Tải bản đầy đủ - 0 (trang)
Notes (Troubles�hooting, Tips, and Tricks)

Notes (Troubles�hooting, Tips, and Tricks)

Tải bản đầy đủ - 0trang


Visualisation of Thermal Changes in Freely Moving Animals


Fig. 3. Same thermal image displayed with rainbow, iron, and grey colour palettes. Image acquired and processed with the

same equipment as in Fig. 1b, and cropped with standard image editor software. Notice that the difference between interscapular and lumbar back temperature is more evident in the left (rainbow), and less evident in the right (grey) image.

3. Grey is used mainly for print. It gives little contrast, which

makes it difficult to spot small differences of temperature.

Moreover, at times seeing the profile of the objects of interest

is also made difficult. Usually black is used for the coldest temperature, white for the warmest, and shades of grey for everything in between.

4. Rainbow is the palette that gives most contrast, making it easiest to discern each temperature in a scale. Light red is at the

top of the temperature scale, and dark blue at its bottom. It is

the palette of choice for use during an experimental session, for

it will allow the experimenter to easily discern temperatures

with the naked eye. It is usually very good for colour prints, as

long as the objects in the picture are more or less uniform in

their temperature and the image is not too busy. Complex

objects with too many colours being displayed at once might

become too difficult to interpret. The main problem with the

rainbow palette is that it becomes uninterpretable when printed

in black and white.

5. Iron is the most versatile palette. Light yellow indicates the

highest temperature, and dark blue the lowest. It gives more

contrast than black and white, yet when printed in monochrome, all temperatures will be rendered correctly, making it

an ideal option for publishing. It is also the least confusing

when dealing with busy images.

6. The tail and paws are the most difficult regions to see when

vasoconstricted and shot against a cold background. While it is

impossible to avoid missing data in these conditions, there are

steps that could be used to minimise it. Make sure the focus is

on the tail in the beginning of the experiment, so it can be seen

even when its temperature is close to that of the background.

Rough bedding, urine, and faecal boli will often make an


D.M.L. Vianna and P. Carrive

uneven thermal background, and an interruption of the

expected shape of those sometimes gives away a cold tail that

would be invisible otherwise (see right image, Fig. 1b).

7. Infrared gun. A cheap alternative to a digital infrared camera is

an infrared gun. This is a simple device that works according to

the same principles as an infrared camera, except that it will

give only a single temperature measure for a restricted field

instead of an array of measures. The minimum requirement for

it to be used in physiological research is to have an inbuilt laser

pointer that will show the experimenter the exact limits of the

field being measured. When purchasing, make sure the working distance of the device matches the distance to the object

during the experiment.

8. Sensitivity vs. accuracy. Most digital infrared cameras will give

a very acceptable sensitivity of less than 0.1°C, which means

that relative changes within an experiment can be detected

with a ±0.1°C error. This is sufficient resolution to study physiological responses. On the other hand, accuracy is not its

strength, so a measured value of 32°C might refer to a real

temperature of 30–34°C (accuracy being in the range of ±2°C).

Hence, if absolute temperature measurements, rather than

relative changes, are critical for an experiment, one would

probably be safer to confirm the measurement made by the

infrared camera with a conventional thermocouple. Regular

calibration of the camera in this case is also important.


The authors wish to thank Dr. Marcio L. Vianna for his advice

regarding the physical principles mentioned in the text.


1. Habler HJ, Stegmann JU, Timmermann L,

Janig W (1998) Functional evidence for the

differential control of superficial and deep

blood vessels by sympathetic vasoconstrictor

and primary afferent vasodilator fibres in rat

hairless skin. Exp Brain Res 118:230–234

2. Marks A, Vianna DM, Carrive P (2009)

Nonshivering thermogenesis without interscapular brown adipose tissue involvement

during conditioned fear in the rat. Am J

Physiol Regul Integr Comp Physiol 296:


