±0.36 2.26±0.34 -0.28±0.29 <0.001

ACA 19 1.47±0.52 1.44±0.49 -0.02±0.21 0.627

47

5 Discussion

This dissertation focused on cardiac-related intracranial pulsatility using a

unique combination of multi-modal MRI and invasive lumbar infusion tests.

The analysis provided insight into the pulsatile forces exerted on the brain,

and how the compensating mechanisms of the CSF system and venous blood

cooperate in attenuating these rapid excitations. The approach taken in this

dissertation has improved the understanding of the stability of PCMRI

measurements, the mechanisms involved in intracranial pulsatility and how

intracranial pulsatility may be related to pathophysiology.

5.1 Measurement aspects

Paper I investigated the repeatability of 2D PCMRI net flow and pulsatility

measurements of cerebral arterial, venous and CSF flows (Table 2). The

volume of arterial pulsatility was sensitive to variations in heart rate and

decreased as the subject remained in the MRI machine. This could be related

to a decrease in anxiety and blood pressure as the subject gradually adapted

to the unusual environment. This favored standardization of the acquisition

order, e.g. always performing the PCMRI measurements last in the MRI

protocol.

The 2D PCMRI net flow measurements were repeatable for the ICA, VA and

IJV. The volume of spinal and aqueduct CSF pulsatility also had sufficient

precision. This indicated that aqueduct stroke volume measurements for

selecting shunt candidates in normal pressure hydrocephalus (Bradley et al.,

1996; Kahlon et al., 2007) are adequate from a measurement repeatability

perspective. However, the small size of the aqueduct makes the

measurements susceptible to partial volume overestimations, whereby as

high resolution as possible should be used. The net aqueduct CSF flow

showed high variability between measurements and interpretations based on

single measurements of this quantity should be avoided (Table 2).

48

Nevertheless, the average aqueduct net CSF flow reported in Paper I is

similar to invasive estimations (Ekstedt, 1978) and a previous 2D PCMRI

study (Huang et al., 2004). Besides limitations associated with the

measurement technique, slow changes in cerebrovascular caliber related to

physiological and neuronal activity likely causes slow oscillations of CSF at

the aqueduct, which will also will be reflected as variability in the

measurements. Therefore, the stability in this measure may be improved by

longer acquisition times.

The 4D PCMRI measurements produced estimations of the volume of

arterial pulsatility (Figure 16) and net flow consistent with results from 2D

PCMRI (Paper V). This supports using a single 4D PCMRI measurement to

assess pulsatility in many intracranial arteries, despite using a velocity

sensitivity adapted for vessels containing high velocities, such as the MCA.

The increase in 2D and 4D PCMRI correlation when excluding

measurements with large differences in heart rates (Paper V) indicated that a

large part of the random differences between measurements is a result of

actual physiological variations. This is in agreement with the results from

Paper I where it was demonstrated that physiological variability influences

the repeatability.

The standard deviation of the difference between 2D and 4D PCMRI for the

volume of ICA pulsatility and net flow was 0.06 ml and 0.40 ml/s,

respectively (Paper V), which agreed with the corresponding values provided

from repeated 2D PCMRI measurements of Paper I (0.07 ml and 0.32-0.33

ml/s, respectively, Table 2). This indicated that 2D and 4D PCMRI were

associated with the same level of measurement variability, supporting that

4D PCMRI can replace 2D PCMRI, at least for the ICA.

Interestingly, the 2D PCMRI measurements of ICA net flow and the volume

of ICA pulsatility in Paper V were smaller than corresponding values from

49

Paper I (approximately one sixth and one third lower for net flow and

pulsatility, respectively). Machine dependence may have contributed to this

discrepancy. However, phantom measurements indicated that for large

vessels, such as the ICA, the accuracy was better than 10 % for both the

Philips (Paper I) and GE system (unpublished data). The large relative

difference in ICA pulsatility could originate from differences in

measurement location. In Paper I, ICA measurements were made more

proximally, whereby the arterial pulsatility likely was less attenuated by

arterial compliance. This is in line with previous observations regarding the

dampening of arterial pulsatility along the ICA (Schubert et al., 2011).

Additionally, the true temporal resolution of the 2D PCMRI sequence used

in Paper V is lower than the sequence in Paper I-IV (a consequence of

employing multiple views per segment) (Lotz et al., 2002). This smoothens

the reconstructed arterial waveform, something that potentially decreases

the estimated volume of arterial pulsatility.

The 4D PCMRI workflow required minimal prospective planning. This is

very different from 2D PCMRI where skillful observations and

interpretations are needed to place the measurement section perpendicular

to the vessel of interest. Moreover, the 4D PCMRI analyses were highly

automated (Figure 9), and only required basic anatomical knowledge from

the user. This simplification in the PCMRI workflow will make the technique

more accessible for future research projects, aiming to determine cerebral

hemodynamics in multiple cerebral vessels.

5.2 Physiological aspects

In Paper I the correlation between arterial pulsatility and the sum of venous

and cervical CSF pulsatility supported the notion that venous blood and CSF

escapes the cranium in response to an increased arterial volume (Figure 13).

