Comparison of quantitative dynamic susceptibility-contrast MRI perfusion estimates obtained using different contrast-agent administration schemes at 3 T

Comparison of quantitative dynamic susceptibility-contrast MRI perfusion estimates obtained using different contrast-agent administration schemes at 3 T

European Journal of Radiology 75 (2010) e86–e91 Contents lists available at ScienceDirect European Journal of Radiology journal homepage: www.elsevi...

498KB Sizes 1 Downloads 25 Views

European Journal of Radiology 75 (2010) e86–e91

Contents lists available at ScienceDirect

European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad

Comparison of quantitative dynamic susceptibility-contrast MRI perfusion estimates obtained using different contrast-agent administration schemes at 3 T Ronnie Wirestam a,∗ , Oliver Thilmann a,1 , Linda Knutsson a , Isabella M. Björkman-Burtscher b , Elna-Marie Larsson c , Freddy Ståhlberg a,b a

Department of Medical Radiation Physics, Lund University, University Hospital, SE-22185 Lund, Sweden Department of Diagnostic Radiology, Lund University, University Hospital, SE-22185 Lund, Sweden c Division of Radiology, Department of Oncology, Radiology and Clinical Immunology, Uppsala University, Akademiska sjukhuset, SE-75185 Uppsala, Sweden b

a r t i c l e

i n f o

Article history: Received 15 July 2009 Accepted 31 July 2009 Keywords: Magnetic resonance imaging Perfusion Cerebral blood flow Contrast agent Dosage

a b s t r a c t Absolute cerebral perfusion parameters were obtained by dynamic susceptibility contrast magnetic resonance imaging (DSC-MRI) carried out using different contrast-agent administration protocols. Sixteen healthy volunteers underwent three separate DSC-MRI examinations each, receiving single-dose (0.1 mmol/kg b.w.) gadobutrol, double-dose gadobutrol and single-dose gadobenate-dimeglumine on different occasions. DSC-MRI was performed using single-shot gradient-echo echo-planar imaging at 3 T. The arterial input functions (AIFs) were averages (4–9 pixels) of arterial curves from middle cerebral artery branches, automatically identified according to standard criteria. Absolute estimates of cerebral blood volume (CBV), cerebral blood flow (CBF) and mean transit time (MTT) were calculated without corrections for non-linear contrast-agent (CA) response in blood or for different T2* relaxivities in tissue and artery. Perfusion estimates obtained using single and double dose of gadobutrol correlated moderately well, while the relationship between estimates obtained using gadobutrol and gadobenate-dimeglumine showed generally lower correlation. The observed degree of CBV and CBF overestimation, compared with literature values, was most likely caused by different T2* relaxivities in blood and tissue in combination with partial-volume effects. The present results showed increased absolute values of CBV and CBF at higher dose, not predicted by the assumption of a quadratic response to contrast-agent concentration in blood. This indicates that the signal components of measured AIFs were not purely of arterial origin and that arterial signal components were more effectively extinguished at higher CA dose. This study also indicates that it may not be completely straightforward to compare absolute perfusion estimates obtained with different CA administration routines. © 2009 Elsevier Ireland Ltd. All rights reserved.

1. Introduction Attempts to measure perfusion parameters in absolute terms by dynamic susceptibility contrast magnetic resonance imaging (DSC-MRI) have, despite application of apparently appropriate theory, often been characterized by elevated values of cerebral blood volume (CBV) and cerebral blood flow (CBF) [e.g., 1,2], in this report referred to as pseudo-absolute estimates. It is reasonable to ascribe these systematic overestimations to problems with the registration of the arterial input function (AIF), and a corresponding underestimation of the arterial concentration time integral. A number of issues may be of relevance to reduced AIF curve areas, for example, partial volume effects, arterial signal saturation and local