3. Vianna DM, Carrive P (2005) Changes in cutaneous and body temperature during and after





conditioned fear to context in the rat. Eur J

Neurosci 21:2505–2512






Thermoregulatory competence and behavioral

expression in the young of altricial species–

revisited. Dev Psychobiol 33:107–123

Blumberg MS, Stolba MA (1996) Thermogenesis,

myoclonic twitching, and ultrasonic vocalization

in neonatal rats during moderate and extreme

cold exposure. Behav Neurosci 110:305–314

Gordon CJ (1990) Thermal biology of the

laboratory rat. Physiol Behav 47:963–991

Blessing WW (2003) Lower brainstem pathways regulating sympathetically mediated


Visualisation of Thermal Changes in Freely Moving Animals

changes in cutaneous blood flow. Cell Mol

Neurobiol 23:527–538

8. Hayward JS, Ball EG (1966) Quantitative

aspects of brown adipose tissue thermogenesis

during arousal from hibernation. Biol Bull


9. Farrell WJ, Alberts JR (2000) Ultrasonic vocalizations by rat pups after adrenergic manipulations of brown fat metabolism. Behav Neurosci


10. Blumberg MS, Efimova IV, Alberts JR (1992)

Thermogenesis during ultrasonic vocalization

by rat pups isolated in a warm environment: a

thermographic analysis. Dev Psychobiol



11. Benoist JM, Pincede I, Ballantyne K, Plaghki

L, Le Bars D (2008) Peripheral and central

determinants of a nociceptive reaction: an

approach to psychophysics in the rat. PLoS

One 3:e3125

12. Carrive P, Churyukanov M, Le Bars D (2011)

A reassessment of stress-induced “analgesia” in

the rat using an unbiased method. Pain


13. Cerri M, Zamboni G, Tupone D, Dentico D,

Luppi M, Martelli D, Perez E, Amici R (2010)

Cutaneous vasodilation elicited by disinhibition of the caudal portion of the rostral ventromedial medulla of the free-behaving rat.

Neuroscience 165(3):984–995

Chapter 13

Perfusion Magnetic Resonance Imaging

Quantification in the Brain

Fernando Calamante


Cerebral perfusion, the rate of blood delivery to brain tissue, plays an important role in tissue viability and

brain function. The most commonly used magnetic resonance imaging (MRI) method to assess brain

perfusion and tissue haemodynamics in clinical investigations is known as dynamic susceptibility-contrast

(DSC) MRI. Among the main reasons for its widespread use are its fast acquisition time, good contrastto-noise ratio, and wealth of information available from DSC-MRI data. A description of the typical steps

involved in a DSC-MRI study, the practical decisions that need to be taken, the problems that can be

encountered, and the approaches that have been developed to overcome or minimise them will be the

topic of this chapter. In particular, the implications of all these issues for perfusion quantification will be


Key words: Perfusion, Cerebral blood flow, Magnetic resonance imaging, Contrast agent,

Arterial input function, Deconvolution, Mean transit time, Cerebral blood volume, Dynamic susceptibility contrast

1. Introduction

Cerebral perfusion (also known as cerebral blood flow, CBF)1 refers

to the rate of blood delivery to brain tissue. It is the blood flow at

the capillary level, where exchange of oxygen and nutrients, and

the removal of waste products, take place. Therefore, perfusion

plays an important role both in tissue viability and in brain



In the context of Perfusion MRI, the terms perfusion, cerebral blood flow, and its acronym CBF are commonly used

indistinguishably. This convention will be also used throughout this chapter.

Emilio Badoer (ed.), Visualization Techniques: From Immunohistochemistry to Magnetic Resonance Imaging, Neuromethods,

vol. 70, DOI 10.1007/978-1-61779-897-9_13, © Springer Science+Business Media, LLC 2012



F. Calamante

Fig. 1. Dynamic susceptibility-contrast (DSC)-magnetic resonance imaging (MRI) data on a patient with a vascular abnormality in the right anterior, posterior, and middle cerebral arteries (left side of the images). The patient had no infarctions

at the time of the MRI examination. The nine images in the top row are nine time samples of the gradient-echo echo-planar

images acquired during the passage of the bolus of contrast agent (TR = 1.5 s). These images show the transient decrease

in signal intensity induced by the passage of the bolus. Note the asymmetric behaviour due to the abnormalities on the

right major cerebral arteries. The bottom graph shows the signal intensity time course for a region of interest in the contralateral cortical gray matter (see inset). Three different periods can be observed: the baseline (before the arrival of the

bolus to the region, approximately from 0 to 14 s), the first passage (approximately from 14 to 26 s), and the recirculation

(approximately for times >26 s) (figure reproduced from Calamante (86), with permission from John Wiley & Sons, Inc).

