J

O

PobIt

NES

a

b

Ec

lf

0d

ournal of Neuroradiology (2012) 39, 308—316

Available online at

www.sciencedirect.com

RIGINAL ARTICLE

erfusion magnetic resonance imaging: Comparisonf semiologic characteristics in first-pass perfusion ofrain tumors at 1.5 and 3 Tesla

RM de perfusion : comparaison des caractéristiques sémiologiques desumeurs cérébrales à 1,5 T et 3 T

atacha Mauza, Alexandre Krainika, Irène Tropresb, Laurent Lamalleb,lodie Sellierc, Omer Ekera, Florence Tahona, Jean-Francois Le Basa,ylvie Granda,∗

Clinique universitaire de neuroradiologie et IRM, CHU de Grenoble, BP 217, 38043 Grenoble cedex 9, FranceStructure fédérative de recherche, université Joseph-Fourier—Site santé, Grenoble institut de neurosciences, bâtimentdmond-J.-Safra, chemin Fortuné-Ferrini, 38706 La Tronche cedex, FranceUnité de qualitique et d’évaluation médicale, CHU de Grenoble, BP 217, 38043 Grenoble cedex 9, France

KEYWORDSMRI;Perfusion;Brain tumors;1.5 T;3 T

SummaryObjectives: To investigate whether using 3 Tesla (T) instead of 1.5 T modifies the data obtainedfrom first-pass perfusion in relation to the quantitative values of cerebral blood volume (CBV)and estimation of micro-vascular leakage (MVL). To describe the differences in data in thesetting of neuro-oncology cases and propose explanations based on the discrepancies.Material and methods: In total, 21 patients presenting an intracranial intra-axial space-occupying lesion underwent two MRI explorations, one at 1.5 T and another at 3 T, including afirst-pass perfusion sequence using sequence parameters, defined by the manufacturer Philips.Using a gamma variate analysis, the ratio of cerebral blood volume (rCBV) in tumor, peritumoral,and normal appearing areas was first assessed. After a global analysis, a subgroup analysis wasconducted according to the rCBV value measured at 1.5 T. Lastly, MVL was assessed based on

the signal intensity recorded above baseline after the passage of the contrast medium.Results: At 3 T, compared to 1.5 T data that are currently the reference, rCBV was constantlyand significantly over-evaluated (P = 0.0041 for all tumors), while MVL was constantly and sig-nificantly under-evaluated (P < 0.0001 for all tumors).

∗ Corresponding author.E-mail addresses: natachamauz@hotmail.fr (N. Mauz), akrainik@chu-grenoble.fr (A. Krainik), Irene.tropres@ujf-grenoble.fr (I. Tropres),

aurent.lamalle@ujf-grenoble.fr (L. Lamalle), esellier@chu-grenoble.fr (E. Sellier), oeker@chu-grenoble.fr (O. Eker),tahon@chu-grenoble.fr (F. Tahon), jflebas@chu-grenoble.fr (J.-F. Le Bas), sgrand@chu-grenoble.fr (S. Grand).

150-9861/$ – see front matter © 2012 Elsevier Masson SAS. All rights reserved.oi:10.1016/j.neurad.2011.12.004

Perfusion magnetic resonance imaging: Comparison of semiologic characteristics 309

Discussion: The increase in magnetic field strength along with the associated modifications insequence parameters led to variations in rCBV and MVL when measured using first-pass perfu-sion. In some cases, such as lymphomas, there was a loss of diagnostic information. It thereforeappears necessary to optimize the acquisition parameters to allow for radiologic semiology tobecome relevant again.© 2012 Elsevier Masson SAS. All rights reserved.

hpsme

P

P

O2wl2wilgm

afibd

tptAp

db

mdn

A

Im

cwcontrast-enhanced T1-w, T2-w, T2 -w and FLAIR-w. Other

Introduction

Using 3 Tesla (T) instead of 1.5 T does not significantly modifyimage contrast as far as conventional morphological inves-tigations are concerned. However, the use of a 3 T machineinvolves an increase in the magnetic susceptibility effect(T2*) and relaxation time (T1). These changes are linkedto modifications in the acquisition parameters defined bymanufacturers (Philips in our study), thus altering the semi-ology in functional investigations, particularly in relation tofirst-pass perfusion imaging data.