∗ Corresponding author. Fax: +46 46 17 85 40. E-mail address: [email protected] (R. Wirestam). 1 Presently at Siemens Healthcare, Germany. 0720-048X/$ – see front matter © 2009 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ejrad.2009.07.038

geometric distortions at peak concentration, as well as different effective transversal relaxivities of the contrast agent in arterial and capillary environments [3]. Another potential complication is the observation of a quadratic rather than linear relationship in blood between the contrast-agent induced change in the transversal relaxation rate (R2*) and the concentration when gradient-echo pulse sequences are employed [4]. Even though the majority of clinical DSC-MRI examinations are applied to studies of focal perfusion changes in relative terms, there are still a number of justifications for investigating the characteristics of absolute DSC-MRI perfusion estimates. Firstly, there is indeed an interest in using absolute or pseudo-absolute DSC-MRI perfusion estimates in humans, for example, for assessment of therapeutic response [5], for monitoring of the response to hemodynamic challenges [1] and for studying correlations between CBF and other physiological conditions [6]. Studies of this type are likely to benefit from further information about, for example, repeatability and contrast-agent dose dependence. Secondly, and perhaps even more

R. Wirestam et al. / European Journal of Radiology 75 (2010) e86–e91

importantly, experimentally obtained absolute perfusion estimates reflect the net effect of a number of complications of quite different nature. Hence, experimental absolute perfusion estimates serve as a complement to theoretical predictions and simulations, and may contribute to the understanding of the dominating sources of error present during a standard DSC-MRI experiment. For example, a quadratic arterial R2* response to increased contrast-agent concentration [4], in combination with the tissue T2* relaxivity predicted by Kjølby et al. [3], implies that CBV (and CBF) estimates would decrease with increasing CA dose, if the non-linearity is not corrected for, and it might be interesting to see how this effect compares with the competing partial-volume effects and increased arterial signal decay at higher dose [7]. Finally, DSC-MRI investigations aiming at absolute quantification are generally still rather sparse at 3 T. Repeated DSC-MRI experiments using the same contrast-agent dose or different contrast-agent doses and/or molarities have previously been accomplished in humans [1,8–13]. However, most investigations originate from 1.5 T MRI units, and the degree of absolute quantification varies. Some reports included only relative values [8–11,13], and in some cases only population averages and standard deviations were presented [1], i.e. it is not possible to assess the degree of correlation between individual data points from different experiments. At 3 T, Manka et al. compared DSC-MRI results obtained using 0.05 and 0.1 mmol/kg body weight (b.w.) of gadolinium contrast agent, but conventional hemodynamic parameters were not calculated since the study design did not include AIF registration or deconvolution [11]. Similarly to the present study, Alger et al. [12] compared individual absolute estimates of CBV and CBF, obtained at gadolinium contrast-agent doses of 0.1 and 0.2 mmol/kg b.w., but their study was conducted at 1.5 T and without repositioning of the subject between the two measurements. In the present study, 16 healthy volunteers underwent three 3 T DSC-MRI investigations each, receiving 0.1 mmol/kg b.w. gadobutrol (Gadovist® 1.0), 0.2 mmol/kg b.w. gadobutrol and 0.1 mmol/kg b.w. gadobenate-dimeglumine (MultiHance® ) on different occasions. Results regarding concentration-versus-time curves, arterial bolus widths, grey-to-white matter ratios and the diagnostic usefulness of relative CBV and CBF maps from these examinations have previously been published [14]. In light of the emerging interest in absolute perfusion quantification by DSCMRI, as well as novel findings concerning AIF properties [7,15], varying T2* relaxivities in blood and tissue [3] and R2*-versusconcentration relationships in blood [4], an additional analysis of this rather extensive volunteer material, with respect to the global average of the absolute CBV, CBF and mean transit time (MTT) estimates, seemed highly justified. These recent investigations, exploring the basic mechanisms of DSC-MRI, have made it quite clear that the dependence of DSC-MRI perfusion estimates on the employed contrast-agent administration protocol cannot be fully assessed by a comparison in relative terms only.