This finding was in line with the Monro-Kellie principle and previous

analyses (Enzmann and Pelc, 1993; Greitz, 1993; Balédent et al., 2001; Kim

50

et al., 2007). Essentially, this emphasizes the role of venous blood, as an

important part of craniospinal compliance, in absorbing intracranial

pulsatility.

The PVI is a factor describing the magnitude of craniospinal compliance

(Marmarou et al., 1975; Shapiro et al., 1980). It describes how well arterial

volume pulsations are absorbed at a certain pressure within the system

(Avezaat and van Eijndhoven, 1986). Different infusion methods produce

disparate PVI values (Tans and Poortvliet, 1989; Juniewicz et al., 2005;

Sundström et al., 2010) whereby new methods are desirable. The method

dependence in the PVI estimate may in part be a consequence of slow

vasogenic responses to the infusion (Fridén and Ekstedt, 1983) that

complicate the time series analysis used in the calculation. The presented

method to assess PVI avoids the time series analysis by utilizing the pulsatile

CSF dynamics, averaged over many cardiac cycles, both for infusion and

MRI data. This quantification is likely less prone to slow vasogenic

responses, but can still be obtained regardless of infusion pattern. We note

that the measurement stability for the volume of arterial pulsatility was

poorer (Paper I) than assumed in Paper II. The effect will be a less reliable

estimate of PVI. Given the observations from Paper I, a possible

improvement could be obtained from modeling effects of physiological

variability such as heart rate. Alternatively, several measurements can be

averaged to obtain a more stable value.

The PVI value provided in Paper II was the first estimation in healthy

subjects based on the lumped parameter model in equation 2. The value is

lower than previously reported in adult subjects (Shapiro et al., 1980). We

hypothesize that this difference is related to the inclusion of the constant

term in the mathematical model of the CSF system. Indeed, the studies using

the constant term (Czosnyka et al., 2001; Juniewicz et al., 2005) have

produced lower PVI estimates than studies without the constant term

(Shapiro et al., 1980; Maset et al., 1987).

51

Venous compression was treated as part of the craniospinal compliance. This

is important since otherwise the venous waveform should be subtracted

from the arterial waveform when calculating the volume of the excitation.

This calculation would yield the cyclic change in cerebral blood volume, i.e.

the part of the arterial volume increase that is not directly compensated by

venous outflow. The different approaches have been the subject of

controversy as the method applied in Paper II indicated that the capacity of

the spinal compliance is less than that of the intracranial compartment. This

result has been argued not to conform with anatomical considerations, i.e.

that the spinal dura mater is not as strictly confined in bone as the dura

mater in the cranium (Tain et al., 2011), and with previous observations in

subjects with a cervical block (Magnaes, 1989). However, by considering that

the volume of spinal CSF pulsatility was approximately one third of the

volume of total arterial pulsatility, it is clear that the dominating

compensating mechanism, for processes in the time frame of a cardiac cycle,

was located intracranially.

Interestingly, the proportion of spinal compliance appeared different

between young and elderly subjects (Paper I and II). In elderly subjects a

larger part of the total arterial pulse was absorbed by intracranial

compensating mechanisms. The volume of total arterial pulsatility was also

higher among elderly, something that may be a reflection of reduced aortic

compliance in the elderly, inducing larger pulsatility in distal intracranial

arteries.

5.3 Pathophysiological aspects

With focus on IJV flow in the supine position there were no signs of altered

venous hemodynamics specific to MS (Paper III). The great variability in the

lateralization of the IJV flow appeared similar between cases and controls

(Figure 14). In most cases and controls, the right IJV conducted the major

52

part of the internal jugular drainage, which is in accordance with the

literature (Stoquart-ElSankari et al., 2009).

We were unable to find a difference in the aqueduct CSF flow between MS

and controls. Here, it is important to note that the aqueduct measurements

were associated with partial volume overestimations (Paper I). However,

because only results from sequences with identical settings were compared,

this limitation hardly caused any group specific biases in the CSF waveform.

An IJV reflux was only observed in vessels carrying small flow rates, and

only during a fraction of the cardiac cycle. As CCSVI has been described,

with IJV insufficiency in 94 % (33 of 35) of relapse-remitting MS cases

(Zamboni et al., 2009b), and almost never present in controls, this case-

control evaluation should have had the ability to detect a difference. The

failure to do so challenges the view of CCSVI as an important factor in MS.

Paper III was one of the first scientific evaluations of CCSVI ever performed.

Today, there is accumulating evidence that the role of IJV congestion does

not have a causative role in MS (Doepp et al., 2010; Centonze et al., 2011;

Doepp et al., 2011; Zivadinov et al., 2011). Nevertheless, although the

association between CCSVI and MS is weaker than initially reported, more

research is required to exclude a role of altered venous hemodynamics in the

MS etiology.

Paper IV investigated relationships between intracranial pulsatile stress and

regional brain volumes. The sizes of several vital brain structures were

related to arterial and CSF pulsatility (Figure 15). The volumes of the frontal

cortex, temporal cortex, hippocampus and ventricles were related to two or

more measures of intracranial cardiac-related pulsatility (Table 3). These

brain regions are reported to have rapid age-related structural alterations

(Tisserand et al., 2004; Fjell et al., 2009a; 2009b), stressing the importance

of adjusting for age in the statistical analysis performed in this study.

