quantitative nuclearmagnetic resonance imaging ... · when imaging the cold lesions, the...

Journal of Neurology, Neurosurgery, and Psychiatry 1987;50:125-133

Quantitative nuclear magnetic resonance imaging:characterisation of experimental cerebral oedemaD BARNES, W I McDONALD, G JOHNSON, P S TOFTS, D N LANDON

From the Department of Clinical Neurology, Institute ofNeurology, Queen Square, London UK

SUMMARY Magnetic resonance imaging (MRI) has been used quantitatively to define the character-istics of two different models of experimental cerebral oedema in cats: vasogenic oedema producedby cortical freezing and cytotoxic oedema induced by triethyl tin. The MRI results have beencorrelated with the ultrastructural changes. The images accurately delineated the anatomical extentof the oedema in the two lesions, but did not otherwise discriminate between them. The patterns ofmeasured increase in T1' and T2' were, however, characteristic for each type of oedema, andreflected the protein content. The magnetisation decay characteristics of both normal andoedematous white matter were monoexponential for T1 but biexponential for T2 decay. The relativesizes of the two component exponentials of the latter corresponded with the physical sizes of themajor tissue water compartments. Quantitative MRI data can provide reliable information aboutthe physico-chemical environment of tissue water in normal and oedematous cerebral tissue, and are

useful for distinguishing between acute and chronic lesions in multiple sclerosis.

Magnetic resonance imaging (MRI) is of provenvalue for the detection of abnormalities in the brain ina variety of diseases of the central nervous system.Despite its sensitivity to pathological changes, how-ever, the abnormal images do not provide specificinformation about the nature of the underlyingpathology.'

In addition to producing images which demon-strate qualitative changes, MRI can also providequantitative data about a tissue. The relaxation times,T, and T2 and magnetisation decay characteristicsare of particular interest as they reflect not only theamount of water in a tissue but also its physico-chemical environment.2 These facts raise the possi-bility that specific pathological changes in a tissuemight be characterised in terms of the resultantchanges in its nuclear magnetic resonance (NMR)properties. We have investigated this possibility bystudying experimental models of cerebral oedema invivo using quantitative MRI techniques; the NMRcharacteristics have been compared with the mor-phological features of the lesions. Two models ofcerebral oedema have been studied in cats. First, a

Address for reprint requests: Prof WI McDonald, Institute ofNeurology, Queen St, London WCIN 3BG, UK.

Received 24 April 1986.Accepted 25 May 1986

cortical freezing lesion has been used to produce vaso-genic cerebral oedema in the underlying white matter.The lesion is characterised by expansion of the extra-cellular space which contains protein-rich fluid as aresult of damage to the blood vessels.3The second model is cytotoxic oedema induced by

the systemic administration of triethyl tin (TET) sul-phate. This type of oedema is characterised by split-ting of the myelin sheaths at the intraperiod line fol-lowed by progressive accumulation of oedema fluidbetween the separated lamellae to form intramyelinicvacuoles.4 Unlike vasogenic oedema, the oedemafluid in this form of cytotoxic oedema lacks asignificant protein content.5

In a previous paper' we described in detail the cor-relation between the NMR and morphological fea-tures of cytotoxic oedema. In this study we have com-pared the findings in vasogenic and cytotoxic cerebraloedema in order to determine whether it is possible togain an insight into the nature of pathologicalchanges in tissue water using quantitative MRI.

Methods

Twenty-six adult cats (weighing 2-3 5 kg) were used in theseexperiments. Vasogenic oedema was induced by means of acold lesion in 16 animals. The cold lesion was produced bythe following method: general anaesthesia was induced usingpentobarbitone sodium (40 mg/kg body weight) by intra-

125

Protected by copyright.

on October 28, 2020 by guest.

http://jnnp.bmj.com

/J N

eurol Neurosurg P

sychiatry: first published as 10.1136/jnnp.50.2.125 on 1 February 1987. D

ownloaded from

http://jnnp.bmj.com/

126

peritoneal injection, and the animal was positioned in astereotactic head frame. A burr hole was then made18-20mm in front of the external auditory meatus and8-10mm to the right of the midline and the exposed duraprotected by a piece of thin plastic film. A circular brasschuck, 6mm in diameter, was cooled in liquid nitrogen andapplied to the dural surface for 25 seconds. After removal ofthe chuck, the frozen cortex was allowed to thaw, the plasticfilm removed and the soft tissues sutured in layers.