2. Materials and methods 2.1. Subjects The study included 16 healthy volunteers (8 males and 8 females) of mean age 22.9 years (range 20–25 years) and with a mean weight of 70 kg (range 58–81 kg). Each volunteer underwent three separate DSC-MRI examinations, receiving, in randomized order, single-dose (0.1 mmol/kg b.w.) and doubledose (0.2 mmol/kg b.w.) gadobutrol (Gadovist® 1.0, Schering AG, Berlin, Germany) and single-dose gadobenate-dimeglumine (MultiHance® , Bracco S.p.A., Milano, Italy) on different occasions. Two subsequent examinations were separated by a minimum

e87

of 3 days (mean 4.9 days) to ensure complete excretion of the contrast agent, and the time between the first and the last examination was on average 11.9 days (maximum 18 days). A plasma creatinin test, with a required concentration below 90 ␮M, was performed to ensure proper excretion of the contrast agent within the period between the examinations. The volunteers were instructed to adhere to their normal caffeine intake patterns on the days of the examinations, but to avoid caffeine intake within one hour before the start of the examination. The study was approved by the local ethics committee and written informed consent was obtained from each volunteer before participation. 2.2. MRI experiments DSC-MRI was performed using a 3 T MRI unit (Magnetom Allegra, Siemens Medical Systems, Erlangen, Germany). Dynamic single-shot gradient-echo (GRE) echo-planar imaging (EPI) was employed during 90 s at a temporal resolution of 1.5 s, with flip angle 90◦ , echo time (TE) 21 ms, slice thickness 5 mm, interslice gap 1.5 mm, field of view 210 mm × 210 mm, acquisition matrix 72 × 128 and image matrix 128 × 128. The volume of interest consisted of 20 transversal-to-coronal slices, positioned so that the second lowest slice intersected the pons/medulla junction and the root of the nose. The contrast agent was administered in an arm vein at an injection rate of 5 ml/s, followed by a 40 ml saline flush. Injection was initiated 18 s after the start of the dynamic imaging sequence and the same side of injection was used in all three examinations. The mean contrast-agent injection volume was 7 ml for single-dose gadobutrol and 14 ml for double-dose gadobutrol and single-dose gadobenate-dimeglumine. 2.3. Theory and data analysis The haemodynamic parameters CBV, CBF and MTT were calculated according to standard DSC-MRI theory:

∞ Ctissue (t) dt CBV = kH  0∞ 0

CBF =

Cartery (t) dt

CBV CBV max[R(t)] = ∞ MTT R(t) dt 0

(1a)

(1b)

The constant kH = (1 − Hlarge )/[(1 − Hsmall )] = 0.705 ml/g was employed, where Hlarge and Hsmall are the haematocrit values in large and small vessels, respectively, and  is the brain-tissue density. Tracer concentrations C in tissue and artery were calculated using the relationship C(t) = −k·(1/TE)·ln[S(t)/S0 ], where S(t) is the signal at time t, S0 is the pre-contrast signal and k is a proportionality constant assumed to be equal for tissue and large vessels. R(t) is the tissue residue function and max[R(t)] is the peak value of this function. Deconvolution was performed using a block-circulant singular value decomposition algorithm. The employed global AIFs were averages (4–9 pixels) of arterial curves from middle cerebral artery (MCA) branches in the sylvianfissure region, automatically identified by the perfusion software according to standard criteria (early bolus arrival, reasonably narrow concentration curve, high peak concentration and no obvious distortion of the curve at peak concentration) [16]. All arterial curves retrieved by the software were visually inspected, and physiologically unrealistic curves were manually excluded from the AIF calculation, but the user interaction was kept to a minimum. CBV and CBF estimates, in pseudo-absolute terms, were obtained using two different AIF time-integral calculations: Firstly, the arterial time integral obtained directly from the automatically identified AIFs was employed. Secondly, intra-individual AIF time-integral variations (e.g., due to varying partial-volume effects)

e88

R. Wirestam et al. / European Journal of Radiology 75 (2010) e86–e91

Table 1 Mean values from the whole population of whole-brain average CBF, CBV and MTT, with and without AIF time-integral correction. Original AIF time integral

Gadovist® 1.0 (single dose) Gadovist® 1.0 (double dose) MultiHance® (single dose)

Rescaled AIF time integral

CBF [ml/(min 100 g)]

CBV [ml/100 g]

MTT [s]

CBF [ml/(min 100 g)]

CBV [ml/100 g]