Dynamic susceptibility-contrast (DSC) magnetic resonance

imaging (MRI), also known as bolus-tracking MRI, is currently

the most commonly used MRI method to assess brain perfusion

and tissue hemodynamics in clinical investigations. One of the

main reasons for its widespread use is that bolus-tracking MRI data

can be acquired very quickly (in approximately 1 min acquisition

time), while still retaining good contrast-to-noise ratio (1).

DSC MRI involves an intravenous bolus injection of a paramagnetic contrast agent (typically a gadolinium-based contrast

agent), and the rapid measure of the signal changes during its passage through the brain. The bolus of contrast agent induces a transient signal-drop on T2*-weighted images (Fig. 1) (and see

methods), which can be used to infer the time-dependent concentration of the contrast agent (Fig. 2). Quantification of perfusion

using DSC-MRI involves measurement of the so-called arterial

input function (AIF, which describes the contrast agent input to


Perfusion Magnetic Resonance Imaging Quantification in the Brain


Fig. 2. Concentration time course on a patient with vascular abnormality in the right major cerebral arteries (same patient

as in Fig. 1). The nine images in the top row are nine time samples of the concentration of the contrast agent during the

passage of the bolus (TR = 1.5 s); the images show the transient increase in contrast concentration with the arrival of the

bolus. Note the asymmetric behaviour due to the abnormalities on the right major cerebral arteries. The bottom graph

shows the concentration time course for a region of interest in the contralateral cortical gray matter (see inset). As expected,

the concentration is zero before the arrival of the bolus (during the baseline period) (figure reproduced from Calamante

(86), with permission from John Wiley & Sons, Inc).

the tissue of interest), and a deconvolution analysis to remove,

from the tissue concentration time course, the temporal spread

contribution associated with the AIF (2). There are, however,

many issues regarding the potential of DSC-MRI to accurately

quantify perfusion (e.g. see (3, 4)), which will be discussed in this


2. Methods

A typical DSC-MRI study involves the following steps:

1. Selecting an appropriate image acquisition protocol.

2. Selecting an appropriate contrast agent injection protocol.

3. Assessing the need of motion correction as a pre-processing step.


F. Calamante

4. Estimating the time-dependent contrast agent concentration.

5. Measuring the AIF.

6. Performing the deconvolution analysis to remove the AIF


7. Quantifying the hemodynamic parameters of interest.

8. Scaling the measurements appropriately, if values in absolute

units are required.

A description of each of these steps, the practical decisions that

need to be taken, the problems that can be encountered, and the

approaches that have been developed to overcome or minimise

them, will be discussed. In particular, the implications of all these

issues for perfusion quantification will be emphasised.

2.1. Selecting an

Appropriate Image

Acquisition Protocol

The most commonly used data-acquisition sequence for DSC-MRI

data is based on T2*-weighted imaging (i.e. a gradient-echo based

method). In particular, since the transit time of the blood through

the vasculature is only of the order of a few seconds, a fast T2*weighted imaging method, such as gradient-echo echo-planar

imaging (EPI (5)), is typically employed. This imaging technique

allows acquisition of 15–20 slices with a time resolution of approximately 1.5 s, thus providing sufficient temporal resolution to characterise the bolus passage with good brain coverage. However,

even with a fast imaging sequence such as EPI, there are a number

of key parameters that must be chosen as a compromise between

spatial resolution, temporal resolution, spatial coverage, and signal-to-noise ratio (SNR). The most relevant parameters are: the

echo time (TE), the repetition time (TR), the flip angle (q), the

spatial resolution, and the number of slices. A set of recommendations was recently published by an international consortium of

experts as part of the “Acute Stroke Imaging Research Roadmap”

(6). In particular, it was recommended that TE = 35–45 ms (for

studies performed at 1.5 T) or 25–30 ms (for 3 T studies),

TR £ 1.5–2 s, q = 60–90° (for 1.5 T studies) or 60° (for 3 T studies), in-plane resolution 1.8–2 mm, slice thickness 5 mm, and

whole-brain coverage achieved using ³12 slices. It should be noted

that the choice of flip angle should not only be based on the commonly used Ernst angle criteria (i.e. cosq = exp(−TR/T1)), but it

should be chosen to avoid introducing T1-enhancement effects

(7). In practice, the flip angle should therefore be slightly smaller

than the Ernst angle (see Note 1).