Cytotoxic oedema was induced in 10 animals by the intra-peritoneal injection of one to four doses of TET sulphate(1 mg/kg body weight).

ImagingNMR imaging was performed on a Picker International sys-tem operating at 0 5 Tesla. A home-made saddle-shapedreceiver coil 10cm in both diameter and length was used,and the imaging sequences were modified to produce a 15 cmfield of view. The reconstruction method used was2-dimensional Fourier transformation. The pixel dimensionswere 0-6 x 1 2mm and the slice thickness, 5 mm.Of the 16 animals in which cold lesions were made, all

were scanned at 24 hours, and six were re-examined at 72hours after the operation. Those with cytotoxic oedema wereexamined at intervals up to 4 days after the first dose ofTETsulphate. Each imaging session lasted approximately 3 hoursduring which a series of spin-echo (SE) (T. from 40 to1120 ms) and inversion-recovery (IR (Ti from 100 to 700 ms)images' were acquired in the coronal plane. This plane wasfound to be the most useful since it provided the best anat-omical detail, and the depth of abnormal tissue relative tothe slice thickness of 5 mm minimised partial volume effects.When imaging the cold lesions, the antero-posterior

extent of the underlying oedema was initially ascertainedusing sagittal views to enable the series of coronal images tobe centred on the oedematous white matter. The cytotoxicoedema, which involved the cerebral white matter diffusely,was imaged at the same level to minimise the effects ofregional variations in the NMR characteristics of whitematter when comparing the two lesions.The signal intensity of all imaging sequences depends to

varying extents upon protein density, T1 and T2. Thus SEimages are particularly sensitive to differences in T2 betweentissues, whereas IR images depend more on differences in T1for image contrast. The SE sequences were modified by theaddition of an extra 900 pulse after the echo whichcompletely saturated the magnetisation. This resulted in anequal T,-dependence of all SE regardless of T.6.The relaxation times were read directly from calculated

images produced from a two-point method by computeralgorithms: the T1 from an SE (40/2000) and an IR(700/2000/40), and the T2 from two SE (40/2000 and120/2000) images. The values for T1 and T2 obtained in thisway must be regarded as approximations to the true relax-ation behaviour of the tissue for reasons which are dealt withlater, and will therefore be referred to as T1' and T2'. Inaddition to the relaxation times, the magnetisation decaycharacteristics of the abnormal tissue for both T1 and T2relaxation were determined by measuring the signal intensityof the region of interest from the screen and fitting the datato the best mono-, bi- and triexponential decay functions

Barnes, McDonald, Johnson, Tofts, Landon

using a computed least-squares fitting procedure. An F-testwas then applied to determine the most appropriatefunction.

Determination of water contentThree animals with vasogenic and three with cytotoxicoedema were killed after the final imaging session by anoverdose of anaesthetic following which the brains wereremoved and sectioned. Samples of tissue (005-01 mls)were taken from the base of the oedematous gyrus for wet-dry weight determination of water content. They were driedin an oven at 60°C until their weights had been stable for48 hours.

Electron microscopyTwo animals with cytotoxic and six with vasogenic oedemawere killed by perfusion-fixation through the ascendingaorta via the left ventricle with 3% glutaraldehyde in 0 1 Mcacodylate buffer at a pH of 7-4. 1500mls of fixative wereperfused over a period of 15 minutes at an effective pressureof 180mm Hg. The brains were removed after a further 48hours in fixative and tissue samples taken from the regionsindicated in fig 1. They were post-fixed in 1% osmiumtetroxide, dehydrated in graded ethanol solutions andembedded in epoxy resin for microscopy. Semi-thin sections(0 5 microns) were stained with toluidine blue for lightmicroscopy and ultra-thin sections (60 nm) with methanolicuranyl acetate and Reynold's lead citrate for transmissionelectron microscopy.