First-pass perfusion has developed considerably over thepast 10 years so that it has become part of the initial inves-tigations of brain tumors. The technique can be used forseveral purposes [1—4]:

• diagnosis: the characterization of brain tumor vascular-ization is complementary to the histological approach. Asemiology of different brain tumors was described usingdata obtained at the magnetic field strength of 1.5 T,which is now well-established. In some cases, such ascerebral lymphoma, combining the perfusion curve andmorphological data allows for an almost definite diagnosisprior to stereotactic biopsy [5,6]. For mixed tumors thatare both cystic and nodular, first-pass perfusion is ableto differentiate hypervascular hemangioblastomas frompilocytic astrocytomas, which are not associated to anincrease in cerebral blood volume (CBV) [7];

• prognosis: cerebral perfusion was shown to play an impor-tant role in grading gliomas. Indeed, the histological gradeis correlated with the ratio of cerebral blood volume(rCBV), permeability of tumor vessels [8—11], and extentof cellular proliferation [12];

• therapeutic management and follow-up of gliomas: thestereotactic biopsy is guided by the visualization ofhotspots, which indicate highly vascularized and con-sequently, more aggressive areas. Tumor response tochemotherapy and antiangiogenic agents may also beassessed by variations in rCBV [13].

The development of devices using a 3 T magnetic fieldled to changes in the acquisition parameters provided bymanufacturers, modifying the semiology in relation to boththe rCBV and curve position after the passage of the contrastmedium.

While some semiologic studies have been publishedcomparing different magnetic field strengths, they mostlyconcerned conventional MRI sequences, organs other than

the brain, and various pathologies not including brain tumors(mainly vascular, inflammatory, and psychiatric patholo-gies) [14—19]. To the extent of our knowledge, no study

ftc

as compared the semiology of brain tumors using first-ass perfusion sequences at 1.5 T and 3 T. The aim of ourtudy was therefore to describe the differences in CBV andicro-vascular leakage (MVL) quantitatively and propose

xplanations for the observed discrepancies.

atients and methods

atients

ur prospective study compared the perfusion data of1 patients at 1.5 T and 3 T, comprising ten men and 11omen presenting intracranial intra-axial space-occupying

esions. Investigations were performed between September9, 2009, and October 26, 2010. The median age of patientsas 56 years (range: 34—84). The space-occupying lesions

ncluded 12 high-grade gliomas, four non-Hodgkin malignantymphomas, two lung metastases, one abscess, one low-rade glioma (hypervascular oligodendroglioma), and oneedulloblastoma.During the first examination, the tumor was identified

nd initially characterized, with 13 patients undergoing therst-pass perfusion at 1.5 T and eight at 3 T. The choiceetween both magnetic field strengths was made randomly,epending on the availability of the devices.

The second exploration was presurgical (biopsy or resec-ion) and it was only performed if required based on theatients’ clinical state or therapeutic care. The device withhe alternate magnetic field strength was used in each case.

second examination was not performed for the sole pur-ose of the study.

The median delay between the two examinations was 7ays, except for four patients for whom the delay was longer,ut inferior to 45 days.

Prior to surgery, patients received a symptomatic treat-ent consisting of corticoids, analgesics and antiepilepticrugs, which remained unmodified between the two exami-ations.

cquisitions

mages at 1.5 T were acquired using a Philips Achievaachine, and those at 3 T using a Philips Intera machine.During the first examination, the perfusion sequence

ompleted an initial assessment of the cerebral tumorith following morphological sequences: T1-weighted (-w),

*

unctional sequences, such as diffusion-weighted and spec-roscopy, were sometimes acquired depending on theirlinical relevance to the diagnosis.

3 N. Mauz et al.

oufi

(aa

ep

•

•

wl

P(

Te

petwn

•

•

•

pfRrWti

fa

Figure 1 Modelization of the first-pass perfusion curve for theq

aIrR

s

mpV

obanWo(

l

•

•

10

The second examination was aimed at planning the biopsyr surgical resection. In addition, a contrast-enhanced vol-metric sequence was performed immediately after therst-pass perfusion sequence.

The injected gadolinium dose was 0.2 mL/kg0.1 mmoL/kg) of gadoteric acid (Dotarem, Guerbet®)t a speed of 6 mL/s in a 20-gauge catheter with anutomatic injector, according to the usual protocols.