53

Although we studied healthy subjects, the array of anatomical features that

were related to large pulsatility resembled features linked to Alzheimer’s

disease, i.e. small hippocampus and large ventricles (Fjell et al., 2009a). The

findings should encourage investigating intracranial cardiac-related pulsatile

stress as a potential risk factor that may cause or contribute to cognitive

impairment and dementia.

There are no previous studies on arterial and CSF pulsatility and regional

brain volumes. However, cortical thickness correlates with risk factors for

cerebrovascular disease (Leritz et al., 2011). Surprisingly, the pattern of age-

related tissue reduction may appear without reduction in local cerebral

perfusion rates (Chen et al., 2011). Furthermore, in Alzheimer’s disease,

where vascular risk factors are increasingly being recognized, the spatial

distribution of gray matter thinning is associated with a matching pattern of

hyperperfusion consistent with the paths of large cerebral vessels and their

early divisions (Alsop et al., 2008). These paradoxical relationships may be

explained if cardiac-related intracranial pulsatile stress harms the brain. In

this scenario, high tissue perfusion rates and proximity to a large artery

likely increases the risk of damage (O'Rourke and Hashimoto, 2007).

The cortical regions with volumes related to intracranial pulsatility were

located near large cerebral vessels and their first branches, where vascular

pulsatility likely is more intense. Still, the unique relationship between

frontal cortex size and CSF pulsatility suggests that other mechanisms than

pure intravascular damage might be active. Additionally the finding of a

relationship between CSF ∆P and ventricular volume further emphasizes the

potential importance of intracranial pulsatility in normal pressure

hydrocephalus.

It is relevant to summarize the factors shaping intracranial pulsatility. The

magnitude of the cardiac stroke volume and compliance of the aorta and pre-

cranial arteries determine the magnitude of cerebral arterial pulsatility.

54

Therefore, large vessel stiffness may induce increased intra-vascular

pulsatile stress. The craniospinal compliance determines the corresponding

rise in CSF pressure, which constitutes a force of extra-vascular pulsatile

stress inside the cranium.

The volume of CSF pulsatility and CSF pulse pressure were both related to

the volume of ICA pulsatility, but they were not related to each other. This

was a manifestation of the variability in magnitude and distribution of

craniospinal compliance, in agreement with Paper II. This indicated that

although all measures of pulsatility were related to the same phenomena,

each variable held a considerable amount of unique information, supporting

the choice to run correlation analyses between regional brain volumes and

all available measures related to intracranial cardiac-related pulsatile stress

(Paper IV).

Given the results of Paper IV, it is important to further study the mechanism

behind the relationships between regional brain volumes and intracranial

cardiac-related pulsatility. Potentially, this provides one additional piece of

evidence of the link between arterial and cerebral aging. Furthermore, it may

open an avenue of new treatment options aimed towards cognitive decline.

In perspective of the aging population (Lutz et al., 2008) it will be

increasingly important to evaluate factors contributing to cerebral aging,

something that starts earlier than previously believed (Singh-Manoux et al.,

2012). This dissertation suggests that PCMRI may be an important tool for

that work.

55

6 Conclusion

We found that 2D PCMRI was a feasible tool to study the pulsatile waveform

of cerebral arteries, veins and CSF. However there were differences between

measurements that largely depended on physiological variability in the

parameters of interest. To reduce the impact of this limitation, all flow

quantifications should be performed last in the MRI protocol, so that the

patient can adapt to the potentially stressful environment.

The volume of summed cerebral venous and spinal CSF pulsatility was

related to the volume of intracranial arterial pulsatility. This verified the

exchange in volume that occurs between the main intracranial constituents

during the cardiac cycle, and also supported the view that intracranial veins

contribute to craniospinal compliance.

We showed how PCMRI and CSF infusion tests could be combined to

estimate PVI. The intracranial contribution to the overall craniospinal

compliance was larger than previously thought. The distribution of

craniospinal compliance was also different between age groups.

Using 2D PCMRI we were not able to reproduce the findings associated with

the vascular MS hypothesis, CCSVI. If CCSVI is an entity associated with MS,

the association is likely weaker than previously reported. Importantly, this

did not support endovascular procedures in the veins of MS subjects.

56

Among elderly, CSF pulse pressure and the volume of arterial and CSF

pulsatility were related to the sizes of several vital brain structures. The

anatomical features associated with cardiac-related intracranial pulsatility

resemble features that are strongly linked to Alzheimer’s disease. This

finding should stimulate further research exploring intracranial cardiac-

related pulsatility as a contributing factor to cognitive decline and dementia.

4D PCMRI provided a viable methodology to estimate cerebral arterial

pulsatility. The results were consistent with 2D PCMRI estimations. The

workflow required minimal prospective planning and the output allowed

comprehensive analysis of intracranial hemodynamics. Therefore, 4D

PCMRI is an important tool for future investigations of altered intracranial

hemodynamics and associated disorders.

57

7 Acknowledgements

I wish to express my sincere gratitude to:

Anders Eklund, my main supervisor, for giving me the opportunity to work

in this captivating field. You have been a great source of inspiration and I am

forever grateful for your dedicated mentorship. Jan Malm, my supervisor,

for your enthusiasm and commitment to push our research field further.