._.~~~~~~~~~~~~~~~~~. ... ' ;}

JU

* '.4s

dy. x

*) -

Fi 1r Sceai cooa secio of th ca bri -hwn

th sit ofaplcto of th col prob an prxmtie

fro whic tisusapewrtknfo ekctro micrscop(U)~ ~ ~ ~



http://jnnp.bmj.com

/J N

eurol Neurosurg P


ownloaded from


Quantitative nuclear magnetic resonance imaging: characterisation of experimental cerebral oedema

Fig 2 Light micrograph ofthe suprasylvian gyrus showing the localised superficial cortical lesion and widening and pallor ofthe subjacent white matter (Luxol Fast Blue-Cresyl Fast Violet; x 13).

Light microscopyOne animal with cytotoxic, and one with vasogenic oedemawere perfused with formol-saline to provide paraffin sectionsfor light microscopy.

Results

Although every attempt was made to standardise themethod of production of the cold lesions throughoutthe series of experiments, the nature and severity ofthe pathological changes seen at different times werevariable. Similarly, animals showed variable sus-ceptibility to TET sulphate but the dosage regimeswere adjusted to produce comparable abnormalitieson the NMR images.

Microscopic appearancesVasogenic oedema. A superficial necrotic corticallesion marked the site of application of the coldprobe. The subjacent suprasylvian gyrus was widenedand showed marked pallor when compared with thenormal side (fig 2). The changes were maximal imme-diately subjacent to the cortical lesion, graduallybecoming less marked in the deeper white matter. The

lateral gyrus and internal capsule were involved to alesser extent, as was the ipsilateral side of the corpuscallosum, although the oedema did not appear to

/~~~~~~~~

Fig 3 Light micrograph ofwhite matter affected bycytotoxic oedema showing the spongy appearance impartedby the intramyelinic vacuoles (Toluidine Blue;Bar = 100 pIn).

127



http://jnnp.bmj.com

/J N

eurol Neurosurg P


ownloaded from


Barnes, McDonald, Johnson, Tofts, Landon

,}<< -n.

> 'if-., .. ... u/.}XB p

x ,gi

W'ad J R r | =H.. ..Fig 4 Electron micrographs of vasogenic oedema at (a) I day and (b) 3 daysfollowing the cold lesion. The extracellularspace is enlarged and contains extravasatedplasma protein (Bars = I pm).

cross the midline.At higher magnification, the myelinated fibres were

seen to be widely separated from each other in con-trast to the normal compact configuration. Theappearances were identical at 24 and 72 hours.

Cytotoxic oedema The cerebral white mattershowed diffuse pallor compared with normal; thatwithin the gyri, the corpus callosum and the internalcapsule adjacent to the basal ganglia being mostabnormal in appearance. At higher magnification,the predominant abnormality was a spongiformappearance imparted to the white matter by the pres-

ence of numerous vacuoles (fig 3). Individual vacuolesappeared to follow the direction of the myelinated

fibres, those running longitudinally appearing cigar-shaped, and those seen in transverse section, circular.

Electron microscopyVasogenic oedema Three lesions were examined at Iday and three at 3 days following the production ofthe cold lesion. In each lesion, the oedematous pro-cess was characterised by expansion of the extra-cellular space in the involved white matter resulting inabnormal separation of the nerve fibres and glial ele-ments (fig4). The changes were most marked in thesuprasylvian gyrus subjacent to the site of applicationof the cold probe, and appeared progressively lesssevere in samples from the lateral gyrus and internalcapsule, more remote from. it. The size of the extra-

128



http://jnnp.bmj.com

/J N

eurol Neurosurg P


ownloaded from


Quantitative nuclear magnetic resonance imaging: characterisation ofexperimental cerebral oedema......................................................................... -; } -- --- ......... _S_I_ 1|1 __|_ -I | gU- ::; & ^WsPwi> eN, ___ w_| I I | | * _ 1E _l I I | _l 1|_ _ ; _| | __ , * * - _ , I * _ r. . _. w -| I _ l * * | _ | 3_ l l l * _e _ _.| l l 13 ! | 1 l | | _* |-l | -1U-R Xi | l 3 _ _I_ -.E8BB .. _ ... *_ _a, _|e ew r S-eS!1 T.tZ P- - t.t.-

;...'M.f^'^'...-..;..