Other methods based on T2-weighted imaging (i.e. a spin-echo

based acquisition), or multi-echo sequences have also been

employed, although they are much less common. Similarly, alternative imaging methods to EPI have been used to achieve, for

example, true full-brain coverage or higher spatial resolution. See

Note 2 for a description of some of these alternative imaging methods used in DSC-MRI.


Perfusion Magnetic Resonance Imaging Quantification in the Brain


When data quality is very poor (e.g. very noisy data due to a

sub-optimal choice of TE or bolus dose), the use of image denoising as a pre-processing step may be the only alternative in order to

achieve usable results (8, 9). Various denoising methods have been

used, from simple methods which denoise each image volume

independently to more rigorous methods that consider DSC-MRI

as a four-dimensional (three-dimensional space + time) data (8–

10). However, any denoising method has limited power and,

whenever possible, the imaging protocol should be optimised to

increase data quality during the acquisition stage.

2.2. Selecting an

Appropriate Contrast

Agent Injection


Once the MRI protocol has been chosen, the next important step

in DSC-MRI is the selection of an appropriate contrast agent injection protocol. The most relevant factors to consider in this respect

are the contrast agent material, the dose, the injection site, access

and rate, as well as the bolus flush. Once again, the previously

mentioned international consortium has compiled very good

guidelines regarding many of these factors (6). In particular, they

have recommended the use of a standard gadolinium-based contrast agent, with a so-called “single” dose (for half-molar agents,

corresponding to 0.1 mmoL/kg or 0.2 mL/kg body weight),

administered by an intra-venous injection (preferably in the antecubital vein), using a 18–20 gauge line, at a rate of 4–6 mL/s using

an MR-compatible power injector; this bolus should be followed

by a 20–40 mL of saline flush. Flushing the bolus of contrast agent

with saline (preferably with the same injection rate) is important to

minimise the spread of the bolus of contrast agent (due to the low

flow in veins).

It is important to note that these gadolinium-based contrast

agents effectively behave as intravascular contrast agents, provided

the blood–brain barrier (BBB, which is a physical barrier that limits

the transport of substances from the blood into the central nervous

system) remains intact. This assumption is essential for the validity

of the common quantification kinetic model (see Sect. 2.6


In recent years, there has been increasing concerns with some

potential adverse reactions to contrast agents (11). Several studies

have suggested a possible association between gadolinium-based

contrast agents and an increased risk of nephrogenic systemic

fibrosis (NSF), particularly in patients with renal insufficiency, or

with glomerular filtration rate (GFR) < 30 mL/min/1.73 m2 (11).

This has led to changes in the investigation of these high-risk

patient groups, for which the risk–benefit of DSC-MRI studies

should be carefully considered. Furthermore, not all contrast

agents were found to have the same association to increased risk,

and the choice of contrast agent should be carefully considered

(see Note 3 for more details).


F. Calamante

2.3. Assessing the

Need of Motion

Correction as a

Pre-processing Step

Although DSC-MRI has a relatively fast acquisition time (1–2 min),

its data can suffer from motion-related artefacts. These artefacts do

not manifest in the same way as in other conventional MRI methods (e.g. as image ghosting) because a single-shot imaging method

(such as EPI) is commonly used. It is the dynamic nature of DSCMRI that is affected by motion instead: if motion is present, different image volumes in the time-series will not match in space. It is

therefore essential to assess the data for the presence of motion,

particularly for the image volumes acquired in and around the passage of the bolus.

The most common source of motion is subject movement at

the time of the injection. If the motion is primarily in-plane, standard motion-correction strategies have been shown to eliminate

the error (12). On the other hand, if the motion has a significant

through-slice component, these strategies will not work; they are

not able to correct for the spin history effects due to the multi-slice

two-dimensional acquisition strategies commonly employed in

DSC-MRI. The use of a three-dimensional based sequence (e.g.

see 3D-PRESTO MRI sequence in Note 2 below) could offer a

viable solution; these 3D sequences may be worth considering if

very ill or uncooperative subjects are studied.

