64

One of the principle advantages of MagneticResonance Imaging (MRI) over other diagnos-tic imaging modalities is the intrinsically largebandwidth of soft tissue contrast. Traditionally,clinical imaging exploits the contrast due todifferences in spin density (ρ), spin-latticerelaxation (T1), and spin-spin relaxation (T2)between normal tissues and pathology.

However, other contrast mechanisms such assusceptibility, magnetization transfer, chemicalshift, diffusion, perfusion, flow, and endogenousand exogenous agents can be used to createspecific and subtle changes in contrast. One ofthese mechanisms, magnetization transfer, hasbeen actively researched during the last coupleof years, following Wolff and Balaban’spublication of their first results4.

Magnetization Transfer Contrast (MTC) inmagnetic resonance imaging is the result ofselectively observing the interaction of bulkwater protons with the protons contained inmacromolecules of tissue.

Magnetization Transfer (MT) is based onwell-defined biophysical and biochemicalproperties. Since different tissues have differentmacromolecular compositions, the degree ofinteraction, or Magnetization Transfer (MT),can differ widely, generating very high tissuecontrast.

MTC is obtained by combining saturationtransfer techniques with standard MRI proce-dures. The basic use of the saturation transfertechnique in MR spectroscopy was alreadydescribed in 1963 by Forsén and Hoffman1, whoused it to calculate the exchange rate of specificchemical reactions. In the 1970’s Edzes andSamulski2, 3 proposed that cross-relaxationbetween 1H pools was a dominant relaxationmechanism in biological systems such ascollagen and muscle. However, it was not until1989 that Wolf and Balaban4 became the firstto apply these principles to in vivo measure-ments. They calculated the pseudo-first-orderrate constant for the transfer of magnetization

in kidney and skeletal muscle, and showedimages where the contrast was a function ofmagnetization transfer. In their first publication4

they predicted that: ‘Since the image contrastgenerated using this method is specific to theexchange between mobile and restrictedprotons, this technique may prove valuable indiagnosis or characterization of cancer, oedema,or other pathologies where the specific relaxa-tion mechanism may be useful in determiningthe nature of the disease’.

More experiments5 were performed on a 4.7T system and substantiated the original findings.In 1991, the first magnetization transfer imagesof a cat head at 1.5 T were produced by Wolf,Eng, and Balaban6. Their conclusion that ‘thistechnique for increasing tissue contrast has the potential to be applied to imaging humanbeings at clinically relevant magnetic fields’ canbe considered as the beginning of a steady stream of scientific papers on MR physics andclinical applications of magnetization transfercontrast. A number of different review papershave recently been published on the variousaspects of MTC7, 10. In this review, simplifiedschematic drawings are used to explain thephysical properties of MTC for those readerswho are less specialized in mathematics andphysics.

MR physics of magnetization transfer: relaxa-tion mechanismsIn biological systems, hydrogen nuclei or pro-tons (1H) can be described as existing in twopools. The so-called ‘free pool’ consists of relatively mobile protons (correlation times ≈3 . 10-12 s) in free bulk water and some lipid-containing tissues. This pool has a narrowspectral line (10-100 Hz) and a relatively longT2 (> 10-100 ms), and is labelled in mathema-tical descriptions as the A pool. With standardMRI this pool provides the bulk of the signal.

The second pool or ‘bound pool’ consists ofrestricted protons bound in proteins, other large

Magnetization transfer contrastPart 1: MR physics

R.W. de Boer1)

1) Philips Medical Systems, Best, the Netherlands.

65

macromolecules, and cellular membranes. Thispool has a very broad spectral line (10-50 kHz)and subsequently a very short T2 (< 0.1 ms), andis labelled in mathematical descriptions as theB pool. With conventional MRI this pool is not‘MR visible’, because its signals are not directlydetectable due to the very short T2.

Note that both pools have approximately thesame central resonance frequency (Fig. 1) andthat the only difference is the larger spread inresonance frequencies of the restricted protons(correlation times ≈ 10-8 s).

Under normal MR conditions magnetizationis exchanged from the ‘free pool’ to the ‘boundpool’ and vice versa, resulting in an equilibriumsituation characteristic for that type of tissue.One should realize that if the magnetization ineither pool is perturbed (e.g. by the RF excita-tion pulses in standard MRI), changes occur inboth pools. The normally observed spin-latticerelaxation of the ‘free pool’ is, in fact, a complexfunction of the recovery of both pools to theequilibrium state (relaxation coupling).