Perfusion images were obtained using a gradient-echocho-planar imaging (GE—EPI) sequence with the followingarameters:

at 1.5 T, repetition time (TR) was 507 ms, echo time (TE)30 ms, flip angle 40◦, field of view 230 mm, matrix 90*128,and slice thickness 5.5 mm with a multi-shot technique.In total, 12 contiguous slices were acquired in axial planeevery 1.5 sec for 60 sec;

at 3 T, TR was 1635 ms, TE 40 ms, flip angle 75◦, field ofview 224 mm, matrix 112*128, and slice thickness 4.0 mmwith one-shot technique. In total, 25 contiguous sliceswere acquired in axial plane every 1.5 sec for 60 sec.

A gradient-echo (GE) rather than spin-echo (SE) sequenceas chosen due to its better sensitivity to both microcircu-

ation and larger vessels [20].

ost-processing with relative valuesgamma-variate function)

he perfusion software provided by Philips was used (Philipsxtended MR workspace R2.3.6.1).

Examinations were analysed chronologically for eachatient, starting with conventional MRI sequences (contrast-nhanced T1-w and FLAIR-w sequences) in order to localizehe tumor and characterize its morphology. Perfusion imagesere then analysed before the operator read the results ofeuroradiologic and histological investigations.

Perfusion data was then analysed:

selection of the tumor region of interest (ROI), basedon the maximal CBV value using CBV map and contrast-enhanced T1-w images. The analysis was thus based onhomogeneous areas, corresponding to the hotspots oftumor neovascularization. Tumor feeder arteries were notincluded as ROIs;

selection of peritumoral ROI, defined as the area appear-ing FLAIR hyperintense but not contrast-enhancing;selection of the reference ROI in the normal appearingwhite matter.

In most cases, the tumor ROI was copied before beingasted in the peritumoral or white matter region. For aew patients, the shape of peritumoral and white matterOIs had to be adapted in order to stay strictly within theseegions, although we tried to maintain the number of pixels.e tried to keep the ROI size as similar as possible between

he different patients, around 15 pixels but we had to adapt

t to the hotspots’ size and shape in each tumor.

The software provided by the manufacturer did not allowor superimposition of ROIs between images taken at 1.5 Tnd 3 T, so the second images were placed on slices as closed

•

uantification of micro-vascular leakage.

s possible to the first ones based on anatomical criteria.n the case of hotspots appearing between the two explo-ations, the maximum CBV value was chosen to position theOI.

For each ROI, both the CBV and distance between thetabilization level of the curve and baseline were studied.

The rCBV was measured between the (a) tumor and nor-al appearing white matter (rCBVt = CBVt / CBVwm), and (b)eritumoral area and normal appearing white matter (rCB-pt = CBVpt / CBVwm).

The distance to baseline was measured on the y-axisn each curve and expressed in percentage. The distanceetween the stabilization level of the curve and baselinellows for an estimation of MVL, which is an essential diag-ostic criterion that should be used in daily practice [21,22].e calculated the distance using three different formulas inrder to determine which would be most useful in diagnosisFig. 1).

In the following formulae, a is the baseline, b the peakevel, and c the level at which the curve stabilizes:

first formula: distance = (c−a) / a. The formula does nottake into account the depth of the peak, thus being devoidof major measurement errors. However, the formula is notvery relevant from a clinical point of view, as it does nottake account of signal loss extent and thus the importanceof the hyper-vascularization. Therefore, this formula wasnot included in our statistical analysis;

second formula: distance = (c−a) / (a−b). This formulacorresponds to the ratio between the MVL extent (dis-tance between the stabilization level and baseline) andthe vascularization extent (depth of the peak). This for-mula was used in our statistical analysis because, in ourview, it was the most pertinent from a clinical point ofview as it takes into account the two most importantparameters in the analysis of a first-pass curve;

third formula: distance = (c−a) / (c−b). This formula com-

pares the distance between the stabilization level andbaseline with the distance between the stabilization andpeak levels. This formula was not used in our analysis, as

logi

t(t

D

OeeTsocsrwunpi

Cs

Ooal[a

tgtvatbcab

Perfusion magnetic resonance imaging: Comparison of semio

the MVL appears twice in the formula, both in the numer-ator (c−a) and the denominator (c−b = [c−a) + (a−b]).

Statistical analysis

Statistical analysis was performed using Statview software5.0. Non-parametric tests were used because our cohortdistribution was not following the Gauss curve.