171 ± 52 233 ± 67 158 ± 27

14 ± 2.9 20 ± 3.8 13 ± 2.2

5.4 ± 0.92 6.0 ± 1.0 5.5 ± 0.84

64 ± 22 88 ± 22 61 ± 21

5.1 ± 1.2 7.7 ± 1.7 4.9 ± 1.1

between different experiments were reduced using an AIF rescaling approach, introduced by Knutsson et al. [2], based on the calculation of a venous concentration–time curve area from the posterior superior sagittal sinus for rescaling of the original global AIF time integral (with retained global AIF shape). Large-vessel concentration time curves are typically distorted at peak concentration by, for example, signal saturation and/or local geometric distortions. Hence, before AIF rescaling, the venous concentration time integral was corrected by multiplying an undistorted smallvein concentration time curve with an appropriate correction factor so that the base and flanks of the small-vein concentration time curve corresponded to the base and flanks of the distorted concentration time curve from the posterior superior sagittal sinus. Mean estimates of CBV, CBF and MTT were calculated from all brain voxels in the parametric maps obtained from the 20 DSC-MRI slices. In the CBV and CBF averaging procedures, largevessel contributions were reduced by excluding all pixel values of the parametric maps that exceeded 2.5 times the median CBV or CBF value of the entire volume. Effects of contrast-administration scheme on the observed perfusion estimates were statistically

analyzed by an analysis of variance (ANOVA), supplemented by two-sided, paired t-tests. 3. Results Estimates from the entire population of whole-brain average CBF, CBV and MTT, with and without AIF time-integral correction, are given in Table 1. The statistical analyses (ANOVA) indicated that a significant effect of contrast-agent administration scheme was present for CBF (p < 0.000002), CBV (p < 0.000001) and MTT (p < 0.02), with as well as without AIF time-integral correction. Supplementing paired t-tests indicated that the doubledose gadobutrol experiment showed significantly higher CBF (p < 0.0001), CBV (p < 0.0001) and MTT (p < 0.05) values than the single-dose gadobutrol administration while the two single-dose experiments did not exhibit any significant differences with respect to the population averages (p > 0.25). Comparisons between different contrast-administration schemes, on an individual basis, were made by the use of Bland–Altman plots: CBF, CBV and MTT results obtained using uncorrected AIF time integrals are given in Fig. 1, while the CBF and

Fig. 1. Bland–Altman analyses of CBF, CBV and MTT results, obtained using original AIFs, from comparisons of three different contrast-agent administration schemes. Solid lines indicate mean difference between methods and dashed lines represent mean difference ± 1.96 SD. GVs = Gadovist® 1.0, single dose; GVd = Gadovist® 1.0, double dose; MH = MultiHance® , single dose.

R. Wirestam et al. / European Journal of Radiology 75 (2010) e86–e91

e89

Fig. 2. Bland–Altman analyses of AIF time-integral corrected CBF and CBV results from comparisons of three different contrast-agent administration schemes. Solid lines indicate mean difference between methods and dashed lines represent mean difference ± 1.96 SD. GVs = Gadovist® 1.0, single dose; GVd = Gadovist® 1.0, double dose; MH = MultiHance® , single dose. Table 2 Linear equations and correlation coefficients obtained by linear-regression analyses of original CBF, CBV and MTT data from comparisons of three different contrast-agent administration schemes. Parameter

Linear-regression equation and correlation coefficient MultiHance® single dose (y) versus Gadovist® 1.0 single dose (x)

MultiHance® single dose (y) versus Gadovist® 1.0 single dose (x)

Gadovist® 1.0 double dose (y) versus Gadovist® 1.0 single dose (x)

CBF [ml/(min 100 g)]

y = 0.29x + 109 r = 0.55

y = 0.26x + 96 r = 0.65

y = 0.87x + 84 r = 0.68

CBV [ml/100 g]

y = 0.22x + 10 r = 0.29

y = 0.020x + 13 r = 0.033

y = 0.82x + 9.2 r = 0.62

MTT [s]

y = 0.46x + 3.1 r = 0.50

y = 0.58x + 2.1 r = 0.70

y = 0.83x + 1.5 r = 0.75

Table 3 Linear equations and correlation coefficients obtained by linear-regression analyses of AIF time-integral corrected CBF and CBV data from comparisons of three different contrast-agent administration schemes. Parameter

Linear-regression equation and correlation coefficient MultiHance® single dose (y) versus Gadovist® 1.0 single dose (x)

MultiHance® single dose (y) versus Gadovist® 1.0 single dose (x)

Gadovist® 1.0 double dose (y) versus Gadovist® 1.0 single dose (x)