Richard Birgander, my supervisor, for sharing and teaching your expertise

regarding complex neuroanatomy. Khalid Ambarki, a great colleague and a

close friend. Working with you has been so much fun. My co-authors, Jon

Hauksson, Peter Sundström, Noam Alperin, Kevin Johnson and Oliver

Wieben. Kristin Nyman, Sonja Edvinsson, Hanna Ackelind and Ann-Catrine

Larsson for skillful measurements and for always being helpful. My

colleagues in the group as well as fellow PhD candidates, Kennet Andersson,

Anders Behrens, Tomas Bäcklund, Gabriel Granåsen, Hanna Isarelsson,

Niklas Lenfeldt, Sara Qvarlander and Nina Sundström. The staff at the

biomedical engineering and radiophysics departments. Anna Wernblom, for

administrative support. Olof Lindahl, Ronnie Lundström and the late Stefan

Karlsson for leading the department and providing a nice environment to

conduct research in. Anders Garpebring and Greger Orädd for fun

experiments, interesting discussions and good collaboration. Göran

Mannberg, for #!/bin/bash and your helpfulness. The staff at AMI centre,

for the angiogram on the cover. All my friends that make the spare time fun.

My relatives (including those I count as relatives) and especially my

grandmother. You are all so supportive and kind. My mother and father. You

are true role models and continue to inspire me. My wife, Mirjam, for your

constant support and encouragement.

58

This project was supported by the Swedish research council; VINNOVA; and

the Swedish Foundation for Strategic Research through their common

initiative: ‘Biomedical engineering for improved health’; Swedish research

council Grant No. 621-2011-5216; EU, Objective 2 Norra Norland, CMTF;

The County Council of Västerbotten; and the Swedish Heart and Lung

Foundation.

59

8 References

Alperin, N., Lee, S.H., Sivaramakrishnan, A., and Hushek, S.G. (2005).

Quantifying the effect of posture on intracranial physiology in humans by

MRI flow studies. J Magn Reson Imaging 22, 591–596.

Alsop, D.C., Casement, M., de Bazelaire, C., Fong, T., and Press, D.Z. (2008).

Hippocampal hyperperfusion in Alzheimer's disease. Neuroimage 42, 1267–

1274.

Ambarki, K., Israelsson, H., Wåhlin, A., Birgander, R., Eklund, A., and

Malm, J. (2010). Brain ventricular size in healthy elderly: comparison

between Evans index and volume measurement. Neurosurgery 67, 94–99;

discussion99.

Ambarki, K., Wåhlin, A., Birgander, R., Eklund, A., and Malm, J. (2011). MR

imaging of brain volumes: evaluation of a fully automatic software. AJNR

Am J Neuroradiol 32, 408–412.

Andersson, K., Manchester, I.R., Malm, J., and Eklund, A. (2010). Real-time

estimation of cerebrospinal fluid system parameters via oscillating pressure

infusion. Med Biol Eng Comput 48, 1123–1131.

Andersson, N., Malm, J., Bäcklund, T., and Eklund, A. (2005). Assessment

of cerebrospinal fluid outflow conductance using constant-pressure infusion-

-a method with real time estimation of reliability. Physiol. Meas. 26, 1137–

1148.

Anile, C., De Bonis, P., Albanese, A., Di Chirico, A., Mangiola, A., Petrella, G.,

and Santini, P. (2010). Selection of patients with idiopathic normal-pressure

hydrocephalus for shunt placement: a single-institution experience. J.

Neurosurg. 113, 64–73.

60

Avezaat, C.J., and van Eijndhoven, J.H. (1986). Clinical observations on the

relationship between cerebrospinal fluid pulse pressure and intracranial

pressure. Acta Neurochir (Wien) 79, 13–29.

Avezaat, C.J., van Eijndhoven, J.H., and Wyper, D.J. (1979). Cerebrospinal

fluid pulse pressure and intracranial volume-pressure relationships. J.

Neurol. Neurosurg. Psychiatr. 42, 687–700.

Avolio, A.P., Chen, S.G., Wang, R.P., Zhang, C.L., Li, M.F., and O'Rourke,

M.F. (1983). Effects of aging on changing arterial compliance and left

ventricular load in a northern Chinese urban community. Circulation 68,

50–58.

Balédent, O., Henry-Feugeas, M.C., and Idy-Peretti, I. (2001). Cerebrospinal

fluid dynamics and relation with blood flow: a magnetic resonance study

with semiautomated cerebrospinal fluid segmentation. Invest Radiol 36,

368–377.

Bateman, G.A., Levi, C.R., Schofield, P., Wang, Y., and Lovett, E.C. (2008).

The venous manifestations of pulse wave encephalopathy: windkessel

dysfunction in normal aging and senile dementia. Neuroradiology 50, 491–

497.

Bernstein, M.A., King, K.F., and Zhou, X.J. (2004). Handbook of MRI pulse

sequences (Academic Press).

Bloch, F. (1946). Nuclear induction. Physical Review 70, 460–474.

Boutouyrie, P., Laurent, S., Benetos, A., Girerd, X.J., Hoeks, A.P., and Safar,

M.E. (1992). Opposing effects of ageing on distal and proximal large arteries

in hypertensives. J Hypertens Suppl 10, S87–S91.