A special situation occurs when the ‘boundpool’ is selectively saturated. This will result inno net magnetization of the ‘bound pool’ andthus a difference in magnetization between thepools is created. Due to cross relaxation proces-ses, magnetization is transferred from the ‘freepool’ to the ‘bound pool’.

In order to obtain a first order impression ofthe magnetization transfer phenomenon, asimplified biological tissue with only oneimaging voxel is shown in Figure 2. This hypo-thetical tissue has a ‘free pool’ of four watermolecules with four proton spins, and a ‘boundpool’ of one macromolecule with three protonspins. In Figures 2-6 the arrow up denotes oneunit of measurable MR signal for the ‘free pool’,and the arrow down indicates zero measurableMR signal. The ‘free pool’ and ‘bound pool’ cantransfer magnetization to and from each otherand are in equilibrium. In the present constella-tion the signal to noise ratio (S/N) = 4 arbitrary units (AU). Note that the three spins of the‘bound pool’ do not contribute to the S/N ofthis tissue.

The second step of the MTC experiment isto selectively saturate the bound protons (Fig.3). Note that the four spins of the ‘free pool’ arenot (yet) affected. As an example for thisspecific tissue, the next step in the MT processis that two spins exchange magnetization. Theresult of this cross relaxation is that the S/N isdecreased to 2 AU (Fig. 3). Note that the twospins of the ‘bound pool’ are ‘MR invisible’ andtherefore do not contribute to the S/N of thetissue. If the ‘bound pool’ is held in saturation,a new equilibrium situation has been created

where the tissue has less available longitudinalmagnetization. Note that the apparent longitu-dinal relaxation time T1sat has been shortenedcompared to the normal T1: after excitation itnow takes less time to relax back to the new(smaller) equilibrium magnetization (Fig. 4).How this mechanism can actually be used toimprove contrast in MRI is shown in Figure 5.We now have two hypothetical tissues. Theoutside tissue has only four free water moleculeswith four spins, whereas the inner tissue has botha ‘free pool’ and a ‘bound pool’ with also a totalof four ‘MR visible’ spins. Both tissues have aS/N = 4 AU, and therefore there is no contrast.

However, after selective saturation of the‘bound pool’ the S/N of the inner tissue isreduced to 2 AU as a result of magnetization transfer. The other tissue is not affected (S/N =4 AU). Consequently, in the new equilibriumsituation, there is a large contrast between thetwo tissues: namely the Magnetization TransferContrast (MTC) (Fig. 5).

Fig. 1. The spectral linesof the ‘free pool’ and the‘bound pool’ are bothcentred at the sameLarmor frequency f0. The‘free pool’ has a narrowline (relatively long T2);the ‘bound pool’ has ahomogeneous broadenedline (T2 < 0.2 ms).

Fig. 2. Simplifiedbiological tissueconsisting of one imagingvoxel. This hypotheticaltissue has a ‘free pool’ offour water molecules withfour proton spins, and a‘bound pool’ of onemacromolecule with threeproton spins. An arrowup indicates one unit ofmeasurable MR signal(S/N) for the ‘free pool’.An arrow down means nomeasurable MR signal(S/N = 0).

'free pool'

'bound pool'

f0 Frequency

H H

O

macro- molecule

H

H

H H

O

H H

O

H H

O

H

1

2

66

Cross relaxation: dipole-dipole interactionsAlthough the magnetization transfer mecha-nism is relatively well understood at the tissue

level, much less detailed knowledge is availableon a cellular and molecular level. Continuedadvances in modelling of magnetization transfereffects will improve the detailed picture of themechanism underlying this phenomenon11, 21.The most comprehensive models, as summa-rized by Balaban and Ceckler7, describe the ‘freeproton pool’ and the macromolecule protonpool (‘bound pool’). The ‘free pool’ is thoughtto include both mobile bulk water molecules andwater molecules bound to the macromolecularsurface, the so-called ‘hydration water’ or‘surface water layer’. These two water fractionsexist in a state of rapid chemical equilibrium,such that all the tissue water can be thought ofas a single proton pool. According to Eng et al.5the surface layer of water communicates withthe bulk water by diffusion. The actual transferof magnetization (see also Figure 6) from thesurface layer of water with the macromoleculeprotons occurs by means of dipole-dipoleinteractions through space (cross relaxation).