The Wilcoxon test was used to compare the 1.5 T and 3 Tsequences in relation to the rCBVs and distance to baseline.

Firstly, rCBV [tumor] and rCBV [peritumoral] were com-pared between 1.5 T and 3 T. In addition, rCBV [tumor] wascompared between 1.5 T and 3 T in two subgroups: markedlyhypervascular lesions (MHL, n = 16, rCBV [tumor] > 1.8quantified at 1.5 T), including gliomas, metastases, andmedulloblastoma, and slightly hypervascular lesions (SHL,n = 5, rCBV [tumor] < 1.8 quantified at 1.5 T), including lym-phomas and abscess. The threshold of 1.8 chosen for thesubgroup analysis was appropriated according of its diag-nosis and prognostic value, particularly concerning gliomas[23]. The same process was followed regarding the distanceto baseline, with a comparison of all tumors, MHLs, andSHLs. In order to determine if the distance to baseline wasassociated with a particular subgroup (MHL or SHL) at 1.5 Tand 3 T, the Kolmogorov-Smirnov test was used.

The significance level was set at P < 0.05.

Results

Table 1 presents the data of rCBV and distance to baselinefor tumor areas along with the associated P-values.

Analysis of ratio of cerebral blood volume (rCBV)

The rCBV [tumor] was significantly higher at 3 T than at 1.5 Tfor all subgroups (P = 0.0041 for all tumors, P = 0.0174 forMHL, and P = 0.0431 for SHL) (Figs. 2—4, Table 1). The dif-ference in rCBV [peritumoral] between 1.5 T and 3 T was notstatistically significant (P = 0.9584).

Analysis of distance to baseline

Signal intensity after the peak was significantly lower at 3 Tthan at 1.5 T for all subgroups (P < 0.0001 for all tumors,P = 0.0007 for MHL, and P = 0.0431 for SHL) (Figs. 2—3, Fig. 5,Table 1). The difference in the distance to baseline based on

E

Ta

Table 1 Medians for the ratio of cerebral blood volume (rCBV) [tuand subgroups. P-values are also expressed.

rCBVt 1.5 T

Median for all tumors 5.6

P = 0.0014

Median for MHL (gliomas,metastases, medulloblastoma)

6.7

P = 0.0174

Median for SHL (lymphomas,abscess)

1.5

P = 0.0431

c characteristics 311

he tumor subgroup (MHL and SHL) was significant at 1.5 TP = 0.0010), but not at 3 T (P = 0.1996) (Kolmogorov-Smirnovest).

iscussion

ver the past 15 years, many studies have shown the ben-fits of perfusion MRI in the diagnosis and pre-treatmentvaluation of brain tumors, as well as during follow-up.he majority of these studies were performed using 1.5 Tystems. The recent growth of 3 T MRI has led to the semiol-gy of perfusion imaging being modified in terms of botherebral blood volume and curve position after the pas-age of contrast medium, although morphological data hasemained similar. Our study showed an increase in rCBVhen measured at 3 T compared to 1.5 T in addition to anndervaluation of MVL. These differences, due to the mag-etic field strength and associated variations in acquisitionarameters, need to be recognized and taken into accountn clinical practice.

oherence of 1.5 T data with the acceptedemiology

ur data obtained at 1.5 T conformed to the accepted semi-logy of brain tumors [10,24]: Gliomas and metastases had

high rCBV [tumor] (median = 6.7) and low rise above base-ine (median = +7%); lymphomas and abscesses had a rCBVtumor] approaching 1 (median = 1.5) and very high risebove baseline (median = +120%) [5,6,25].

Several studies [26,27] quantified MVL, i.e., the distanceo baseline (T1 effect, cf. infra). This quantification was notenerally made to determine its diagnostic value, but rathero correct the T1 effect and thus, obtain more accurate CBValues. A few studies [27] suggested that this quantitativessessment may be of diagnostic value or correlated withumor grade [8,10]. However, these calculations were eitherased on a double-echo sequence [8] or used permeabilityoefficient measures [10]. As these methods were not avail-ble in our study, we chose a calculation method that woulde both relevant and easy to use in daily clinical practice.