CBF [ml/(min 100 g)]

y = 0.54x + 26 r = 0.57

y = 0.49x + 18 r = 0.50

y = 0.84x + 34 r = 0.85

CBV [ml/100 g]

y = 0.54x + 2.2 r = 0.62

y = 0.34x + 2.3 r = 0.54

y = 1.0x + 2.6 r = 0.73

CBV plots obtained using corrected AIF time integrals are provided in Fig. 2. Figs. 1 and 2 show that the bias was most pronounced when double-dose and single-dose estimates are compared. Results from linear-regression analyses of the corresponding scatter plots (graphs not shown) are given in Tables 2 and 3. In these pair-wise comparisons, results from the two gadobutrolbased experiments consistently showed the highest degrees of correlation with each other, with linear equations typically closer to proportionality (i.e., showed smaller intercepts with the y-axis). For CBF and CBV data, the population averages were reduced and the coefficients of correlation improved after application of the AIF time-integral correction. However, the double-dose gadobutrol versus single-dose gadobutrol relationship maintained a stronger

correlation after AIF time-integral correction, as compared with the gadobenate-dimeglumine versus gadobutrol relationships. In summary, the perfusion estimates obtained using single and double dose of gadobutrol showed moderate to good degrees of correlation, while the relationships between estimates obtained using gadobutrol and single-dose gadobenate-dimeglumine tended to show somewhat lower correlation and, in most cases, larger deviation from proportionality. 4. Discussion The observed degree of CBV and CBF overestimation, without AIF time-integral correction, was not unexpected, and seemed to

e90

R. Wirestam et al. / European Journal of Radiology 75 (2010) e86–e91

be reasonably consistent with the relaxivity differences between artery and tissue at 3 T, predicted by Kjølby et al. [3], combined with some degree of partial-volume effects [7]. After AIF timeintegral correction, the CBV and CBF estimates were reduced but still somewhat overestimated. Based on reference CBF measurement techniques, the global average CBF in younger adults can be expected to amount to approximately 40–50 ml/(min 100 g) [2,17], compared with the present values of approximately 60–90 ml/(min 100 g). MTT is not expected to be directly influenced by the relaxivity differences between tissue and blood, although other complications, influencing the shape of the measured AIF (such as partial-volume effects and non-linear contrast-agent response), may potentially have a detrimental impact on MTT estimates. However, the MTT values observed in the present study were indeed quite reasonable. Reported literature values of CBV and CBF imply that MTTs in normal human brain tissue should typically range between approximately 4 and 7 s according to the central volume theorem [17,18]. The MTT values observed in the present study were around 5.5–6 s, i.e. well within the above interval, as can be expected for a mean value over the entire brain. In practice, the signal composition of typically selected AIFs in realistic DSC-MRI experiments is unknown, which makes it difficult to make any a priori assessment of the relevance of a potential correction approach. In the present study, absolute estimates of CBV, CBF and MTT were thus calculated without any explicit corrections for the non-linear arterial response to the contrast agent or for different relaxivities in tissue and artery. Without such corrections, a quadratic R2*-versus-concentration relationship [4], in combination with a high tissue T2* relaxivity as predicted by Kjølby et al. [3], implies that CBV (and CBF) estimates would exhibit a decrease with increasing contrast-agent dose. The present results showed the opposite trend, most likely due to partial-volume effects [7], i.e. the signal components of measured AIFs were not purely of arterial origin and the arterial signal components were more effectively extinguished at higher contrast-agent dose (i.e., signal data for a larger number of time points were influenced by arterial signal saturation at higher dose). This observation complicates attempts to correct for the non-linear contrast-agent response in arteries [4] and further questions the validity of typical AIFs employed in DSC-MRI (often selected on the basis of a reasonable appearance according to certain criteria as described above). Alger et al. [12] performed sequential DSC-MRI measurements (with TE > 60 ms) at 1.5 T using gadolinium contrast-agent doses of 0.1 and 0.2 mmol/kg b.w. delivered at a fixed injection rate of 5 ml/s. Among the observed results was that brain tissue displayed a less than doubled R2* time integral at double dose, partially explained by the authors as a consequence of tissue signal saturation at higher CA dose, while a doubling in the arterial R2* time integral was seen when the contrast-agent dose was doubled. Correspondingly, the CBV and CBF estimates were subtly lower at 0.2 mmol/kg b.w., except for the lowest values of CBV and CBF. These findings are not consistent with the present results, or with the previous evaluation of our data by Thilmann et al. [14], but the discrepancies may, in part, be explained by weaker susceptibility effects at 1.5 T leading to less pronounced arterial signal saturation at lower field strength. The significantly longer MTT observed using double dose of gadobutate indicates that the AIFs selected at higher dose must have shown a different shape from the single-dose experiments, but it is difficult to assess whether this was due to dose-dependent partial-volume effects (including the rapid decay of arterial signal components) [7] or on the non-linear R2*-versus-concentration relationship in blood (or on a combination of both these effects). Bleeker et al. [15] recently showed that GRE-based AIFs obtained from pixels outside large vessels may show the desired linear response to the contrast agent, provided that specific selection criteria are fulfilled and that the vessel orientation relative the