Research on the actual binding sites isongoing. Eng et al.5 suggest that lipid bilayers

Fig. 3. During an MTCexperiment themeasurable MR signal isdecreased. The upper partshows the initial magneticenvironment of thehypothetical tissue in Fig.2, the middle part showsthe spins after selectivesaturation of the ‘boundpool’, and the bottomimage shows the newequilibrium situation witha lower signal of the ‘freepool’.

Fig. 4. The apparentlongitudinal relaxationtime T1sat has beenshortened compared withthe normal T1: afterexcitation it now takesless time to relax back tothe equilibriummagnetization MAsat.

MA0

-MA0

MAsat

-MAsat

Time

T1sat

normal T1

4

3

H H

O

macro- molecule

H

H

H H

O

H H

O

H H

O

H

Pre-saturation: S/N = 4

H H

O

H

macro- molecule

H H

O

H H

O

H H

O

H

H

Selective saturation of bound protons

H

H

O

H

macro- molecule

H H

O

H H

O

HH

O

H

H

Magnetization transfer post-saturation: S/N = 2

67

are effective macromolecular structures formagnetization transfer between ‘bound’ and‘free’ protons.

It has also been suggested that chemicalexchange between protons in hydration waterand macromolecule protons could be theprimary mechanism for magnetization transfer.However, experiments with ‘broad’ 2H spinsand water 2H in vivo, as well as experimentswith the spin 1⁄2 isotope tritium, did not revealmagnetization transfer effects. Furthermore, inmodel lipid systems and other macromolecules,similar magnetization transfer effects have been observed independently of pH values from5 to 8.

From these studies it is concluded thatchemical exchange is not a significant factor inthe magnetization transfer process5.

It is to be expected that ongoing studies willexplore the biochemical and biophysical path-ways underlying the magnetization transfermechanism, in order to optimize the specificityof MR tissue characterization with this uniquephysical phenomenon9.

5

6

Fig. 5. Magnetizationtransfer can increasecontrast between tissueswith different biophysicalproperties. The images onthe left represent the pre-MTC situation; those onthe right the post-MTCsituation. The upper set of drawings show theproton spins in thedifferent tissues. Thelower set of images aregrey scaled to representthe corresponding signallevel.

Fig. 6. Different physicaland chemical processesplay a role inmagnetization transfer:diffusion, chemicalexchange, and crossrelaxation. For furtherdetails, please refer to themain article.

H H

O

macro- molecule

H

H

H H

O

H H

O

H H

O

H

H H

O

H H

O

H H

O

H H

O

H H

O

H H

O

H H

O

H H

O

H

H

O

H

macro- molecule

H H

O

H H

O

HH

O

H

H

S/N=4

S/N=4

S/N=4

S/N=2

H

H

H

H

HH

H

H

H

O

O

O

O

macro- molecule

cross- relaxation

chemical exchange

H

H

O H

H

O

H

H

O

H

HO

H

H

H

O

diffusion

diffusion

H

68

Dynamics of the magnetization transfer mechanismThe dynamics of the magnetization transferprocess were already formulated in 1963 byForsén and Hoffman1. A two-pool spinexchange system, shown in Figure 7 by the‘coupled bathtubs’ diagram, has been analysedto form a quantitative basis of the previouslydescribed mechanisms22-27, 32. The longitudinalmagnetization of the A pool is labelled MA, andfor the B pool MB. Under fully relaxedconditions, this is MA0 and MB0.

The spin-lattice relaxation rates in theabsence of exchange (intrinsic relaxation) are labelled R1A (R1A = 1/T1A) and R1B (R1B =1/T1B).

The exchange rates of magnetization trans-fer between the A and B pools are labelled Kforand Krev.

Fig. 7. The ‘coupledbathtub’ model. The Apool is the ‘free pool’, theB pool is the ‘boundpool’. The longitudinalmagnetization of the Apool is labelled MA0, andfor the B pool MB0. Thespin-lattice relaxationrates in the absence ofexchange are labelled R1A(R1A = 1/T1A) and R1B(R1B = 1/T1B). Theexchange rates ofmagnetization transferbetween the A and B poolare labelled Kfor and Krev.The water level in thebathtubs represents theamount of magnetization.

Fig. 8. When the B pool isselectively saturated, anew equilibrium iscreated where the A poolhas less availablelongitudinalmagnetization (labelledMAsat). See also Figure 7.