xplanations for the 1.5 T and 3 T differences

1 corresponds to the longitudinal spin relaxation timefter stimulation (Table 2). The longer the TR, the

mor] and distances to baseline at 1.5 T and 3 T for all tumors

rCBVt 3 T BLt 1.5 T BLt 3 T

7.6 +10.0% −33.0%P < 0.0001

10.1 +6.5% −34.5%P = 0.0004

1.9 +120.0% −27.0%P = 0.0431

312 N. Mauz et al.

Figure 2 Glioblastoma explored at 1.5 T (A) and 3 T (B); T1-weighted sequence (upper left), cerebral blood volume (CBV) map(upper right), first-pass curve (below); rCBV = 8.1 at 1.5 T and rCBV = 11.8 at 3 T. The first-pass curve returned to its initial levelafter the gadolinium bolus at 1.5 T, but remained significantly below baseline at 3 T.

Figure 3 Cerebral lymphoma explored at 1.5 T (A) and 3 T (B); T1-weighted sequence (upper left), cerebral blood volume (CBV)map (upper right), first-pass curve (below); rCBV = 1.5 at 1.5 T, rCBV = 3.06 at 3 T. The first-pass curve exceeded baseline after thegadolinium bolus at 1.5 T, but remained significantly below baseline at 3 T.

Perfusion magnetic resonance imaging: Comparison of semiologic characteristics 313

Figure 4 Medians for the ratio of cerebral blood volume (rCBV) in tumor area for all tumors (blue), markedly hypervascular lesions(red), and slightly hypervascular lesions (green).

Figure 5 Medians for the distance between the stabilization level of the curve and baseline in tumor area for all tumors (blue),markedly hypervascular lesions (red), and slightly hypervascular lesions (green).

Table 2 Evolution of T1 and T2* effects depending on field strength, sequence parameters, and gadolinium concentration;consequences on global contrast and explanation of the differences observed between 1.5 T and 3 T.

1.5 Tesla 3 Tesla

Sequence weighting T1-weighting increased bygadolinium injection

T1-weighting ↓↓ (field ↑, TR ↑, �↑)T2* weighting ↑↑ (field ↑, E ↑)

During the Bolus pass High [Gd]T1 effect ↑, T2 effect ↑↑T1 effect < T2* effectLoss of signal

High [Gd]→T1 effect weaker than at 1.5 T, 2*effect more pronounced→T1 effect « T2* effect→Loss of signal +++→Area above curve ↑→Calculated CBV ↑

After the Bolus pass (leakage ofgadolinium in extracellularextravascular space)

Low [Gd]→T1 effect ↓ et T2* effect ↓↓→T1 effect » T2* effect→Signal ↑→Ascension above the baselevel

Low [Gd]T1 effect ↓ et T2* effect ↓↓But marked base T2* weighting 2*effect remains > T1 effectNo ascension above the base level

3

wfuafitTw

fiac

mfe

faaae(lTittws

oTr

atTairwwwi

battG

pca

mgsmwm

Cp

rAopqheieDiu

BTlalaibomeia

P

Aucvao

vetatdal

uctrwf

14

eaker was the contrast between two tissues of dif-ering T1, i.e., the tissue with the longer T1 ‘‘catchesp’’ to the shorter-T1 tissue. A long TE corresponds ton increase in T2* contrast (transverse relaxation andeld inhomogeneities). With a gradient-echo sequence,he flip angle � must also be taken into account:he greater � is, the more T1-weighted the sequenceill be.

Even with similar sequence parameters, an increase ineld strength alone leads to a convergence of T1 relax-tion curves of different tissues, therefore decreasing T1ontrast.

Increasing the field strength also results in an increase inagnetic susceptibility effect (T2* effect), which is there-

ore higher at 3 T. To obtain the same T2*-weighting at 1.5 T,cho time should be doubled [26].

In our study, predefined sequence parameters were dif-erent at 1.5 T and 3 T. The choice of a one-shot techniquet 3 T, which is systematically suggested by manufacturersnd in line with the development of the MRI systems, wasssociated with an increase in TR. These variations influ-nce the global signal weighting. The lengthened TE at 3 T40 ms vs. 30 ms at 1.5 T) increased T2*-weighting, while theengthened TR at 3 T (1635 ms vs. 507 ms at 1.5 T) decreased1-weighting. The wider flip angle at 3 T (75◦ vs. 40◦ at 1.5 T)

ncreased T1-weighting, but not enough to compensate forhe TR-lengthening. With the predefined sequence parame-ers at 3 T, T2*-weighting of the signal was enhanced overall,hile T1-weighting was reduced, independently of the field

trength.The injection of gadolinium intensified the competition

f T1 and T2* effects on the signal intensity, with a rise in2* contrast resulting in signal loss, but a rise in T1 contrastesulting in signal increase.