Fig 5 MRI (coronal section) ofa control animal showingthe normal distribution ofgrey and white matter(Inversion-recovery; T; = 200 ms, Tr = 2000 ms) (seefig 1).

cellular space in the oedematous white matter wasgenerally greater at day 1 than at day 3. The oedemadid not affect the contralateral hemisphere.The expanded extracellular space contained fluffy

electron-dense material with an appearance typical ofplasma protein. This material was of variable densityin different regions of individual lesions, but wasalways much more dense at I day than at 3 days afterinduction of the lesion (fig4a and b). The astrocyticprocesses in the oedematous region appeared swollenat both time intervals, but more so on day 3 than onday 1. The oligodendrocytes and myelinated nervefibres appeared normal.Cytotoxic oedema As previously described,6 thepredominant abnormality was severe disruption ofthe myelin sheaths. The myelin lamellae were sepa-rated at the intraperiod line and the spaces betweenthem were expanded to form vacuoles which oftenexceeded 15 pm in diameter. The contents of thevacuoles were devoid of electron-dense material incontrast to the oedema spaces in vasogenic oedema.The extracellular space was not visibly enlarged, andthe astrocytic processes appeared normal.

MRI appearancesA coronal image taken from a healthy control isshown in fig5 in which the normal pattern of whitematter can be seen.Vasogenic oedema The quantitative NMR datafrom the cold lesions was taken from the base of thesuprasylvian gyrus in each experiment. The appear-ances of vasogenic oedema are shown in fig 6.

Although abnormal signal enhancement was Visiblein the cortex at the site of application of the coldprobe, the remainder of the grey matter appeared

Fig 6 MRI taken 24 hours after the cold lesion. The rightsuprasylvian gyrus is widened and shows marked signalenhancement which extends into the the internal capsulebut does not cross the midline (Spin-echo; T. = 80ms,Tr = 2300 ms).

unaffected. The most abnormal signal was seen in theunderlying white matter, its anatomical distributionconforming to previous descriptions.8 It was possibleto identify the region of the cortical lesion itself and todelineate accurately the extent of the oedematouswhite matter as determined histologically. Whereas inthe IR images it was not possible to distinguish be-tween the superficial necrotic lesion and underlyingvasogenic oedema, the SE images not onlydifferentiated between these, but also provided moreaccurate information about the extent and severity ofthe oedema.Cytotoxic oedema The image appearance of this le-

Fig 7 MRI demonstrating diffuse signal enhancementfromthe cerebral white matter due to severe cytotoxic oedema(Spin-echo; T. = 240ms, Tr = 2000ms).

129



http://jnnp.bmj.com

/J N

eurol Neurosurg P


ownloaded from


Table 1 Absolute valuesfor T1' and T2'for normal and oedematous white matter and the percentage increase and ratiobetween themfor each experimental group

Experimental group T1 (ms) T2 (ms) % T1 % T2 % T1/T2

1 day cold (n = 16) 887 + 128 172 + 37 87 + 21 103 + 31 0-87 + 1-4(715-1160) (118-265) (50-150) (48-200) (0-72-1-15)

3daycold(n = 6) 770 ± 54 159 + 16 58 + 13 87 + 17 0-67 + 004(699-860) (142-184) (46-81) (72-123) (0-64-073)

TET(n = 10) 575 + 46 115 + 12 27 + 7 52 + 12 046 ± 007(510f640) (97-130) (11-33) (2*463) (0-36-059)

Control (n = 36) 476 + 10 81 + 4

sion has been described in detail previously.6 The sig-nal from the cerebral white matter was enhanceddiffusely (fig 7), but in keeping with the histologicalchanges, the degree of enhancement was greatest inthe gyri and the white matter adjacent to the basalganglia.

It was not possible to distinguish between regionsof cytotoxic and vasogenic oedema of comparableseverity on the basis of their image appearances alone.

Relaxation timesTable 1 summarises the relaxation data from both thenormal and experimental groups of animals.We have considered the relaxation times obtained

from experimental animals in terms of percentage in-creases above their pre-morbid values in order to fa-cilitate comparison between individual animals withdifferent normal values. Within each group the rela-tive proportional increases in the relaxation times re-main constant regardless of oedema severity. In 1 dayvasogenic oedema the relaxation times were increasedin approximately the same proportion for all degreesof severity, whereas in 3 day vasogenic oedema, T2'was increased by approximately 50% more than T1'.Cytotoxic oedema, at all stages of development1601

420^

80-

40-

0

a

as

00 * A

A&.m

* a

.*

4T2

showed an increase in T1' which was at least twicethat in T1'. The difference in the ratio of percentageincreases in T1' to T2' between 1 and 3 day vasogenic(p < 0 001), and between 3 day vasogenic and cyto-toxic oedema (p < 0 01) were significant using theMann-Whitney test for non-parametric, unpaireddata. The relaxation data for individual animals aregiven in fig 8.