Fig. 9. A larger exchangerate Kfor will result in alower value for MAsat.The labels Kfor =1 andKfor =2 indicate themagnitude of thelongitudinalmagnetization MAsat inthe two situations. Seealso Figures 7 and 8.

Fig. 10. A largerrelaxation time T1A willresult in a lower value forMAsat. The labels 0.5 Tand 1.5 T indicate themagnitude of thelongitudinalmagnetization MAsat atdifferent field strengths.Kfor is the same for bothsituations. See alsoFigures 7 and 8.

LATTICE

R1A=1/T1A R1B=1/T1B

Kfor

KrevA B

MA0 MB0

MAsat

LATTICE

R1A=1/T1A

Kfor

A B

Kfor =1

Kfor =2

R1A=1/T1A

LATTICE

Kfor

A B

0.5 T

1.5 T

R1A=1/T1A

LATTICE

Kfor

A B

7

8

9

10

69

The water level in the bathtubs (Figs 7-10)represents the amount of magnetization.

At equilibrium the transfer of magnetizationfrom A to B (MA . Kfor) is equal to the transferof magnetization from B to A (MB . Krev).

In more general terms, the rate of change ofmagnetization in the A pool, dMA(t)/dt, is givenby:

dMA(t)/dt = (MA0 - MA(t)) . R1A –MA(t) . Kfor + MB(t) . Krev (1)

When the B pool is selectively saturated (andkept in complete saturation!) a new equilibriumis created where the A pool has less availablelongitudinal magnetization (labelled MAsat) andalso a shorter longitudinal relaxation time con-stant (labelled T1sat), with:

MAsat = MA0 / (1 + Kfor . T1A) (2)

and:

1/T1sat = R1A + Kfor (3)

Re-arranging (2) and (3) results in

Kfor = (1/T1sat) . (1 - MAsat/MA0) (4)

or:

MAsat/MA0 = 1 - Kfor . T1sat (5)

Note that magnetization transfer is a dynamicprocess. After selective saturation of the B pool,the equilibrium longitudinal magnetization MA0is reduced exponentially to MAsat with a timeconstant T1sat. Moreover, the analytical solu-tions above assume complete and indefinitesaturation of the B pool. More details are welldescribed by Hajnal et al.32. More complex anddetailed MTC models are still being develo-ped15, 19, 20.

A more intuitive understanding of these effects is shown in Figure 8. The B pool is nowempty (complete saturation has taken placesince there is no water in bathtub B) and waterflows from A to B. A new equilibrium waterlevel is established when the inflow (via tubeR1A) and outflow (via tube Kfor) are in balance.The amount of water which flowed from A toB cannot flow backwards, because the B pool iscontinuously emptied (kept in saturation). It isalso easy to understand that a larger exchangerate Kfor will result in an even lower value forMAsat (mathematically described in Formula 2).

Figure 9 shows that a larger outflow tube(larger Kfor) will result in a lower equilibriumwater level. This effect is comparable to the

situation where the exchange rate Kfor isunchanged but the T1A is larger.

The smaller inflow tube (R1A) will result ina lower equilibrium water level (Fig. 10). Thiseffect is also known from clinical practice - ahigher field strength will result in a largermagnetization transfer effect9.

Ceckler et al.7 showed experimentally thatthe exchange rate Kfor is field strength inde-pendent. The lower value of MAsat is the resultof the higher T1 relaxation values with highermain magnetic fields (Fig. 10).

Saturating pre-pulsesThere are basically two methods to selectivelydestroy the ‘bound pool’31-41.

The first method employs the differences inwidths of the resonance lines (Fig. 11). Fre-quency selective RF pulses are transmitted at afrequency offset ∆f with respect to the central1H Larmor frequency. Using these off-reso-nance pulses it is possible to substantiallysaturate the broad resonance of the ‘bound pool’while hardly affecting the ‘free pool’. Note thatsuch a pulse will not only destroy the spins withthe specific off-resonance frequency: due to spindiffusion all the macromolecule protons will beaffected. The lower water peak is the result ofthe magnetization transfer process.

The second method employs the differencesin T2 between the pools. Binomial RF pulseswith a resultant zero degree pulse angle (‘jumpand return’) are transmitted on-resonance. Themost often used pulse is a 90ºx 90º-x 90º-x 90ºxpulse, with a pulse length < 0.5 ms per element.During such a pulse-train, the magnitude of the‘free proton’ magnetization is not changed, butits direction is changed according to the flipangle of each RF element.

