Magnetic Resonance Insights

Open Access
Full open access to this and
thousands of other papers at
Review
http://www.la-press.com.

Measuring Restriction Sizes Using Diffusion Weighted

Magnetic Resonance Imaging: A Review

Melanie Martin
Department of Physics, University of Winnipeg and Department of Radiology, University of Manitoba.
Corresponding author email: [email protected]

Abstract: This article reviews a new concept in magnetic resonance as applied to cellular and biological systems. Diffusion weighted
magnetic resonance imaging can be used to infer information about restriction sizes of samples being measured. The measurements rely
on the apparent diffusion coefficient changing with diffusion times as measurements move from restricted to free diffusion regimes.
Pulsed gradient spin echo (PGSE) measurements are limited in the ability to shorten diffusion times and thus are limited in restric-
tion sizes which can be probed. Oscillating gradient spin echo (OGSE) measurements could provide shorter diffusion times so smaller
restriction sizes could be probed.

Keywords: magnetic resonance imaging, nuclear magnetic resonance, diffusion, axon diameter, restricted diffusion, porous materials,
central nervous system, oscillating gradients, diffusion-time

Magnetic Resonance Insights 2013:6 59–64

doi: 10.4137/MRI.S11149

This article is available from http://www.la-press.com.

© the author(s), publisher and licensee Libertas Academica Ltd.

This is an open access article published under the Creative Commons CC-BY-NC 3.0 license.

Magnetic Resonance Insights 2013:6 59

Martin

Introduction 180
This short invited article reviews diffusion techniques
used to infer sizes of structures in samples that lead 90 g
to restricted diffusion. It briefly explains the time
dependence of the MRI-measured apparent diffusion
­coefficient. Then it explains how the time dependence
Figure 1. Pulsed gradient spin echo (PGSE) sequence. The basic
is used to infer sizes of structures, such as axon diam- PGSE sequence consists of two RF pulses, shown in blue, and two
eters, in the sample. It then discusses the limitations gradient pulses, shown in black. The diffusion time, ∆, is the time from
the start of the first gradient pulse to the start of the second gradient
of current methods inferring small sizes and briefly pulse.
explains one possible alternative to overcome these
limitations.
phase based on their position at the time of the pulse.
Diffusion A π RF pulse is then applied to the system which
Molecules in a liquid undergo diffusion. Their mean reverses the phase of the spins. Another identical gra-
square displacement depends on the diffusion time, dient pulse is applied to the sample which changes the
∆, as described by Einstein’s relation ,r2. = 2D∆ (in phase of the spins based on their position at the time
one dimension) where D is the diffusion ­coefficient.1 of the second pulse (∆). If no diffusion has occurred,
Molecules diffusing in a uniform medium with no the phase acquired from the second pulse will be
barriers experience unrestricted diffusion. In non- equal and opposite to the phase of the spin just before
­uniform media (eg, porous samples and cellular the pulse, resulting in a net phase of zero. If diffusion
­tissues) barriers restrict molecular displacements so occurs, the mean squared phase of all spins will be
that the diffusion depends on the time scale of the nonzero and cause a loss in MR signal which can be
study and the permeability of the barriers. Assuming used to calculate the ADC.4
Einstein’s relation, ,r2. during the time ∆ can be Alternatively, Callaghan proposed q-space ­imaging
used to find an apparent diffusion coefficient (ADC) using PGSE with short gradient pulses to measure the
that is less than D because of restrictions or hin- conditional probability, P(r′—r, ∆), that a spin initially
drances to motion. Understanding the influence of the at r has migrated to r′ over the time ∆. By measuring
restrictions or hindrances on the ADC gives informa- the signal as a function of gradient strength, one can
tion about the geometry of the boundaries of the sur- calculate the probability distribution of displacements
rounding medium which, for MRI, could be tissues, of spins.5 Using this probability distribution, one can
rocks, concrete, cement, polymers, gases, etc.2 infer from MR images tissue geometric information,
As an example, the ADC in a simple system con- such as axon diameter distributions and axonal and
sisting of molecules with diffusion coefficient D that cellular volume fractions or densities.5–9 It is difficult,
are entirely restricted to a single pore of diameter a however, to find P in white matter because of the vari-
will depend on ∆. For ∆ « a2/(2D), the diffusion will ability of axon diameter.2
appear almost unrestricted. For ∆ » a2/(2D), all mol- In vivo measurements using echo planar imaging
ecules, regardless of their starting position, will be (EPI) sequences are used to acquire images more
found anywhere in the pore.3 Measurements as a func- rapidly.2 This allows full diffusion tensor imaging
tion of ∆ thus provide information about the structure (DTI) in vivo although the resolution is often not
in which the molecules are diffusing. Specifically, as good as standard imaging sequences.10 Perfusion
there will be a change in the ADC when measuring can also be measured using MRI and the effects of
through the time ∆  =  a2/(2D) which can be used to perfusion on diffusion measurements must be taken
determine the pore size. into account.11–14
Traditional MR measurements of the ADC in dif- Many changing cellular structures, due to disease
ferent samples use the Pulsed Gradient Spin Echo or injury, have already been shown to affect ADC
(PGSE) sequence,4 see Figure  1. After excitation, a measurements. Specifically for white matter imaging,
magnetic field gradient pulse is applied to the sam- some have measured changes in DTI metrics with
ple for a short time. This causes the spins to obtain a changing myelin content or myelin damage.2,15–22