The relationship between gadolinium concentration (Gd)nd tissue T1 is not linear [26—28]. When Gd concen-ration increases but remains below a certain threshold,1 decreases, leading to an increase in the T1 effectnd thus, signal intensity. However, once the thresholds overcome, T1 increases along with Gd concentration,educing the influence of T1 in the global sequenceeighting. In our study, as in most clinical situations, itas highly probable that this threshold was not reached,ith the T1 effect being greater as Gd concentration

ncreased.The relationship between Gd concentration and T2* may

e considered to be inversely proportional at 1.5 T [20,27],s T2* decreases with the increase in Gd. Consequently,here is an increase of T2* contrast, with the T2* influence inhe global sequence weighting increasing linearly with thed rise.

At 3 T, it appears that the relationship is not inversely pro-ortional [26], with the T2* effect increasing faster than Gdoncentration. The T2* effect thus increased more markedlyt 3 T compared to 1.5 T.

Thus, the more elevated the Gd concentration, theore marked is the influence of T2* effect in the

lobal sequence weighting, particularly when the field

trength is increased to 3 T. The T2* effect is maxi-al during the bolus pass, quickly diminishing thereafterhen gadolinium is only present as extravasated contrastedium.

uute

N. Mauz et al.

onsequences on the interpretation in clinicalractice

CBVt 3 T, a constant and statistically significant overvaluationf rCBV was observed compared to 1.5 T. In contrast to mor-hological analyses obtained with conventional sequences,uantitative results obtained with perfusion sequences wereighly dependent on the magnetic field and sequence param-ters. Using 3 T instead of 1.5 T thus requires caution whenntegrating the results of the perfusion sequence in a pre-xisting database in order to avoid interpretation mistakes.uring the follow-up of tumor evolution and aggressiveness

n terms of neovascularization, it also appears necessary tose the same machine and sequences.

aselinehe signal intensity after the peak, i.e., distance to base-

ine, was inferior at 3 T compared to 1.5 T. This loss of risebove baseline led to a decreased correlation with the histo-ogical analysis: For instance, increased MVL represented by

rise above baseline was correlated to fenestrated capillar-es under the microscope. There was a significant associationetween the distance to baseline and tumor subgroup (MHLr SHL) at 1.5 T (P = 0.0010), but not at 3 T (P = 0.1996). Theove to 3 T and its implications in terms of sequence param-

ters may lead to a loss of diagnostic information, especiallyn the cases of cerebral lymphomas, abscesses, pilocyticstrocytomas, and even pseudo-tumoral lesions.

ost-processing

t present, gamma-variate analysis is the most commonlysed method in the post-processing setting in daily clini-al practice. This analysis provides relative values for tumorascularization in comparison to the perfusion of normalppearing white matter, and allows for a partial correctionf contrast medium recirculation.

After defining an arterial input function (AIF), the decon-olution analysis would supply absolute values of CBV,xpressed in mL of blood per 100 g of brain tissue. Because ofhe use of this unit, the approach has clear advantages, as itllows for independent quantitative data analysis. However,hese absolute CBV values are not widely used, and theiriagnostic predictive value is not well known. This may beccounted for by the difficulty in positioning the AIF and theack of appropriate software.

Another way to obtain absolute CBV values would be tose the arterial spin labeling (ASL) technique. This methodonsists in marking the circulating spins before they crosshe ROI with an inverse magnetization. As the signal/noiseatio remains quite low with the ASL method but increasesith the magnetic field, it is a clear advantage to per-

orm it at 3 T. If the deconvolution analysis were easier to

se, it would be interesting to compare the absolute val-es obtained with these two methods in order to determinehe most useful one in terms of diagnostic and prognosticvaluation of the tumor.

logi

[

[

[

[

[

[

[

[

[

[

[

[

Perfusion magnetic resonance imaging: Comparison of semio

Strengths and weaknesses of the study

Despite the small number of patients, our results are highlysignificant.