Magnetisation decay characteristicsThe magnetisation decay characteristics of a tissuemight be expected to provide information about sepa-

105 1

-Q104.

._1-4-1

C

.G1 3.Vn 10-

102

10 260

Fig 8 Graph showing % increase in T1' versus T2'for eachanimal (- = cytotoxic; * = vasogenic at I day; andA = vasogenic at 3 days).

A0

* o

@0*0 a

0

0

0

0

0 0

0

I I

80 240 400 560Echo time / ms

720

Fig 9 Graph showing the progressive changes in themagnetisation decay characteristics ofwhite matter duringthe development ofcytotoxic oedema (O = control, 0 = 2days and A = 4 days after thefirst dose ofTET sulphate).

Barnes, McDonald, Johnson, Tofts, Landon130



http://jnnp.bmj.com

/J N

eurol Neurosurg P


ownloaded from


Quantitative nuclear magnetic resonance imaging: characterisation of experimental cerebral oedemarate microscopic tissue water compartments. In thepresent experiments the T1 magnetisation decay wasalways found to be monoexponential, whereas in 5/6normal animals and all those with oedematous whitematter produced non-exponential T2 magnetisationdecay curves. An example is shown in fig9.When the curve-fitting procedure was applied, the

addition of a third exponential function did notsignificantly improve the fit in any experiment andtherefore all magnetisation decay curves were treatedas biexponential, comprising a short-T2 and a long-T2 component. The relaxation times derived from thesignificantly biexponential decay functions for bothcomponents and the percentage of the observed signalderived from the long-T2 component are shown intable 2.

There were no significant differences between anyof the groups in the relaxation time of their short-T2components, whereas that of the long-T2 componentwas considerably greater in each experimental groupthan in healthy controls. The relative amplitude ofthis component, which comprised about 21% of thesignal in the control group, was increased to 55-65%in vasogenic oedema.

Water contentThe water content of normal and oedematous whitematter, as determined by estimation of wet and dryweights, are shown in table 3.The mean increment in water content of tissue

affected by cytotoxic oedema was 6-3% and bv vaso-genic oedema 12-3%.

Discussion

The morphological characteristics of vasogenic oe-dema reported in these experiments conform to pre-vious descriptions.9 10 The time course and severity ofthe oedema, however, showed marked variability de-spite the standardised method of production of thecold lesions. The degree of expansion of the extra-cellular space was generally greater at 1 day than at 3days, although a more reliable distinguishing featurewas the much greater density of extravasated plasmaprotein in the extracellular space at I day after thecold lesion. This finding is in keeping with the lowerTable 2 T2 relaxation times ofeach component, and theproportion ofthe total signal contributed by the long-T2component in normal and oedematous white matter

Experimental Short-T2 Long-T2 %group (ms) (ms) Long-T2

Control (n = 4) 67 + 4 190 + 32 21 + 7I day cold (n = 7) 74 + 19 337 + 60 65 ± 103 day cold (n = 3) 56 + 8 285 + 109 55 ± 7TET sulphate (n = 5) 70 + 13 346 + 66 24 + 4

protein content of oedematous white matter at 3days,11 12 which is the result of uptake and metabo-lism of the protein by astrocytes and vascular endo-thelial cells."3The animals with cytotoxic oedema showed vari-

able susceptibility to TET sulphate, and somemodification of the dosage regime was required toproduce comparable changes in the MRI appearancesof the oedematous white matter. The experimentswere terminated at 4 days since the pathologicalchanges were well developed at this time and the clin-ical effects were not yet severe. Cytotoxic oedema wasless severe than vasogenic in terms of the size of theultrastructural oedema space and smaller increase inwater content of the oedematous tissue. Furthermore,the protein content of the oedema fluid was differentin that cytotoxic oedema fluid was unassociated withplasma protein in keeping with previous reports thatthe fluid resembles a plasma ultrafiltrate.5MRI was sensitive in demonstrating both types of

oedema, but on the basis of image appearance alone itwas not possible to distinguish reliably between re-gions affected by cytotoxic and vasogenic oedema ofsimilar severity. The anatomical extent of the imageabnormalities corresponded with those seen on histo-logical examination.