60 Magnetic Resonance Insights 2013:6

 Measuring restriction sizes using DWI

Axon Diameter Measurements from Another method, ActiveAx, extended and optimized
Diffusion Weighted MRI these methods to determine the accuracy and preci-
There are several methods for estimating axon diam- sion with which this important new biomarker, axon
eter distributions and densities using MRI and most diameter, can be estimated in live human ­subjects.9 The
use PGSE with ADC(∆) or q-space. One method, work from this group combined a simplified version
AxCaliber, uses a framework that combines com- of CHARMED with high-angular-resolution diffusion
posite hindered and restricted models of water dif- imaging (HARDI) and a model with a single axon diam-
fusion (CHARMED) and PGSE measurements.23 eter.9 It has been modified to be robust in the presence
This method uses one fixed gradient direction and of orientation dispersion.26 The same group has also
multiple diffusion times and gradient strengths combined two-shell HARDI with a three-compartment
to make the CHARMED model more sensitive to tissue model to create neurite orientation dispersion
axon diameter.8 It fits the MR signal to an equation and density imaging (NODDI).27 Table 1 summarizes
which has components due to restricted diffusion, the literature for some of the axon diameter measure-
hindered diffusion, and later free diffusion.24 Water ments made with MRI. Diffusion times used ranged
in each axon size will experience restricted diffu- from 7 to 305 ms measuring diameters from 0.4–14 µm.
sion at different ∆. For example, water in an axon Measuring similar restriction sizes in porous objects at
with a small diameter will experience restriction for room temperature would likely need longer imaging
much smaller ∆ than water in a larger axon.2 Thus by times because of the slower diffusion at the lower room
shortening ∆, smaller and smaller axons move from temperature compared to body temperature.
the restricted component to the hindered component Measuring axon diameter distributions with MRI,
allowing the inference of axon diameter of smaller even with intact ex vivo brains, has advantages over
and smaller axons. traditional ex vivo histological techniques. Ex vivo
The framework used in AxCaliber for dividing the histological techniques are cumbersome, require tissue
models into different types of diffusion was based on sectioning and are subject to inaccuracies such as cell
another model, which describes bovine optic nerve shrinkage.9 With MRI, the brain remains intact and
tissue as a three-compartment system, (axons, glial images can provide measurements over large regions
cells, and extracellular space). Each compartment has of the brain. Because of the difficulties with ex vivo his-
its own diffusion coefficient, size, volume fraction, tological measurements of axon diameter distributions,
membrane permeability, and NMR relaxation times. the variation of fiber composition over the population
It uses multiple ∆s and gradient strengths in a pulse and during development is largely unstudied.9
sequence similar to PGSE (stimulated echo) which Although fixation can affect samples, studies have
allows for longer diffusion times and measures the been done to compare ADC and other DTI metrics
mean axon diameter, but not the distribution.25 between in vivo and fixed ex vivo samples.22,28–30 While

Table 1. A summary of some previous MRI measures of axon diameters. The measured diameters ranged from 0.4 to
14 μm using diffusion times ranging from 7 to 305 ms