61

Bradley, W.G., Scalzo, D., Queralt, J., Nitz, W.N., Atkinson, D.J., and Wong,

P. (1996). Normal-pressure hydrocephalus: evaluation with cerebrospinal

fluid flow measurements at MR imaging. Radiology 198, 523–529.

Bulat, M., and Klarica, M. (2011). Recent insights into a new hydrodynamics

of the cerebrospinal fluid. Brain Res Rev 65, 99–112.

Bullitt, E., Zeng, D., Mortamet, B., Ghosh, A., Aylward, S.R., Lin, W., Marks,

B.L., and Smith, K. (2010). The effects of healthy aging on intracerebral

blood vessels visualized by magnetic resonance angiography. Neurobiol.

Aging 31, 290–300.

Centonze, D., Floris, R., Stefanini, M., Rossi, S., Fabiano, S., Castelli, M.,

Marziali, S., Spinelli, A., Motta, C., Garaci, F.G., et al. (2011). Proposed

chronic cerebrospinal venous insufficiency criteria do not predict multiple

sclerosis risk or severity. Ann. Neurol. 70, 52–59.

Chang, W., Landgraf, B., Johnson, K.M., Kecskemeti, S., Wu, Y., Velikina, J.,

Rowley, H., Wieben, O., Mistretta, C., and Turski, P. (2011). Velocity

measurements in the middle cerebral arteries of healthy volunteers using 3D

radial phase-contrast HYPRFlow: comparison with transcranial Doppler

sonography and 2D phase-contrast MR imaging. American Journal of

Neuroradiology 32, 54–59.

Chen, J.J., Rosas, H.D., and Salat, D.H. (2011). Age-associated reductions in

cerebral blood flow are independent from regional atrophy. Neuroimage 55,

468–478.

Compston, A., and Coles, A. (2002). Multiple sclerosis. Lancet 359, 1221–

1231.

62

Czosnyka, M., Czosnyka, Z.H., Whitfield, P.C., Donovan, T., and Pickard,

J.D. (2001). Age dependence of cerebrospinal pressure-volume

compensation in patients with hydrocephalus. J. Neurosurg. 94, 482–486.

Davson, H., Domer, F.R., and Hollingsworth, J.R. (1973). The mechanism of

drainage of the cerebrospinal fluid. Brain 96, 329–336.

Davson, H., Hollingsworth, G., and Segal, M.B. (1970). The mechanism of

drainage of the cerebrospinal fluid. Brain 93, 665–678.

Desikan, R.S., Ségonne, F., Fischl, B., Quinn, B.T., Dickerson, B.C., Blacker,

D., Buckner, R.L., Dale, A.M., Maguire, R.P., Hyman, B.T., et al. (2006). An

automated labeling system for subdividing the human cerebral cortex on

MRI scans into gyral based regions of interest. Neuroimage 31, 968–980.

Doepp, F., Paul, F., Valdueza, J.M., Schmierer, K., and Schreiber, S.J.

(2010). No cerebrocervical venous congestion in patients with multiple

sclerosis. Ann. Neurol. 68, 173–183.

Doepp, F., Schreiber, S.J., Münster, von, T., Rademacher, J., Klingebiel, R.,

and Valdueza, J.M. (2004). How does the blood leave the brain? A

systematic ultrasound analysis of cerebral venous drainage patterns.

Neuroradiology 46, 565–570.

Doepp, F., Würfel, J.T., Pfueller, C.F., Valdueza, J.M., Petersen, D., Paul, F.,

and Schreiber, S.J. (2011). Venous drainage in multiple sclerosis: a

combined MRI and ultrasound study. Neurology 77, 1745–1751.

Edsbagge, M., Starck, G., Zetterberg, H., Ziegelitz, D., and Wikkelso, C.

(2011). Spinal cerebrospinal fluid volume in healthy elderly individuals. Clin

Anat 24, 733–740.

63

Eide, P.K., and Brean, A. (2006). Intracranial pulse pressure amplitude

levels determined during preoperative assessment of subjects with possible

idiopathic normal pressure hydrocephalus. Acta Neurochir (Wien) 148,

1151–1156; discussion1156.

Eide, P.K., and Brean, A. (2010). Cerebrospinal fluid pulse pressure

amplitude during lumbar infusion in idiopathic normal pressure

hydrocephalus can predict response to shunting. Cerebrospinal Fluid Res 7,

5.

Eklund, A., Smielewski, P., Chambers, I., Alperin, N., Malm, J., Czosnyka,

M., and Marmarou, A. (2007). Assessment of cerebrospinal fluid outflow

resistance. Med Biol Eng Comput 45, 719–735.

Ekstedt, J. (1977). CSF hydrodynamic studies in man. 1. Method of constant

pressure CSF infusion. J. Neurol. Neurosurg. Psychiatr. 40, 105–119.

Ekstedt, J. (1978). CSF hydrodynamic studies in man. 2 . Normal

hydrodynamic variables related to CSF pressure and flow. J. Neurol.

Neurosurg. Psychiatr. 41, 345–353.

Enzmann, D.R., and Pelc, N.J. (1991). Normal flow patterns of intracranial

and spinal cerebrospinal fluid defined with phase-contrast cine MR imaging.