2.4. Estimating the


Contrast Agent


Quantification of perfusion using DSC-MRI requires knowledge

of the time-dependent concentration of the contrast agent (see

step 2.6 below). However, DSC-MRI does not measure the concentration directly, and its value is estimated indirectly from the

effect the contrast agent has on the image signal intensity. The

most commonly used model is based on a linear relationship

between the change in the relaxation rate DR2* (where R2* = 1/T2*)

and the contrast agent concentration (13):

C (t ) = k · ΔR2* (t ) = k · (R2* (t ) − R2* (0)) = −

⎛ S (t ) ⎞


· ln ⎜


⎝ S (0) ⎟⎠


where C(t) is the time-dependent concentration (also referred to as

the “peak”, due to its shape, see Fig. 2), R2*(0) indicates the baseline

relaxation rate (i.e. the relaxation rate without contrast agent), S(t)

is the signal intensity and S(0) its baseline value, and k is a proportionality constant that depends on the type of contrast agent, the

magnetic field strength, the pulse sequence, and the tissue type.

Since a measurement of R2* requires multiple images acquired with

different TEs (and therefore, with an associated increased acquisition time), DSC-MRI data are commonly acquired with a single TE

value. Therefore, it should be noted that (1) not only assumes a

linear relationship between the concentration of contrast agent and

the change in relaxation rate, but it also assumes that the change in

relaxation rate can be estimated from the change in signal intensity

measured from single-echo T2*-weighted images (see right-hand

side of (1)). This, in turn, assumes a negligible contribution from T1


Perfusion Magnetic Resonance Imaging Quantification in the Brain


relaxation to the changes in signal intensity. This assumption depends

on several factors (7), including BBB status, sequence parameters

(e.g. TR, q), and sequence type (see also Notes 1 and 8).

The actual value of k cannot be easily measured in practice, and

this has been one of the major criticisms for the accuracy of

Perfusion MRI measurements in absolute units (compared to absolute measurements using Perfusion CT) (4). A discussion about

this issue and its implication for perfusion quantification is included

in Note 4.

2.5. Measuring the

Arterial Input Function

One of the key steps in quantification of DSC-MRI data is the

measurement of the AIF. The information about the contrast agent

input to the tissue of interest is essential to be able to isolate true

tissue information (e.g. cerebral perfusion) from other non-tissue

related contributions, such as the specific injection protocol, cardiac output, and macro-vascular structure. For example, a wider

tissue concentration time course (i.e. a wider “peak”) could be due

to any one, or a combination, of the following conditions: a lower

tissue perfusion, and/or a slower injected bolus, and/or a larger

injected bolus, and/or a slower cardiac output, and/or a longer

transit route of the bolus through a collateral vascular path. The

key role of the AIF is to help isolate, from all these contributions,

the contribution due to tissue perfusion.

Due to practical limitations, a single AIF is commonly used for

each slice (or even for the whole brain). This so-called “global

AIF” is usually measured from the signal changes in and around a

major artery, such as the middle cerebral artery (14). The basic

assumption is that such a global AIF provides a reasonable representation of the input of the bolus to brain tissue, and contains

information regarding the contribution of the injection conditions,

the cardiac output, and the macro-vascular structure. However,

the perceptive reader would anticipate that this assumption can

lead to errors in perfusion quantification. In fact, the measurement

of AIF is one of the most important possible sources of error in

DSC-MRI quantification. The measured AIF can be subject to a

number of possible artefacts, including: partial-volume effects (see

Note 5), bolus delay and dispersion (see Note 6), non-linear effects

with contrast-agent concentration (see Note 4), and truncation

artefacts (see Note 7).

While the AIF used to have to be manually measured by experienced users (e.g. by manually identifying voxels with earlier and

narrower “peaks”), a number of automatic methods have been

developed in the last few years (15–18). These methods not only

reduce demands on the need for highly trained local staff, but they

have also contributed to increase the objectivity, speed, and reproducibility of the DSC-MRI measurements; they are therefore highly

recommended when analysing DSC-MRI data. Figure 3 illustrates

the workflow and results from one of these automatic methods.


F. Calamante

Tài liệu bạn tìm kiếm đã sẵn sàng tải về

Notes (Troubles�hooting, Tips, and Tricks)

Tải bản đầy đủ ngay(0 tr)