Due to the long T2 (much longer than the du-ration of the RF elements) the magnetization isreturned to the z-axis at the end of the binomialpulse. Magnetization of ‘bound protons’ with a

11

off-resonance pulse

f0 Frequency∆f

'free pool' after MT

'free pool' before MT

'bound pool'

Fig. 11. Using off-resonance pulses it ispossible to substantiallysaturate the broadresonance of the ‘boundpool’ while hardlyaffecting the ‘free pool’.Due to spin diffusion allthe macromoleculeprotons will be saturated.The lower water peak isthe result of themagnetization transferprocess.

70

very short T2 is destroyed, because the trans-verse dephasing is much faster than the durationof the RF elements. The net effect of such apulse-train is that the magnetization of the ‘freeprotons’ is not changed and the ‘bound protons’

are destroyed (Figs 12, 13). Both methodsshould be considered as MTC pre-pulses whichcan be interleaved with all known MRI pulsesequences (Fig. 14). Depending on the clinicalapplication, one can either use the inducedchanges in longitudinal magnetization (lowerMAsat) or the reduced longitudinal relaxationtime constant (lower T1sat) or a combination ofboth.

As a starting point for experimentation it isprobably easiest to select a proton densityweighted sequence and add either an on-resonance or an off-resonance MTC pre-pulseto generate the desired signal changes.

Because magnetization transfer is a dynamicprocess the MTC pre-pulses have to be repeatedto keep the ‘bound pool’ saturated. As a rule ofthumb, 10-20 MTC pre-pulses per secondshould be sufficient for complete saturation37.Depending on the software implementation onclinical scanners, the user can influence this bymodifying repetition time TR and number ofslices. On the Philips Gyroscan MR systems theMTC pre-pulses (either on-resonance or off-resonance), followed by a spoiler gradient todephase any residual transversal magnetization,are applied before each excitation RF pulse(with Turbo Field Echo before each shot).

On-resonance and off-resonance pulsesThe off-resonance pulse is a single, shaped RFpulse (e.g. sinc-Gauss) with a relatively narrowexcitation bandwidth (dependent on pulseshape and duration). Two critical parameterswill affect the amount of saturation: the

Fig. 13. The result of anon-resonance binomialpulse (90ºx 90º-x 90º-x90ºx) for the ‘free pool’ isthat the longitudinalmagnetization Mz returnsto its original value. Thespins of the ‘bound pool’with very short T2 onlyexperience decay,resulting in destroyedmagnetization after theon-resonance pulse.

Fig. 12. Simplifiedrepresentation of thechange of longitudinalmagnetization Mz duringa 90ºx 90º-x pulse. Upperrow represents the ‘freepool’, lower rowrepresents the ‘boundpool’. First column is theinitial situation. Thesecond column shows Mzand Mxy after a 90ºxexcitation pulse. The thirdcolumn shows thesituation after some time∆t, with ∆t << T2 of the‘free pool’ and ∆t >> T2 of

the ‘bound pool’. Thefourth column shows Mzand Mxy after a 90º-x resetpulse.

1.0

0.5

0.0

Mz

'free pool'

'bound pool'

121 pulse

0.0 0.5 1.0 1.5 2.0Time (ms)

12

13

Z'

X'

Y'

Mz = 1

Mxy = 0

Z'

X'

Y'

Mz = 0

Mxy = 1

Z'

X'

Y'

Mz = 0

Mxy = 1

Z'

X'

Y'

Mz = 1

Mxy = 0

Z'

X'

Y'

Mz = 1

Mxy = 0

Z'

X'

Y'

Mz = 0

Mxy = 1

Z'

X'

Y'

Mz = 0

Mxy = 0

Z'

X'

Y'

Mz = 0

Mxy = 0

71

magnitude and the frequency offset of the off-resonance pulse31-34, 38, 41.

More B1 power will result in a bettersaturation, but is limited in practice by safetyconsiderations (Specific Absorption Rate orSAR).

Decreasing the frequency offset will also create a better saturation of the ‘bound pool’,but values closer to the 1H Larmor frequencywill also result in direct excitation of the ‘freepool’. A practical value is a 700º pulse angle of10 ms applied at a frequency offset of 1500 Hz.For more details the user should refer to theoperator’s manual of the scanner used.