In vasogenic oedema, both SE and IR sequencesclearly distinguished normal from abnormal tissue,but it was possible to differentiate between the area ofthe necrotic cortical lesion itself and the underlyingoedema more easily on the SE images due to thegreater sensitivity of these sequences to changes in T2.It was also easier to appreciate regional variations inseverity of the oedema on the SE images.The T1-dependant IR sequences were less useful in

TET sulphate-induced oedema, first because increasesin T1' were relatively small, and secondly becausesuch increases only served to reduce the normal con-trast between grey and white matter. Since the pro-portional increase in T2' was twice as great as in T1'in this type of oedema, the more T2-dependent SEsequences were most appropriate for demonstratingthe extent of the pathological changes which could beseen most clearly when echo times greater than 120 mswere used. The correct choice of imaging sequenceswould obviously be important when examining alesion with similar characteristics such as Reye's

Table 3 Water content ofnormal and oedematous whitematter

Experimental group Water content (%)

Control (n = 4) 65-9 + 2TET sulphate (n = 3) 72 2 + 2 9Cold (n = 3) 78.2 + 3-6

131



http://jnnp.bmj.com

/J N

eurol Neurosurg P


ownloaded from


132syndrome'4 and the encephalopathy associated withmitochondrial myopathy.'5

All experiments resulted in increases of both relax-ation times, consistent with the findings of previousstudies of these lesions by NMR spectroscopy.16 17The percentage increase in the relaxation times of thelesions ranged from 45-130%, but the changes ob-served in individual lesions corresponded with thesubjective assessment of the severity of the patholo-gical changes in those case examined by electron mi-croscopy. Go and Edzes16 have demonstrated thatthe severity of the oedema (as determined by the in-crease in tissue water content) is directly related to theresultant increase in T,' and T2', and it is likely thatthe range of increases in relaxation times found in thepresent experiments is a true reflection of the intrinsicvariability of the oedematous process, and not theresult of imprecise measurements.There were significant differences in the pattern of

increase in Tl' and T2' between the experimentalgroups. In early vasogenic oedema in which the fluidwas associated with large amounts of plasma protein,Tl' and T2' increased in approximately the same pro-portion, whereas after 3 days, when much of the pro-tein had been removed from the extracellular space,the proportional increase in T2' was approximately50% greater than that in Tl'. Cytotoxic oedema, inwhich the oedema fluid was unassociated with pro-tein, consistently showed an increase in T2' which wastwice that in T,'. The chemical nature of the oedemafluid thus appears to have a significant effect on thepattern of changes in T1' and T2'. As the plasma pro-tein is removed from the expanded extracellular spaceover a period of days following the cold lesion, so thepattern of change in the relaxation times approachesthat seen in cytotoxic oedema in which the oedemafluid in the vacuoles is unassociated with protein.The pattern of elevation of the relaxation times was

independent of the severity of the oedema in bothmodels; over a wide range of severity the pattern re-mained consistent within each experimental group.Two individual cold lesions which increased in sever-ity between I and 3 days still conformed to the ex-pected pattern of changes in the relaxation timesoutlined above.

These findings may be explicable in terms ofwhat isknown about the relaxation behaviour of proteinsolutions.'8 Protein molecules increase the relaxationefficiency of water protons, shortening both T, andT2 as a result of restriction of the free motion of thewater. When protein is associated with water at theconcentration found in plasma, T2 is much shorterthan T1, but at progressively lower concentrations T2increases more rapidly than T, until they becomeequal for pure water. The oedema fluid in I day vaso-genic, 3 day vasogenic, and cytotoxic oedema is asso-

Barnes, McDonald, Johnson, Tofts, Landonciated with protein at high, intermediate, and zeroconcentrations respectively, and therefore it might beexpected that 1 day vasogenic oedema would result inthe smallest increase in T2' relative to Ti', and cyto-toxic oedema the largest, as we have observed.