Radiology 178, 467–474.

Enzmann, D.R., and Pelc, N.J. (1993). Cerebrospinal fluid flow measured by

phase-contrast cine MR. AJNR Am J Neuroradiol 14, 1301–1307;

discussion1309–discussion1310.

Fischl, B. (2012). FreeSurfer. Neuroimage E-pub ahead of print.

64

Fischl, B., Salat, D.H., Busa, E., Albert, M., Dieterich, M., Haselgrove, C., van

der Kouwe, A., Killiany, R., Kennedy, D., Klaveness, S., et al. (2002). Whole

brain segmentation: automated labeling of neuroanatomical structures in

the human brain. Neuron 33, 341–355.

Fischl, B., van der Kouwe, A., Destrieux, C., Halgren, E., Ségonne, F., Salat,

D.H., Busa, E., Seidman, L.J., Goldstein, J., Kennedy, D., et al. (2004).

Automatically parcellating the human cerebral cortex. Cereb. Cortex 14, 11–

22.

Fjell, A.M., Walhovd, K.B., Fennema-Notestine, C., McEvoy, L.K., Hagler,

D.J., Holland, D., Brewer, J.B., and Dale, A.M. (2009a). One-year brain

atrophy evident in healthy aging. J. Neurosci. 29, 15223–15231.

Fjell, A.M., Westlye, L.T., Amlien, I., Espeseth, T., Reinvang, I., Raz, N.,

Agartz, I., Salat, D.H., Greve, D.N., Fischl, B., et al. (2009b). High

consistency of regional cortical thinning in aging across multiple samples.

Cereb. Cortex 19, 2001–2012.

Fridén, H.G., and Ekstedt, J. (1983). Volume/pressure relationship of the

cerebrospinal space in humans. Neurosurgery 13, 351–366.

Frydrychowicz, A., Landgraf, B., Wieben, O., and Francois, C.J. (2010).

Scimitar Syndrome: Added Value by Isotropic Flow-Sensitive Four-

Dimensional Magnetic Resonance Imaging With PC-VIPR (Phase-Contrast

Vastly Undersampled Isotropic Projection Reconstruction). Circulation 121,

e434–e436.

Greitz, D. (1993). Cerebrospinal fluid circulation and associated intracranial

dynamics. A radiologic investigation using MR imaging and radionuclide

cisternography. Acta Radiol Suppl 386, 1–23.

65

Greitz, D., Wirestam, R., Franck, A., Nordell, B., Thomsen, C., and Ståhlberg,

F. (1992). Pulsatile brain movement and associated hydrodynamics studied

by magnetic resonance phase imaging. The Monro-Kellie doctrine revisited.

Neuroradiology 34, 370–380.

Gu, T., Korosec, F.R., Block, W.F., Fain, S.B., Turk, Q., Lum, D., Zhou, Y.,

Grist, T.M., Haughton, V., and Mistretta, C.A. (2005). PC VIPR: a high-

speed 3D phase-contrast method for flow quantification and high-resolution

angiography. AJNR Am J Neuroradiol 26, 743–749.

Hall, J.E., and Guyton, A.C. (2010). Guyton and Hall Textbook of Medical

Physiology (Saunders).

Huang, T.-Y., Chung, H.-W., Chen, M.-Y., Giiang, L.-H., Chin, S.-C., Lee, C.-

S., Chen, C.-Y., and Liu, Y.-J. (2004). Supratentorial cerebrospinal fluid

production rate in healthy adults: quantification with two-dimensional cine

phase-contrast MR imaging with high temporal and spatial resolution.

Radiology 233, 603–608.

Hussey, F., Schanzer, B., and Katzman, R. (1970). A simple constant-

infusion manometric test for measurement of CSF absorption. II. Clinical

studies. Neurology 20, 665–680.

Irani, D.N. (2009). Cerebrospinal Fluid in Clinical Practice (W B Saunders

Company).

Johnson, K.M., and Markl, M. (2010). Improved SNR in phase contrast

velocimetry with five-point balanced flow encoding. Magn. Reson. Med. 63,

349–355.

66

Juniewicz, H., Kasprowicz, M., Czosnyka, M., Czosnyka, Z., Gizewski, S.,

Dzik, M., and Pickard, J.D. (2005). Analysis of intracranial pressure during

and after the infusion test in patients with communicating hydrocephalus.

Physiol. Meas. 26, 1039–1048.

Kahlon, B., Annertz, M., Ståhlberg, F., and Rehncrona, S. (2007). Is

aqueductal stroke volume, measured with cine phase-contrast magnetic

resonance imaging scans useful in predicting outcome of shunt surgery in

suspected normal pressure hydrocephalus? Neurosurgery 60, 124–9–

discussion129–30.

Katzman, R., and Hussey, F. (1970). A simple constant-infusion manometric

test for measurement of CSF absorption. I. Rationale and method. Neurology

20, 534–544.

Kim, J., Thacker, N.A., Bromiley, P.A., and Jackson, A. (2007). Prediction of

the jugular venous waveform using a model of CSF dynamics. AJNR Am J

Neuroradiol 28, 983–989.