Manuscript Sample Mean neuron Mean neuron Diffusion

diameters based diameters based time (ms)
on histology (μm) on diffusion MRI (μm)
Stanisz25 Bovine optic nerve 2–3 2.6 7–29
Assaf23 Excised rat brain tissue Not measured 1–2 and 6–8 35–305
Assaf8 Porcine sciatic nerve 7.3 6.3 9.17–79.17
Assaf8 Porcine optic nerve 3.48 3.74 9.17–79.17
Barazany24 Rat corpus callosum 0–3 0.4–8 9.9–99.9
Alexander9 Fixed monkey brain 1–3 2–8 12.7–23.7
Alexander9 Human corpus callosum 2–6 6–14 15.7–85.3
Zhang26 Human corpus callosum Not measured 6.8–11.7 15.7–85.3
Zhang26 Human centrum semiovale Not measured 6.8–13.3 15.7–85.3

Magnetic Resonance Insights 2013:6 61

Martin

absolute values of ADC change, partially due to the short diffusion times,34 see Figure 2. In this sequence,
temperature of the sample, fractional anisotropy is sim- the trapezoidal gradient pulses of the PGSE sequence
ilar.31 Thus ex vivo measurements of axon diameters on are replaced with sinusoidally varying gradi-
intact fixed brains can still offer important information. ent pulses. Each period of the sine wave acts as a
­diffusion ­weighting so that the magnetic moments
Limits to the Ability to Measure Small are dephased by the first lobe of the sine wave and
Axon Diameters rephased by the second lobe. These sine waves are
The PGSE sequence requires, however, that ∆ be repeated multiple times so that sufficient diffusion
large with respect to the restriction sizes in biologi- weighting can be obtained. The first measurements
cal tissues, thus limiting the information that could were made of surface-to-volume ratios from packed
be obtained from the measurements as will now be beads in water33 and in vegetables.36
explained. The signal from a PGSE sequence is given However, only in the case of unrestricted Brownian
by ln ( S S0 ) = −γ 2δ 2 g 2 ( ∆ − δ 3) D = −bD, where γ is motion can these multiple diffusion-weighting periods
the gyromagnetic ratio of the hydrogen nucleus, δ is the be considered independent.37,38 For this reason, the sig-
duration of the gradient pulse, and g is the amplitude nal attenuation should be described in terms of a fre-
of the gradient pulse. In order to measure the diffusion quency spectrum rather than simply a scalar value (∆).
coefficient, the difference in signal with, S, and with- More complicated gradient modulations that selec-
out, S0, diffusion gradients has to be large enough com- tively sample a narrow frequency domain of the diffu-
pared to the noise. Thus a large b = γ 2δ 2 g 2 ( ∆ − δ 3) sion spectrum have been proposed.39 These provide a
value is needed. Given that small ∆s are desired, and straightforward means of characterizing the diffusion
δ  ,  ∆ for PGSE, the only remaining factor that can spectrum, but are difficult to implement accurately.35
be increased is the gradient strength. With q-space Thus double-sine- and apodized cosine-modulated gra-
imaging, the smallest displacements which can be dient waveforms, which are modified forms of OGSE,
probed depend on the largest q which can be used.8 have been used and are still termed OGSE ­sequences.35
Again, to make a large q, large gradient strengths need In this so-called temporal diffusion spectroscopy, the
to be used. Gradient coils are limited in power they effective diffusion times are changed by varying dif-
can supply. Moreover, for small duration, large ampli- fusion gradient frequencies, and thus a spectrum of
tude gradient pulses, there will be a large change in diffusion rates, which describe the biological tissue
magnetic field experienced by the subject in a short microenvironment, can be measured. These measure-
amount of time. This induces an electric field in the ments are different from q-space imaging where the
subject, which could cause twitching of the skin, pain, propagator is usually defined at a given long diffu-
or interference with the function of the heart or brain.32 sion time and has spatial and directional dependences
Thus, in practice, it is difficult to make mea- over a large range of q-space.5,40 The contrast between
surements at short diffusion times using the PGSE
sequence. PGSE signals from water molecules in
small axons do not change with diffusion time in the 180
typical range used. Thus methods to determine axon 90 g
diameter from these PGSE measurements are insensi-
tive to these small axons. Shortening ∆ would allow
the inference of restriction sizes, for example rodent
sized axon diameters, and surface-to-volume ratios
in samples. However, such measurements presently
cannot be done with PGSE as described above.32,33
Figure 2. Oscillating gradient spin echo (OGSE) sequence. The basic
OGSE sequence consists of two RF pulses, shown in blue, and two
Oscillating Gradient Spin Echo sinusoidally oscillating gradient pulses, shown in black. In theory, the
higher the frequency of the gradient pulses, the smaller the structures
Sequences that can be probed. More complicated OGSE sequences which ­provide
In 1969, an oscillating gradient spin echo (OGSE) a ­straightforward means of characterizing the diffusion spectrum,
change the form of the sine wave gradient pulses to apodized cosine or
sequence was proposed to make measurements at ­double sine waves.35