In our study, CBV values were not corrected with a pre-bolus injection of contrast medium, which would fill theinterstitial space, thus reducing the bolus leakage throughthe disrupted blood-brain barrier and T1 effect during thebolus pass [29]. This technique allows for a more accurateand reliable measurement of rCBV in the tumor. However,it prevents the rise of the curve above baseline, resultingin the loss of key diagnostic information. In addition, theproblem of higher doses of gadolinium must be stressed.

Other corrections were made without the pre-bolusinjection of gadolinium, but required changes in thesequence parameters or the use of particular post-processing algorithms. We conducted the study as closeto daily clinical practice as possible, thus using sequenceparameters and post-processing algorithms, which are pre-defined by the manufacturer and available in day-to-daypractice.

Conclusion

Compared to 1.5 T, the use of 3 T, along with variationsof predefined sequence parameters, led to a significantincrease in tumor CBV and significant decrease in signalintensity after the bolus pass, via reduced T1 effect andenhanced T2* effect. These factors clearly modified the first-pass curve and hindered its interpretation, especially in thecase of SHLs, in which CBV was shown to be much higher at3 T compared to 1.5 T, with MVL no longer being reflectedon the curve.

Hence, perfusion sequence appeared less efficient at 3 Tcompared to 1.5 T for the diagnosis of brain tumors. Ourstudy emphasized the need for sequences to be adapted inorder to regain the diagnostic information brought by thecurve rise above the baseline and obtain rCBV values thatcan be integrated in existing databases. The radiologist mustacquire the physical knowledge required to interpret thesequence parameters and thus, stop relying completely onthe predefined sequences of manufacturers.

Disclosure of interest

The authors have not supplied their declaration of conflictof interest.

References

[1] Knopp E, Cha S, Johnson G, et al. Glial neoplasms:dynamic contrast-enhanced T2*-weighted MR imaging. Radiol-ogy 1999;211:791—8.

[2] Wong J, Provenzale J, Petrella J. Perfusion MR imaging of brainneoplasms. Am J Roentgenol 2000;174:1147—57.

[3] Rollin N, Guyotat J, Streichenberger N, Honnorat J, TianMinh V, Cotton F. Clinical relevance of diffusion and perfu-sion MRI in assessing intra-axial brain tumors. Neuroradiology2006;48:150—9.

c characteristics 315

[4] Weber M, Zoubaa S, Schlieter M, et al. Diagnostic performanceof spectroscopic and perfusion MRI for distinction of braintumors. Neurology 2006;66:1899—906.

[5] Cotton F, Ongolo-Zogo P, Louis-Tisserand G, et al. IRMde diffusion-perfusion dans l’évaluation des lymphomescérébraux. J Neuroradiol 2006;33:220—8.

[6] Hartmann M, Heiland S, Harting I, et al. Distinguishing ofprimary cerebral lymphoma from high-grade glioma withperfusion-weighted magnetic resonance imaging. Neurosci Lett2003;338:119—22.

[7] Bing F, Kremer S, Lamalle L, et al. Value of perfusion MRI in thestudy of pilocytic astrocytoma and hemangioblastoma: prelim-inary findings. J Neuroradiol 2009;36:82—7.

[8] Law M, Yang S, Babb J, et al. Comparison of cerebral bloodvolume and vascular permeability from dynamic susceptibilitycontrast-enhanced perfusion MR imaging with glioma grade.Am J Neuroradiol 2004;25:746—55.

[9] Lev M, Ozsunar Y, Henson J, et al. Glial tumor grading andoutcome prediction using dynamic spin-echo MR susceptibil-ity mapping compared with conventional contrast-enhanced.MR: confounding effect of elevated rCBV of oligodendrogliomas[corrected]. Am J Neuroradiol 2004;25:214—21.

10] Roberts H, Roberts T, Brasch R, Dillon W. Quantitative mea-surement of microvascular permeability in human brain tumorsachieved using dynamic contrast-enhanced MR imaging: corre-lation with histologic grade. Am J Neuroradiol 2000;21:891—9.

11] Jiang Z, Le Bas J, Grand S, et al. Prognostic value of perfu-sion MR imaging in patients with oligodendroglioma: a survivalstudy. J Neuroradiol 2011;38:53—61.

12] Price S, Green H, Dean A, Joseph J, Hutchinson P, Gillard J. Cor-relation of MR relative cerebral blood volume measurementswith cellular density and proliferation in high-grade gliomas: animage-guided biopsy study. Am J Neuroradiol 2011;32:501—6.