The on-resonance pulse is a composite RFpulse consisting of a series of rectangular (block)pulses. These binomial pulses can be consider-ed as transparent for the ‘free pool’. Several authors35-37, 39-40 studied the optimal form (11, 121, 1331 etc). From the Bloch equationsone can show that higher order binomial pulses,with a shorter duration and a larger B1, willresult in a larger transparent bandwidth. This isthe range of frequencies for which the ‘freespins’ are not affected after the pulse. Due tosafety (SAR) and hardware limits a typical on-resonance pulse is a 121 pulse (4 elements) witha pulse angle of 90º and a duration per elementof 0.3 ms.

In clinical practice, a practical problem at 1.5T is the water-fat shift (WFS ≈ 220 Hz). Thetransparent bandwidth of the on-resonanceMTC pre-pulse should be larger than the WFS,otherwise the fat spins will also be affected. Notethat this has nothing to do with the actualmagnetization transfer (cross relaxation), but isthe direct result of the excitation bandwidth ofthe pulses used. Increasing the flip angle anddecreasing the time duration of the on-resonance pulse elements will be helpful toincrease the transparent bandwidth of the bino-mial pulse (Fig. 15). Again, the user should referto the operator’s manual of the scanner used.

The results of both methods for the magne-

tization transfer effects are basically similar,whereby the amount of saturation is a directfunction of the B1 power of the MTC pre-pulses. Theoretical simulations on the power efficiency41 are probably bypassed by otherfactors in practical scanning, like SAR safetylimits, desired contrast, and time efficiency.

Interleaving a standard MRI pulse sequencewith MTC pre-pulses will increase TR with thetime duration of the MTC pre-pulse and timefor spoiling. This might be a disadvantagedepending on the clinical application.

Incidental MTCDixon et al.28 described the fact that MTC isimplicitly generated in many normal imagingsequences where off-resonance pulses areapplied. For example, in Spin Echo (SE)sequences the slice selective excitation andrefocusing pulses can be considered as off-resonance pulse for neighbouring slices. Thiseffect will also occur in 3D, spectral fatsuppression and REST slabs (out-of-slicesuppression).

In multislice TSE (Turbo SE, also called FastSE) sequences this effect is significant due tothe many 180º refocusing pulses30. Melki et al.29

nicely demonstrated that the signal attenuation

Fig. 15. The transparentbandwidth is the range offrequencies for which the‘free spins’ are notaffected after the pulse.Increasing the flip angleand decreasing the timeduration of the on-resonance pulse increasesthe transparent bandwidthof the binomial pulse(squares). Mz is thelongitudinalmagnetization.

Fig. 14. MTC pre-pulseswhich can be interleavedwith all known MRI pulsesequences. On the PhilipsGyroscan systems theMTC pre-pulses (eitheron-resonance or off-resonance), followed by aspoiler gradient todephase any residualtransverse magnetization,are applied before eachexcitation RF pulse (α).

mtcspatial encoding

and measure

α

mtcspatial encoding

and measure

α

mtcspatial encoding

and measure

α

mtcspatial encoding

and measure

α

TR

Slice 1

Slice 2

Slice 3

14

15

-400 -200 0 200 400

Frequency (Hz)

Frequency profiles 121 pulse

Mz

1.0

0.5

0

-0.5

-1.0

72

for white and grey brain matter in multislicesequences is a direct function of the number of slices. An example in knee joint imaging of theincidental MTC effect is shown in Figure 16.

Visualization of MTCThere are many ways to visualize the effects ofthe magnetization transfer process. The easiestway is to judge the MTC images directly. A moretime-consuming way is to generate two sets ofimages, one with and one without MTC pre-pulses, and compare the changes in contrastqualitatively. A more quantitative approach isto construct synthetic images, for example a

subtraction image or a ratio image. The firstapproach means that for each pixel thedifference between two images is calculated andvisualized; in the latter approach the relativechange in signal intensity for each pixel iscalculated. These subtraction or ratio imagescan be colour coded and overlaid on a high-resolution normal grey scale image. The mostcomplex way is to calculate the actual rate ofexchange of magnetization on a pixel basis.Theoretically, this is the most specific way ofvisualizing the MTC. Experimentally, it is moredifficult because it requires measurement ofseveral physical properties (MA0, MAsat, T1sat,see Equation 4) and requires postprocessing ofthe data. The optimal solution of MTC visuali-zation in routine scanning is dependent on theclinical application, scan time considerationsand software for automatic generation ofcalculated images.