62 Magnetic Resonance Insights 2013:6

 Measuring restriction sizes using DWI

tissues might be greater at a discrete, moderately high manuscript conclusions: MM. Made critical revisions:
frequency than at low frequencies, as suggested by MM. The author reviewed and approved of the final
theoretical consideration of simple geometries and pre- manuscript.
liminary studies in tumors41 and rat brain.42
Using OGSE sequences might improve axon Funding
­measurements. As Alexander notes: “More significant The author acknowledges support from the Manitoba
improvements may come from replacing the stan- Health Research Council and the Natural Sciences
dard PGSE sequence with other diffusion-sensitive and Engineering Research Council of Canada.
sequences such as oscillating gradient … experiments,
which may be more sensitive to microstructural param- Competing Interests
eters. … Combination of the experiment design with Author(s) disclose no potential conflicts of interest.
these other pulse sequences should allow the a priori
range of axon diameters to extend to include smaller Disclosures and Ethics
diameters. This should provide protocols with sensi- As a requirement of publication the author has provided
tivity to wider ranges and provide more discriminative signed confirmation of compliance with ethical and
axon diameter indices.”9 This same group performed legal obligations including but not limited to compli-
experiments with optimized gradient waveforms (GEN) ance with ICMJE authorship and competing interests
and were able to make axons with smaller radii more guidelines, that the article is neither under consideration
distinguishable with GEN than with PGSE.43,44 They for publication nor published elsewhere, of their com-
still suggest that oscillating waveforms would provide pliance with legal and ethical guidelines concerning
valuable information for in vivo studies.43 Developing human and animal research participants (if applicable),
DTI techniques to visualize better smaller fibers is con- and that permission has been obtained for reproduction
sidered an active area of research45 and creating new of any copyrighted material. This article was subject
techniques using short diffusion times will cause water to blind, independent, expert peer review. The review-
in the larger fibers to become less restricted allowing ers reported no competing ­interests. ­Provenance:
for easier visualization of the restricted diffusion of the author was invited to submit this paper.
water in the smaller fibers.
References
OGSE and Axon Diameter 1. Einstein A. Über die von der molekularkinetischen Theorie der Wärme
geforderte Bewegung von in ruhenden Flüssigkeiten suspendierten Teilchen.
Measurements Ann Physik. 1905;4:549–60.
Extending the measurements made with OGSE from 2. Jones DK, editor. Diffusion Theory, Methods and Application. Oxford; 2011.
3. Price WS. Pulsed-field gradient nuclear magnetic resonance as a tool for
studying the variation of ADC with diffusion time or studying translational diffusion part 1: basic theory. Concept Magn Reson.
frequency to create a method which uses this variation 1997:9(5):299–336.
to infer restriction sizes using the shortest possible dif- 4. Stejskal EO, Tanner JE. Spin diffusion measurements: Spin echoes in the pres-
ence of a time-dependent field gradient. J Chem Phys. 1965;42(1):288–92.
fusion times could allow for the probing of smaller 5. Callaghan PT, Eccles CD, Xia Y. NMR Microscopy of Dynamic
axon diameters. In theory, this method overcomes ­Displacements—k-space and q-space Imaging. J Phys E Sci Instrum.
1988;21:820–2.
problems with high gradient strengths necessary in 6. Basser PJ. Inferring microstructural features and the physiological state of tis-
current MR diffusion methods (PGSE and q-space) sues from diffusion-weighted images. NMR Biomed. 1995;8(7–8):333–44.
to allow smaller restriction sizes to be distinguished.35 7. Cohen Y, Assaf Y. High b-value q-space analyzed diffusion-weighted MRS
and MRI in neuronal tissue—a technical review. NMR Biomed. 2002;
The OGSE and PGSE methods could be used together 15(7–8):516–542.
to distinguish a large range of restriction sizes from 8. Assaf Y, Blumenfeld-Katzir T, Yovel Y, Basser PJ. AxCaliber: a method
for measuring axon diameter distribution from diffusion MRI. Magn Reson
very small (OGSE) to fairly large (PGSE). ­Combining Med. 2008;59(6):1347–54.
the two methods might allow for a more complete 9. Alexander DC, Hubbard PL, Hall MG, et  al. Orientationally invariant
understanding of the geometry of the sample. indices of axon diameter and density from diffusion MRI. Neuroimage.
2010;52(4):1374–89.
10. Basser PJ, Jones DK. Diffusion-tensor MRI: theory, experimental design and
Author Contributions data analysis—a technical review. NMR Biomed. 2002;15(7–8):456–67.
11. Le Bihan D, Breton E, Lallemand D, Grenier P, Cabanis E, Lavaljeantet M.
Conceived and designed the concept: MM. Wrote MR imaging of intravoxel incoherent motions: application to diffusion and
the first draft of the manuscript: MM. Agree with perfusion in neurologic disorders. Radiology. 1986;161(2):401–7.