Lauterbur, P. (1973). Image Formation by Induced Local Interactions:

Examples Employing Nuclear Magnetic Resonance. Nature 190–191.

Leritz, E.C., Salat, D.H., Williams, V.J., Schnyer, D.M., Rudolph, J.L.,

Lipsitz, L., Fischl, B., McGlinchey, R.E., and Milberg, W.P. (2011). Thickness

of the human cerebral cortex is associated with metrics of cerebrovascular

health in a normative sample of community dwelling older adults.

Neuroimage 54, 2659–2671.

Lotz, J., Meier, C., Leppert, A., and Galanski, M. (2002). Cardiovascular flow

measurement with phase-contrast MR imaging: basic facts and

implementation. Radiographics 22, 651–671.

67

Löfgren, J., Essen, von, C., and Zwetnow, N.N. (1973). The pressure-volume

curve of the cerebrospinal fluid space in dogs. Acta Neurol. Scand. 49, 557–

574.

Lutz, W., Sanderson, W., and Scherbov, S. (2008). The coming acceleration

of global population ageing. Nature 451, 716–719.

Magendie, F. (1842). Recherches anatomique et physiologique sur le liquide

céphalo-rachidien ou cérebro-spinal (Paris).

Magnaes, B. (1989). Clinical studies of cranial and spinal compliance and the

craniospinal flow of cerebrospinal fluid. Br J Neurosurg 3, 659–668.

Malm, J., Jacobsson, J., Birgander, R., and Eklund, A. (2011). Reference

values for CSF outflow resistance and intracranial pressure in healthy

elderly. Neurology 76, 903–909.

Malm, J., Sundström, N., Cesarini, K.G., Edsbagge, M., Kristensen, B.,

Leijon, G., and Eklund, A. (2012). Implementation of a new CSF dynamic

device: a multicenter feasibility study in 562 patients. Acta Neurol. Scand.

125, 199–205.

Marmarou, A., Shulman, K., and LaMorgese, J. (1975). Compartmental

analysis of compliance and outflow resistance of the cerebrospinal fluid

system. J. Neurosurg. 43, 523–534.

Martins, A.N., Wiley, J.K., and Myers, P.W. (1972). Dynamics of the

cerebrospinal fluid and the spinal dura mater. J. Neurol. Neurosurg.

Psychiatr. 35, 468–473.

68

Maset, A.L., Marmarou, A., Ward, J.D., Choi, S., Lutz, H.A., Brooks, D.,

Moulton, R.J., DeSalles, A., Muizelaar, J.P., and Turner, H. (1987). Pressure-

volume index in head injury. J. Neurosurg. 67, 832–840.

Mitchell, G.F. (2008). Effects of central arterial aging on the structure and

function of the peripheral vasculature: implications for end-organ damage.

J. Appl. Physiol. 105, 1652–1660.

Mitchell, G.F., van Buchem, M.A., Sigurdsson, S., Gotal, J.D., Jonsdottir,

M.K., Kjartansson, Ó., Garcia, M., Aspelund, T., Harris, T.B., Gudnason, V.,

et al. (2011). Arterial stiffness, pressure and flow pulsatility and brain

structure and function: the Age, Gene/Environment Susceptibility--

Reykjavik study. Brain 134, 3398–3407.

Mokri, B. (2001). The Monro-Kellie hypothesis: applications in CSF volume

depletion. Neurology 56, 1746–1748.

Moran, P.R. (1982). A flow velocity zeugmatographic interlace for NMR

imaging in humans. Magn Reson Imaging 1, 197–203.

Nayler, G.L., Firmin, D.N., and Longmore, D.B. (1986). Blood flow imaging

by cine magnetic resonance. J Comput Assist Tomogr 10, 715–722.

O'Rourke, M.F., and Hashimoto, J. (2007). Mechanical factors in arterial

aging: a clinical perspective. J. Am. Coll. Cardiol. 50, 1–13.

Purcell, E., Torrey, H., and Pound, V. (1946). Resonance Absorption by

Nuclear Magnetic Moments in a Solid (Physical review).

Purves, D., Augustine, G.J., Fitzpatrick, D., Hall, W.C., LaMantia, A.-S., and

White, L.E. (2011). Neuroscience (Sinauer Associates Incorporated).

69

Raabe, A., Czosnyka, M., Piper, I., and Seifert, V. (1999). Monitoring of

intracranial compliance: correction for a change in body position. Acta

Neurochir (Wien) 141, 31–36; discussion35–discussion36.

Redzic, Z.B., and Segal, M.B. (2004). The structure of the choroid plexus and

the physiology of the choroid plexus epithelium. Adv. Drug Deliv. Rev. 56,

1695–1716.

Schubert, T., Santini, F., Stalder, A.F., Bock, J., Meckel, S., Bonati, L., Markl,

M., and Wetzel, S. (2011). Dampening of blood-flow pulsatility along the

carotid siphon: does form follow function? American Journal of

Neuroradiology 32, 1107–1112.

Shapiro, K., Marmarou, A., and Shulman, K. (1980). Characterization of

clinical CSF dynamics and neural axis compliance using the pressure-volume

index: I. The normal pressure-volume index. Ann. Neurol. 7, 508–514.