13] Dhermain F, Hau P, Lanfermann H, Jacobs A, van denBent M. Advanced MRI and PET imaging for assessment oftreatment response in patients with gliomas. Lancet Neurol2010;9:906—20.

14] Wattjes M, Lutterbey G, Harzheim M, et al. Higher sensitivityin the detection of inflammatory brain lesions in patients withclinically isolated syndromes suggestive of multiple sclerosisusing high field MRI: an intraindividual comparison of 1.5 T with3.0 T. Eur Radiol 2006;16:2067—73.

15] Notohamiprodjo M, Dietrich O, Horger W, et al. Diffusiontensor imaging (DTI) of the kidney at 3 tesla-feasibility, pro-tocol evaluation and comparison to 1,5 tesla. Invest Radiol2010;45:245—54.

16] Sjobakk T, Lundgren S, Kristoffersen A, et al. Clinical 1H mag-netic resonance spectroscopy of brain metastases at 1.5 T and3 T. Acta Radiol 2006;47:501—8.

17] Kamada K, Kakeda S, Ohnari N, Moriya J, Sato T, Korogi Y. Signalintensity of motor and sensory cortices on T2-weighted andFLAIR images: intraindividual comparison of 1.5 T and 3 T MRI.Eur Radiol 2008;18:2949—55.

18] Keihaninejad S, Heckemann R, Fagiolo G, Symms M, HajnalJ, Hammers A. A robust method to estimate the intracranialvolume across MRI field strengths (1.5 T and 3 T). Neuroimage2010;50:1427—37.

19] Stankiewicz JM, Glanz BI, Healy BC, et al. Brain MRI lesion loadat 1.5 T and 3 T versus clinical status in multiple sclerosis. JNeuroimaging 2011;21:e50—6.

20] Zaharchuk G. Theoretical basis of heomdynamic MR imagingtechniques to measure cerebral blood volume, cerebral bloodflow, and permeability. Am J Neuroradiol 2007;28:1850—8.

21] Cha S, Lupo J, Chen M, et al. Differentiation of glioblastoma

multiforme and single brain metastasis by peak height andpercentage of signal intensity recovery derived from dynamicsusceptibility-weighted contrast-enhanced perfusion MR imag-ing. Am J Neuroradiol 2007;28:1078—84.

3

[

[

[

[

[

[

[

[blood volume maps corrected for contrast agent extrava-

16

22] Mangla R, Kolar B, Zhu T, Zhong J, Almast J, Ekholm S. Per-centage signal recovery derived from MR dynamic susceptibilitycontrast imaging is useful to differentiate common enhancingmalignant lesions of the brain. Am J Neuroradiol 2011 [Epubahead of print].

23] Law M, Young RJ, Babb JS, et al. Gliomas: predicting time toprogression or survival with cerebral blood volume measure-ments at dynamic susceptibility-weighted contrast-enhancedperfusion MR imaging. Radiology 2008;247:490—8.

24] Chiang I, Kuo Y, Lu C, et al. Distinction between high-gradegliomas and solitary metastasis using peritumoral 3 T magneticresonance spectroscopy, diffusion, and perfusion imagings.Neuroradiology 2004;46:619—27.

25] Holmes T, Petrella J, Provenzale J. Distinction betweencerebral abscesses and high-grade neoplasms by dynamicsusceptibility contrast perfusion MRI. Am J Roentgenol2004;183:1247—52.

N. Mauz et al.

26] de Bazelaire C, Rofsky N, Duhamel G, et al. Combined T2*

and T1 measurements for improved perfusion and permeabilitystudies in high field using dynamic contrast enhancement. EurRadiol 2006;16:2083—91.

27] Vonken E, van Osch M, Bakker C, Viergever M. Simultaneouslyquantitative cerebral perfusion and Gd-DTPA extravasationmeasurement with dual-echo dynamic susceptibility contrastMRI. Magn Reson Med 2000;43:820—7.

28] Kuperman V, Alley M. Differentiation between the effects of T1and T2* shortening in contrast-enhanced MRI of the breast. JMagn Reson Imag 1999;9:172—6.

29] Boxerman J, Schmainda K, Weisskoff R. Relative cerebral

sation significantly correlate with glioma tumor grade,whereas uncorrected maps do not. Am J Neuroradiol 2006;25:259—67.