There are potential difficulties in the interpretationof relaxation data calculated from only two se-quences. As noted by Brandt-Zawadzki etal,'9 vari-ation of the sequence repetition time results invariations in the calculated relaxation times. In ourexperiments, the sequence parameters were kept con-stant, and the low variance of the relaxation datafrom the control animals (Tl' = 476 + 10; T2' = 81+ 4) suggests that our calculated values weresufficiently precise.As previously mentioned, there is a further problem

with relaxation time measurement using a two-pointmethod. When tissue magnetisation decays in a non-exponential fashion, calculation of T1' and T2' bysuch a method will inevitably produce data that arenot fully descriptive of the relaxation behaviour of thetissue. While recognising this limitation, these experi-ments have shown that they can provide an insightinto the nature of the fluid in cerebral oedema, andmight be usefully applied to the assessment of abnor-mal permeability of the blood-brain barrier and itstherapeutic modification in man.

Further information about the microscopic envi-ronment of tissue water can be obtained by examiningits magnetisation decay characteristics. The T, decayswere always monoexponential in these experiments.Previous workers have been unable to detect morethan one T, component in vasogenic oedema,'7 20although Go and Edzes'6 reported a biexponentialT, decay when studying cytotoxic oedema in rats byin vitro spectroscopy after large, single intravenousdoses of TET. It is uncertain to what extent such invitro findings can be extrapolated to in vivo tissuerelaxation behaviour in view of the inevitable alterna-tions in the physical properties of tissues which occurat death.We have found the T2 decays to be biexponential in

both normal and oedematous white matter, althoughit is probable that there are further, smaller com-ponents of tissue water which have very short T2relaxation times of the order of microseconds or a fewmilliseconds.2' There were no significant differencesin the relaxation times of the short-T2 componentsbetween any of the groups, and in keeping with therelatively unaltered ultrastructural appearance of theintracellular space in these lesions, it is reasonable toascribe this component to intracellular water which isunable to exchange with either the true extracellularspace or water within intramyelinic vacuoles rapidlyenough to yield as a single relaxation component.The relative amplitude of the long-T2 component



http://jnnp.bmj.com

/J N

eurol Neurosurg P


ownloaded from


Quantitative nuclear magnetic resonance imaging: characterisation of experimental cerebral oedema

of normal white matter, representing extracellularwater, was 21%, which is similar to the size of the invivo extracellular space in this tissue estimated byother methods.22 23 The relative amplitude of thiscomponent may therefore provide a means of deter-mining the degree of expansion of the extracellularspace in vasogenic oedema. In these experiments, themean relative amplitude in 1 day vasogenic oedemawas 65%, reaching 76% in the most severe individuallesion. These values were compatible with the size ofthe expanded extracellular space which we observedelectron microscopically.

In cytotoxic oedema the relaxation time of thelong-T2 component, representing water within thevacuoles (346 ms) was considerably greater than thatof water in the normal extracellular space (190ms),but we were unable to detect separate T2 componentsfor extracellular and vacuolar water. This may havebeen due to compression of the extracellular spaceby the expanding intramyelinic vacuoles, reducingthe size of its signal beyond the sensitivity of ourtechnique.NMR spectroscopy has been used for some time to

study the physico-chemical environment of tissuewater in vitro, and we have shown that it is possible toobtain similar information in vivo using quantitativeimaging techniques. We have been able to demon-strate that while the anatomical extent of a patholo-gical process can be accurately delineated by NMRimages, consideration of the quantitative NMR char-acteristics of the abnormal tissue can provide furtherinsight into the nature of the underlying changes.The application of these principles to patients has

permitted a distinction to be made between acutebrainstem lesions of the kind with which multiplesclerosis can present, and the chronic brainstemlesions of the established disease.24 The further devel-opment of this approach has obvious implications forearly diagnosis and the monitoring of therapy, andmay facilitate the distinction between different classesof pathological process by NMR imaging.

We are grateful to the Cerebral Oedema ResearchGroup for carrying out the measurements of water.content, and to Miss Jenny Small for expert technicalassistance with the morphological studies and illus-trations. The work was carried out on the MRIfacility provided by the Multiple Sclerosis Society,and was supported by a grant from the MRC.