AcknowledgementsThe author gratefully acknowledges the contri-butions of M.K. Kouwenhoven MSc, L.H. Hofland MSc, S. Sheppard MSc, J.A. den BoerPhD, M.S. Silver PhD, A.R. Gillams MD, M.

Vahlensieck MD, D.J. Jensen PhD, H.J. Kooijman PhD and M. Beese MD, in under-standing the MR physics and applications ofMTC.

References

Original publications1. Forsén S, Hoffman RA. Study of Moderately RapidChemical Exchange Reactions by Means of Nuclear Mag-netic Double Resonance. J Chem Phys 1963; 39: 2892-2901.2. Edzes HT, Samulski ET. Cross Relaxation and SpinDiffusion in the Proton NMR of Hydrated Collagen. Na-ture 1977; 265: 521-523.3. Edzes HT, Samulski ET. The Measurement of Cross-

Relaxation Effects in the Proton NMR Spin-Lattice Re-laxation of Water in Biological Systems: Hydrated Colla-gen and Muscle. J Magn Reson 1978; 31: 207-229.4. Wolff SD, Balaban RS. Magnetization Transfer Con-trast (MTC) and Tissue Water Proton Relaxation In Vivo.Magn Reson Med 1989; 10: 135-144.5. Eng J, Ceckler TL, Balaban RS. Quantitative 1H Mag-netization Transfer Imaging In Vivo. Magn Reson Med1991; 17: 304-314.

Fig. 16. MR imagesshowing incidental MTCin knee joint imaging.Note the differences incartilage and muscleintensity between multi-slice SE, multi-slice TSEand single-slice TSE.

16

73

6. Wolff SD, Eng J, Balaban RS. Magnetization TransferContrast: Method for Improving Contrast in Gradient-Recalled-Echo Images. Radiology 1991; 179: 133-137.

Review papers7. Balaban RS, Ceckler TL. Magnetization Transfer Con-trast in Magnetic Resonance Imaging. Magn Reson Q1992; 8(2): 116-137.8. Bryant RG, Lester CC. Magnetic Relaxation Couplingin Heterogeneous Systems. J Magn Reson (series B) 1993;101: 121-125.9. Wolff SD, Balaban RS. Magnetization Transfer Imag-ing: Practical Aspects and Clinical Applications. Radio-logy 1994; 192: 593-599.10. Santyr GE, Mulkern RV. Magnetization Transfer inMR Imaging. Magn Reson Imaging 1995; 5: 121-124.

Relaxation mechanisms11. Morris GA, Freemont AF. Direct Observation of theMagnetization Exchange Dynamics Responsible for Mag-netization Transfer Contrast in Human Cartilage In Vivo.Magn Reson Med 1992; 28: 97-104.12. Brown III RD, Koenig SH. 1/T1ρand Low-Field 1/T1of Tissue Water Protons Arise from Magnetization Trans-fer to Macromolecular Solid-State Broadened Lines.Magn Reson Med 1992; 28: 145-152.13. Sepponen R. Rotating Frame and MagnetizationTransfer. Chapter 8 in: Stark DS, Bradley WG (eds). Mag-netic Resonance Imaging. 2nd ed. St. Louis, Mo: MosbyYear Book, 1992; 204-218.14. Koenig SH, Brown RD, Ugolini R. MagnetizationTransfer in Cross-Linked Bovine Serum Albumin Solu-tions at 200 MHz: a Model for Tissue. Magn Reson Med1993; 29(3): 311-317.15. Henkelman RM, Huang X, Xiang Q, Stanisz GJ,Swanson SD, Brosnkill MJ. Quantitative Interpretation ofMagnetization Transfer. Magn Reson Med 1993; 29: 759-766.16. Yeung HN. On the Treatment of the Transient Re-sponse of a Heterogeneous Spin System to Selective RFSaturation (letter to the editor). Magn Reson Med 1993;30: 201-206.17. Yang H, Schleich T. T1 Discrimination Contributionsto Proton Magnetization Transfer in Heterogeneous Bio-logical Systems. Magn Reson Med 1994; 32(1): 16-22.18. McGowan JC, Leigh JS jr. Selective Saturation inMagnetization Transfer Experiments. Magn Reson Med1994; 32(4): 517-522.19. Henkelman RM, Stanisz GJ, Kim JK, Bronskill MJ.Anisotropy of NMR Properties of Tissues. Magn ResonMed 1994; 32(5): 592-601.20. Morrison C, Henkelman RM. A Model for Magneti-zation Transfer in Tissues. Magn Reson Med 1995; 33:475-482.21. Harrison R, Bronskill MJ, Henkelman RM. Magne-tization Transfer and T2 Relaxation Components in Tis-sue. Magn Reson Med 1995; 33: 490-496.