external magnetic field is appropriate. The present MTT results indicate, not unexpectedly, that conventional AIF-selection routines, in many cases, are unable to accommodate such detailed prerequisites for successful AIF selection. The observed degrees of correlation between the gadobutrolbased results obtained on different occasions (i.e., 0.73 < r < 0.85 with AIF time-integral correction) were not entirely discouraging, bearing in mind that volunteers had been repositioned and some days passed between measurements. However, considering the differing absolute levels of CBV and CBF at different gadobutrol doses, and the weak correlation between estimates obtained with gadobenate-dimeglumine and gadobutrol injections, this study also indicates that it is not completely straightforward to compare absolute or pseudo-absolute perfusion estimates obtained with different contrast-agent administration routines. The reasons for the lower coefficients of correlation in the gadobenatedimeglumine versus gadobutrol relationships were not obvious. A higher T1 relaxivity of the gadobenate-dimeglumine preparation [19] may have contributed to a larger influence from competing T1 effects on the gadobenate-dimeglumine-based data [20], but the T1-relaxivity differences between the employed contrast-agent preparations are not particularly prominent at 3 T [19] and the selected AIF voxels were most likely not entirely of arterial origin. Other factors of potential importance are different concentrationto-noise levels for concentration curves obtained at single and double dose [14], which may have influenced, for example, the performance of the deconvolution algorithm, and different injection volumes for the two single-dose injections. However, in spite of the different injection volumes, previous analyses established that there were no marked differences with respect to curve shape, bolus width, maximal tissue signal drop or maximal tissue concentration-to-noise ratio between the population-averaged concentration curves obtained using single-dose gadobenatedimeglumine and single-dose gadobutol [14]. 5. Conclusions Absolute or pseudo-absolute perfusion estimates obtained using different contrast-agent administration schemes may be difficult to compare. For example, results obtained with different types of contrast-agent preparations showed rather low correlation. The linear correlation between estimates obtained using single and double dose of gadobutrol was reasonably good, but double-dose CBF, CBV and MTT estimates showed significantly higher absolute values. Furthermore, this study confirms that signal components of typically selected AIFs in DSC-MRI are most like not of purely arterial origin and it further accentuates the difficulties in assessing the accuracy of seemingly valid AIFs. Considering the complexity in AIF appearance and origin, experimental observations are likely to serve as useful supplements to theoretical predictions and simulations. Acknowledgements This study was supported by Bayer AB, Bayer Schering Pharma Scandinavia, the Swedish Research Council (grant no. 13514), the German Academy of Sciences Leopoldina, the Crafoord Foundation and the Knut & Alice Wallenberg Foundation (grant no. 1998-0182). References [1] Marstrand JR, Rostrup E, Rosenbaum S, Garde E, Larsson HBW. Cerebral hemodynamic changes measured by gradient-echo or spin-echo bolus tracking and its correlation to changes in ICA blood flow measured by phase-mapping MRI. J Magn Reson Imaging 2001;14:391–400. [2] Knutsson L, Börjesson S, Larsson E-M, et al. Absolute quantification of cerebral blood flow in normal volunteers: correlation between Xe-133 SPECT

R. Wirestam et al. / European Journal of Radiology 75 (2010) e86–e91

[3] [4] [5]

[6]

[7]

[8]

[9]