Singh, A.V., and Zamboni, P. (2009). Anomalous venous blood flow and iron

deposition in multiple sclerosis. J Cereb Blood Flow Metab 29, 1867–1878.

Singh-Manoux, A., Kivimaki, M., Glymour, M.M., Elbaz, A., Berr, C.,

Ebmeier, K.P., Ferrie, J.E., and Dugravot, A. (2012). Timing of onset of

cognitive decline: results from Whitehall II prospective cohort study. Bmj

344, d7622.

Spilt, A., Box, F.M.A., van der Geest, R.J., Reiber, J.H.C., Kunz, P., Kamper,

A.M., Blauw, G.J., and van Buchem, M.A. (2002). Reproducibility of total

cerebral blood flow measurements using phase contrast magnetic resonance

imaging. J Magn Reson Imaging 16, 1–5.

70

Stoquart-ElSankari, S., Lehmann, P., Villette, A., Czosnyka, M., Meyer, M.-

E., Deramond, H., and Balédent, O. (2009). A phase-contrast MRI study of

physiologic cerebral venous flow. J Cereb Blood Flow Metab 29, 1208–1215.

Sundström, N., Andersson, K., Marmarou, A., Malm, J., and Eklund, A.

(2010). Comparison between 3 infusion methods to measure cerebrospinal

fluid outflow conductance. J. Neurosurg. 113, 1294–1303.

Tain, R.-W., Bagci, A.M., Lam, B.L., Sklar, E.M., Ertl-Wagner, B., and

Alperin, N. (2011). Determination of cranio-spinal canal compliance

distribution by MRI: Methodology and early application in idiopathic

intracranial hypertension. J Magn Reson Imaging 34, 1397–1404.

Tan, I.L., van Schijndel, R.A., Pouwels, P.J., van Walderveen, M.A.,

Reichenbach, J.R., Manoliu, R.A., and Barkhof, F. (2000). MR venography

of multiple sclerosis. AJNR Am J Neuroradiol 21, 1039–1042.

Tang, C., Blatter, D.D., and Parker, D.L. (1995). Correction of partial-volume

effects in phase-contrast flow measurements. J Magn Reson Imaging 5, 175–

180.

Tans, J.T., and Poortvliet, D.C. (1989). Relationship between compliance

and resistance to outflow of CSF in adult hydrocephalus. J. Neurosurg. 71,

59–62.

Tisserand, D.J., van Boxtel, M.P.J., Pruessner, J.C., Hofman, P., Evans, A.C.,

and Jolles, J. (2004). A voxel-based morphometric study to determine

individual differences in gray matter density associated with age and

cognitive change over time. Cereb. Cortex 14, 966–973.

71

Tsunoda, A., Mitsuoka, H., Bandai, H., Endo, T., Arai, H., and Sato, K.

(2002). Intracranial cerebrospinal fluid measurement studies in suspected

idiopathic normal pressure hydrocephalus, secondary normal pressure

hydrocephalus, and brain atrophy. J. Neurol. Neurosurg. Psychiatr. 73, 552–

555.

Valdueza, J.M., Münster, von, T., Hoffman, O., Schreiber, S., and Einhäupl,

K.M. (2000). Postural dependency of the cerebral venous outflow. Lancet

355, 200–201.

Westerhof, N., Lankhaar, J.-W., and Westerhof, B.E. (2009). The arterial

Windkessel. Med Biol Eng Comput 47, 131–141.

Wigström, L., Sjöqvist, L., and Wranne, B. (1996). Temporally resolved 3D

phase-contrast imaging. Magn. Reson. Med. 36, 800–803.

Wolf, R.L., Ehman, R.L., Riederer, S.J., and Rossman, P.J. (1993). Analysis

of systematic and random error in MR volumetric flow measurements.

Magn. Reson. Med. 30, 82–91.

Zamboni, P., Galeotti, R., Menegatti, E., Malagoni, A.M., Gianesini, S.,

Bartolomei, I., Mascoli, F., and Salvi, F. (2009a). A prospective open-label

study of endovascular treatment of chronic cerebrospinal venous

insufficiency. Journal of Vascular Surgery 50, 1348–1358.e3.

Zamboni, P., Galeotti, R., Menegatti, E., Malagoni, A.M., Tacconi, G.,

Dall'Ara, S., Bartolomei, I., and Salvi, F. (2009b). Chronic cerebrospinal

venous insufficiency in patients with multiple sclerosis. J. Neurol.

Neurosurg. Psychiatr. 80, 392–399.

72

Zamboni, P., Menegatti, E., Weinstock-Guttman, B., Schirda, C., Cox, J.L.,

Malagoni, A.M., Hojanacki, D., Kennedy, C., Carl, E., Dwyer, M.G., et al.

(2009c). The severity of chronic cerebrospinal venous insufficiency in

patients with multiple sclerosis is related to altered cerebrospinal fluid

dynamics. Funct. Neurol. 24, 133–138.

Zivadinov, R., Marr, K., Cutter, G., Ramanathan, M., Benedict, R.H.B.,

Kennedy, C., Elfadil, M., Yeh, A.E., Reuther, J., Brooks, C., et al. (2011).

Prevalence, sensitivity, and specificity of chronic cerebrospinal venous

insufficiency in MS. Neurology 77, 138–144.