Magnetic Resonance Insights 2013:6 63

Martin

12. Turner R, Le Bihan D, Chesnick AS. Echo-planar imaging of diffusion and 29. Concha L, Livy DJ, Beaulieu C, Wheatley BM, Gross DW. In vivo diffusion
perfusion. Magn Reson Med. 1991;19(2):247–53. tensor imaging and histopathology of the fimbria-fornix in temporal lobe
13. Bammer R. Basic principles of diffusion-weighted imaging. Eur J Radiol. epilepsy. J Neurosci. 2010;30(3):996–1002.
2003;45(3):169–84. 30. Douaud G, Jbabdi S, Behrens TEJ, et  al. DTI measures in crossing-fibre
14. Le Bihan D. Molecular diffusion nuclear magnetic resonance imaging. areas: increased diffusion anisotropy reveals early white matter alteration in
Magn Reson Q. 1991;7(1):1–30. MCI and mild Alzheimer’s disease. Neuroimage. 2011;55(3):880–90.
15. Odrobina EE, Lam TYJ, Pun T, Midha R, Stanisz GJ. MR properties of 31. Sun SW, Neil JJ, Liang HF, et al. Formalin fixation alters water diffusion
excised neural tissue following experimentally induced demyelination. coefficient magnitude but not anisotropy in infarcted brain. Magn Reson
NMR Biomed. 2005;18(5):277–84. Med. 2005;53(6):1447–51.
16. Underhill HR, Rostomily RC, Mikheev AM, Yuan C, Yarnykh VL. Fast 32. Schaefer DJ, Bourland JD, Nyenhuis JA. Review of patient safety in time-
bound pool fraction imaging of the in vivo rat brain: association with myelin varying gradient fields. J Magn Reson Imaging. 2000;12(1):20–9.
content and validation in the C6  glioma model. Neuroimage. 2011;54(3): 33. Schachter M, Does MD, Anderson AW, Gore JC. Measurements of restricted
2052–65. diffusion using an oscillating gradient spin echo sequence. J Magn Reson.
17. Song SK, Yoshino J, Le TQ, et al. Demyelination increases radial diffusivity 2000;147(2):233–7.
in corpus callosum of mouse brain. Neuroimage. 2005;26(1):132–40. 34. Gross B, Kosfeld R. Anwendung der spin-echo-methode der messung der
18. Budde MD, Xie M, Cross AH, Song SK. Axial diffusivity is the primary selbstdiffusion. Messtechnik. 1969;77:171–7.
correlate of axonal injury in the experimental autoimmune encephalomy- 35. Does MD, Parsons EC, Gore JC. Oscillating gradient measurements of
elitis spinal cord: a quantitative pixelwise analysis. J Neurosci. 2009;29(9): water diffusion in normal and globally ischemic rat brain. Magn Reson Med.
2805–13. 2003;49(2):206–15.
19. Xie M, Tobin JE, Budde MD, et al. Rostrocaudal analysis of corpus callo- 36. Gore JC, Anderson AW, Does MD, et al. The relationship of problems in
sum demyelination and axon damage across disease stages refines diffusion biomedical MRI to the study of porous media. Magn Reson Imag. 2001;19():
tensor imaging correlations with pathological features. J Neuropathol Exp 295–300.
Neurol. 2010;69(7):704–16. 37. Stepisnik J. Analysis of self-diffusion measurements by density matrix
20. Sun SW, Liang HF, Trinkaus K, Cross AH, Armstrong RC, Song SK. ­calculation. Physica. 1981;104B:350–64.