[10]

[11]

and dynamic susceptibility contrast MRI. J Magn Reson Imaging 2007;26: 913–20. Kjølby BF, Østergaard L, Kiselev VG. Theoretical model of intravascular paramagnetic tracers effect on tissue relaxation. Magn Reson Med 2006;56:187–97. Calamante F, Connelly A, van Osch MJP. Nonlinear R2* effects in perfusion quantification using bolus-tracking MRI. Magn Reson Med 2009;61:486–92. Price SJ, Jena R, Green HA, et al. Early radiotherapy dose response and lack of hypersensitivity effect in normal brain tissue: a sequential dynamic susceptibility imaging study of cerebral perfusion. Clin Oncol 2007;19:577–87. van Osch MJ, Jansen PA, Vingerhoets RW, van der Grond J. Association between supine cerebral perfusion and symptomatic orthostatic hypotension. NeuroImage 2005;27:789–94. Kjølby BF, Mikkelsen IK, Pedersen M, Østergaard L, Kiselev VG. Analysis of partial volume effects on arterial input functions using gradient echo: a simulation study. Magn Reson Med 2009;61:1300–9. Levin JM, Kaufman MJ, Ross MH, et al. Sequential dynamic susceptibility contrast MR experiments in human brain: residual contrast agent effect, steady state, and hemodynamic perturbation. Magn Reson Med 1995;34:655–63. Griffiths PD, Pandya H, Wilkinson ID, Hoggard N. Sequential dynamic gadolinium magnetic resonance perfusion-weighted imaging: effects on transit time and cerebral blood volume measurements. Acta Radiol 2006;47:1079–84. Tombach B, Benner T, Reimer P, et al. Do highly concentrated gadolinium chelates improve MR brain perfusion imaging? Intraindividually controlled randomized crossover concentration comparison study of 0.5 versus 1. 0 mol/L gadobutrol. Radiology 2003;226:880–8. Manka C, Träber F, Gieseke J, Schild HH, Kuhl CK. Three-dimensional dynamic susceptibility-weighted perfusion MR imaging at 3.0 T: Feasibility and contrast agent dose. Radiology 2005;234:869–77.

e91

[12] Alger JR, Schaewe TJ, Lai TC, et al. Contrast agent dose effects in cerebral dynamic susceptibility contrast magnetic resonance perfusion imaging. J Magn Reson Imaging 2009;29:52–64. [13] Essig M, Lodemann KP, Le-Huu M, Brüning R, Kirchin M, Reith W. Intraindividual comparison of gadobenate dimeglubine and gadobutrol for cerebral magnetic resonance perfusion imaging at 1.5 T. Invest Radiol 2006;41:256–63. [14] Thilmann O, Larsson EM, Björkman-Burtscher IM, Ståhlberg F, Wirestam R. Comparison of contrast agents with high molarity and with weak protein binding in cerebral perfusion imaging at 3 T. J Magn Reson Imaging 2005;22:597–604. [15] Bleeker EJ, van Buchem MA, van Osch MJ. Optimal location for arterial input function measurements near the middle cerebral artery in first-pass perfusion MRI. J Cereb Blood Flow Metab 2009;29:840–52. [16] Thilmann O. LUPE: An extensible modular framework for evaluation of DSCacquired perfusion images. Proc. of the 21st Annual Meeting of the ESMRMB. Magn Reson Mater Phys 2004;16(electronic suppl. 1):537. [17] Leenders KL, Perani D, Lammertsma AA, et al. Cerebral blood flow, blood volume and oxygen utilization, normal values and effect of age. Brain 1990;113: 27–47. [18] Phelps ME, Huang SC, Hoffman EJ, Kuhl DE. Validation of tomographic measurement of cerebral blood volume with C-11-labeled carboxyhemoglobin. J Nucl Med 1979;20:328–34. [19] Rohrer M, Bauer H, Mintorovitch J, Requardt M, Weinmann H-J. Comparison of magnetic properties of MRI contrast media solutions at different magnetic field strengths. Invest Radiol 2005;40:715–24. [20] Calamante F, Vonken EJ, van Osch MJ. Contrast agent concentration measurements affecting quantification of bolus-tracking perfusion MRI. Magn Reson Med 2007;58:544–53.