References

1 Ormerod IEC, Douboulay EGH, McDonald WI.Imaging and Multiple Sclerosis. In: McDonald WI,Silverberg Dy-, eds. Multiple Sclerosis. London:Butterworths, 1986.

2 Mathur-De Vre R. Biochemical implications of therelaxation behaviour of water related to NMRimaging. Br J Radiol 1984;57:955-76.

3 Gazendam J, Go GK, van Zanten AK. Composition of

isolated edema fluid in cold-induced brain edema. JNeurosurg 1979;51:70-7.

4 Lee JC, Bakay L. Ultrastructural changes in the edema-tous central nervous system. Arch Neurol 1965;13:48-58.

5 Bakay L. Morphological and chemical studies in cerebraloedema: Triethyltin-induced oedema. J Neurol Sci1965;2:52-67.

6 Barnes D, McDonald WI, Tofts PS, Johnson G, LandonDN. Nuclear magnetic resonance imaging in experi-mental cerebral oedema. J Neurol NeurosurgPsychiatry 1986;49: 1341.

7 Young IR, Bailes DR, Burl M, et al. Initial clinical evalu-ation of a whole body nuclear magnetic resonancetomograph. J Comput Assist Tomog 1982;6(1): 1-18.

8 Go GK, Ebels EJ, Beks JWF, Ter Weeme CA. Thespreading of cerebral edema from a cold injury in cats.Psychiatry Neurol Neurochir 1967;70:403-1 1.

9 Lee JC, Bakay L. Ultrastructural changes in theedematous central nervous system. Arch Neurol1966;14:36-49.

10 Long DM, Maxwell RE, French LA. The effects ofglucosteroids upon cold induced brain edema.J Neuropathol Exp Neurol 197 1;30:680-97.

11 Cutler RWP, Watters GV, Barlow CF. I"25-labelledprotein in experimental brain edema. Arch Neurol1964;11:225-38.

12 Rasmussen LE, Klatzo I. Protein and enzyme changes incold injury edema. Acta Neuropathol (Berlin)1969;13: 12-28.

13 Crockard HA. Brain swelling, brain oedema and theblood-brain barrier. In: Crockard A, Hayward R,Hoff JT, eds. Neurosurgery. The scientific basis ofclinical practice. Oxford: Blackwell Scientific, 1985.

14 Partin JC, Partin JS, Schubert WK, McLaurin RL. Brainultrastructure in Reye's syndrome. J Neuropathol ExpNeurol 1975;34:425-44.

15 DiMauro S, Bonilla E, Zeviani M, et al. Mitochondrialmyopathies. Ann Neurol 1985;17:521-38.

16 Go GK, Edzes HT. Water in brain edema. Arch Neurol1975;32:462-5.

17 Naruse S, Horikawa Y, Tanaka C, etal. Nuclear mag-netic resonance studies on brain edema. J Neurosurg1982;56:747-52.

18 Daszkiewicz OK, Hennel JW, Lubas B. Protein magneticrelaxation and protein hydration. Nature 1963;200:1006-7.

19 Brandt-Zawadzki M, Bartkowski HM, Ortendahl DA,etal. Am J Neuroradiol 1984;5:125-9.

20 Bakay L, Kurland RJ, Parrish RG, et al. Nuclear mag-netic resonance studies in normal and edematous braintissue. Exp Brain Res 1975;23:241-8.

21 Foster KR, Resing HA, Garroway AN. Bounds on"Bound water": transverse nuclear magnetic relax-ation in barnacle muscle. Science 1976;194:324-6.

22 Rees S, Cragg BG, Everitt AV. Comparison of extra-cellular space in the mature and ageing rat brain usinga new technique. J Neurol Sci 1982;53:347-57.

23 Van Harreveld A. The extracellular space in the verte-brate central nervous system. In: Bourne GH, ed. TheStructure and Function of Nevous Tissue. Vol 4. NewYork and London: Academic Press, 1972.

24 Ormerod IEC, Bronstein A, Rudge P, et al. Nuclearmagnetic resonance imaging in clinically isolatedlesions of the brainstem. J Neurol NeurosurgPsychiatry 1987 (in press).

133



http://jnnp.bmj.com

/J N

eurol Neurosurg P


ownloaded from


quantitative nuclearmagnetic resonance imaging ... · when imaging the cold lesions, the...

Documents