­Noninvasive detection of cuprizone induced axonal damage and demyelina- 38. Stepisnik J. Measuring and imaging of flow by NMR. Prog NMR Spec.
tion in the mouse corpus callosum. Magn Reson Med. 2006;55(2):302–8. 1985;17:187–209.
21. Beaulieu C. The basis of anisotropic water diffusion in the nervous 39. Callaghan PT, Stepisnik J. Frequency-domain analysis of spin motion using
­system—a technical review. NMR Biomed. 2002;15(7–8):435–55. modulated-gradient NMR. J Magn Reson. 1995;117():232–7.
22. Song SK, Sun SW, Ramsbottom MJ, Chang C, Russell J, Cross AH. 40. Mori S, van Zijl PCM. Fiber tracking: principles and strategies—a technical
­Dysmyelination revealed through MRI as increased radial (but unchanged review. NMR in Biomedicine. 2002;15(7–8):468–80.
axial) diffusion of water. Neuroimage. 2002;17(3):1429–36. 41. Gore JC, Xu J, Colvin DC, Yankeelov TE, Parsons EC, Does MD.
23. Assaf Y, Cohen Y. Non-mono-exponential attenuation of water and N-acetyl ­Characterization of tissue structure at varying length scales using temporal
aspartate signals due to diffusion in brain tissue. J Magn Reson. 1998;131(1): diffusion spectroscopy. NMR Biomed. 2010;23(7):745–56.
69–85. 42. Kershaw J, Leuze C, Obata T, Kanno I, Aoki I. Changes to the fractional
24. Barazany F, Basser PJ, Assaf Y. In vivo measurement of axon diameter dis- anisotropy and mean diffusivity of in vivo rat brain measured at short effec-
tribution in the corpus callosum of rat brain. Brain. 2009;132(5):1210–20. tive diffusion-times. Proc Intl Soc Mag Reson Med. 2011;19:409.
25. Stanisz GJ, Szafer A, Wright GA, Henkelman RM. An analytical model of 43. Siow B, Drobnjak I, Chatterjee A, Lythgoe MF, Alexander DC. ­Estimation
restricted diffusion in bovine optic nerve. Magn Reson Med. 1997;37(1): of pore size in a microstructure phantom using the optimized gradient
103–11. waveform diffusion weighted NMR sequence. J Magn Reson. 2012;214(1):
26. Zhang H, Hubbard PL, Parker GJM, Alexander DC. Axon diameter ­mapping 51–60.
in the presence of orientation dispersion with diffusion MRI. Neuroimage. 44. Drobnjack I, Siow B, Alexander DC. Optimizing gradient waveforms
2011;56(3):1301–15. for microstructure sensitivity in diffusion-weighted MR. J Magn Reson.
27. Zhang H, Schneider T, Wheeler-Kinshot CA, Alexander DC. NODDI: 2010;206(1):41–51.
­Practical in vivo neurite orientation dispersion and density imaging of the 45. Rane S, Duong TQ. Comparison of in vivo and ex vivo diffusion tensor
human brain. Neuroimage. 2012;61(4):1000–16. imaging in rhesus macaques at short and long diffusion times. Open Neu-
28. Schmierer K, Wheeler-Kingshot CAM, Tozer DJ, et  al. Quantitative roimag J. 2011;5:172–8.
­magnetic resonance of postmortem multiple sclerosis brain before and after
fixation. Magn Reson Med. 2008;59(2):268–77.

64 Magnetic Resonance Insights 2013:6