Dynamics of magnetization transfer22. Rydzy M, Deslauriers R, Smith ICP, Saunders JK.Optimization of Magnetization Transfer Measurements:Statistical Analysis by Stochastic Simulation. Applicationto Creatine Kinase Kinetics. Magn Reson Med 1990; 15:260-274.23. Komu M. Analysis of Longitudinal Relaxation RateConstants from Magnetization Transfer MR Images ofHuman Tissues at 0.1 T. Magn Reson Imaging 1992; 10:35-40.24. Mora BM, Narasimhan PT, Ross BD. 31P Magne-tization Transfer Studies in the Monkey Brain. MagnReson Med 1992; 26: 100-115.

25. Boer RW de, Elevelt AJ. Turbo-MIX T1 Measure-ments and MTC Exchange Rate Kfor Calculations. Pro-ceedings of The SMRM, 12th annual Meeting New York,1993; 1: 175.26. Schick F, Forster J, Pfeffer M, Lutz. O. Pulsed Mag-netization Transfer for Imaging and Spectroscopic Ap-plications on Whole-Body Imagers. MAGMA 1994; 2:127-137.27. Taitelbaum H, Weiss GH, Spencer RGS. Optimiza-tion of Magnetization Transfer Experiments for KineticRate Measurements. NMR in Biomedicine 1994; 7: 287-292.

Incidental MTC28. Dixon TW, Engels H, Castillo M, Shardashti M. In-cidental Magnetization Transfer Contrast in StandardMultislice Imaging. Magn Reson Imaging 1990; 8: 417-422.29. Melki PS, Mulkern RV. Magnetization Transfer Ef-fects in Multislice RARE Sequences. Magn Reson Med1992; 24: 189-195.30. Santyr GE. Magnetization Transfer Effects in Mul-tislice MR Imaging. Magn Reson Imaging 1993; 11(4):521-532.

Pulse sequences31. Ordidge RJ, Knight RA, Helpern JA. MagnetizationTransfer Contrast (MTC) in Flash MR Imaging . MagnReson Imaging 1991; 9: 889-893.32. Hajnal JV, Baudouin CJ, Oatridge A, Young IR, Bydder GM. Design and Implementation of Magne-tization Transfer Pulse Sequences for Clinical Use. J Comput Assist Tomogr 1992; 16(1): 7-18.33. Dixon WT. Use of Magnetization Transfer Pulse Se-quences for Clinical Use. J Comput Assist Tomogr 1992;16: 7-18.34. Swallow CE, Kahn CE jr, Halbach RE, Tanttu JI, Sepponen RE. Magnetization Transfer Contrast Imagingof the Human Leg at 0.1 T: a Preliminary Study. MagnReson Imaging 1992; 10(3): 361-364.35. Yeung HN, Aisen AM. Magnetization Transfer Con-trast with Periodic Pulsed Saturation. Radiology 1992,183: 209-214.36. Hu BS, Conolly SM, Wright GA, Nishimura DG, Macovski A. Pulsed Saturation Transfer Contrast. MagnReson Med 1992; 26: 231-240.37. Schneider E, Prost RW, Glover GH. Pulsed Magne-tization Transfer versus Continuous Wave Irradiation forTissue Contrast Enhancement. Magn Reson Imaging1993; 3: 417-423.38. Jones RA, Southon TE. A Magnetization TransferPreparation Scheme for Snapshot FLASH Imaging. MagnReson Med 1993; 19: 483-488.39. Pike GB, Glover GH, Hu BS, Enzmann DR. PulsedMagnetization Transfer Spin-echo MR Imaging. MagnReson Imaging 1993; 3: 531-539.40. McGowan JC, Schnall MD, Leigh JS. MagnetizationTransfer Imaging with Pulsed Off-Resonance Saturation:Variation in Contrast with Saturation Duty Cycle. MagnReson Imaging 1994; 4: 79-82.41. Hua J, Hurst GC. Analysis of On- and Off-ResonanceMagnetization Transfer Techniques. Magn Reson Imaging 1995; 5: 113-120.