Magnetic Resonance in Medicine 72:996–1006 (2014)

Dynamic Susceptibility Contrast MRI with a Prebolus Contrast Agent Administration Design for Improved Absolute Quantification of Perfusion Linda Knutsson,1* Emelie Lindgren,1 Andr e Ahlgren,1 Matthias J.P. van Osch,2 Karin Markenroth Bloch,3 Yulia Surova,4,5 Freddy Sta˚hlberg,1,6,7 Danielle van Westen,8 and Ronnie Wirestam1 INTRODUCTION

Purpose: Arterial partial-volume effects (PVEs) often hamper reproducible absolute quantification of cerebral blood flow (CBF) and cerebral blood volume (CBV) obtained by dynamic susceptibility contrast MRI (DSC-MRI). The aim of this study was to examine whether arterial PVEs in DSC-MRI data can be minimized by rescaling the arterial input function (AIF) using a sagittal-sinus venous output function obtained following a prebolus administration of a low dose of contrast agent. Methods: The study was carried out as a test–retest experiment in 20 healthy volunteers to examine the repeatability of the CBF and CBV estimates. All subjects were scanned twice with 7–20 days between investigations. Results: DSC-MRI returned an overestimated average wholebrain CBF of 220 6 44 mL/100 g/min (mean 6 SD) before correction and 44 6 15 mL/100 g/min when applying the prebolus design, averaged over all scans. Average whole-brain CBV was 20 6 2.0 mL/100 g before correction and 4.0 6 1.0 mL/ 100 g after prebolus correction. Conclusion: Quantitative estimates of CBF and CBV, obtained with the proposed prebolus DSC-MRI technique, approached those typically obtained by other perfusion modalities. The CBF and CBV estimates showed good repeatability. Magn C 2013 Wiley Periodicals, Reson Med 72:996–1006, 2014. V Inc. Key words: perfusion; cerebral blood flow; cerebral blood volume; dynamic susceptibility contrast MRI; prebolus

1

Department of Medical Radiation Physics, Lund University, Lund, Sweden. C.J. Gorter Center for high field MRI, Department of Radiology, LUMC, Leiden, Netherlands. 3 Clinical Science, Philips, Lund, Sweden. 4 Department of Clinical Sciences, Lund University, Lund, Sweden. 5 Department of Neurology, Ska˚ne University Hospital, Lund, Sweden. 6 Department of Diagnostic Radiology, Lund University, Lund, Sweden. 7 Lund University Bioimaging Center, Lund University, Lund, Sweden. 8 Center for Medical Imaging and Physiology, Ska˚ne University Hospital Lund, Lund, Sweden. Grant sponsor: Swedish Research Council; Grant numbers: 13514; 2007– 3974; 2010–4454; 2011–2971; Grant sponsor: The Crafoord Foundation; Grant sponsor: The Swedish Cancer Society; Grant number: 2012/597. *Correspondence to: Linda Knutsson, Ph.D., Department of Medical Radiation Physics, Lund University Hospital, SE-221 85, Lund, Sweden. E-mail: [email protected] Received 10 May 2013; revised 26 September 2013; accepted 27 September 2013 DOI 10.1002/mrm.25006 Published online 29 October 2013 in Wiley Online Library (wileyonlinelibrary. com). 2

C 2013 Wiley Periodicals, Inc. V

Reproducible absolute quantification of perfusion and other hemodynamic parameters is difficult to achieve by dynamic susceptibility contrast MRI (DSC-MRI), for example, due to arterial dispersion of the bolus occurring between the site of the recorded arterial input function (AIF) and the position of the true AIF (1), the concentration-dependent difference in T2* relaxivity between tissue and large vessels (2,3) and partial-volume effects (PVEs) affecting the registration of arterial concentration levels (4). In an attempt to remove or reduce PVEs, AIF rescaling using either the first-pass or the steady-state phase of arterial and venous concentration time curves in DSC-MRI has previously been proposed in the literature (5–8). The use of a corrected venous output function (VOF) has been shown to improve the linear correlation between CBF estimates obtained by DSCMRI and Xe-133 SPECT (5), DSC-MRI and dynamic computed tomography (CT) perfusion (9) as well as DSC-MRI and arterial spin labeling (ASL) (10). However, the obtained DSC-MRI CBF values from these studies were elevated compared with the reference methods, as well as with literature values obtained by gold-standard techniques such as positron emission tomography (PET) (11– 13). This may suggest that not all PVEs were successfully removed, alternatively that other effects than PVEs contribute significantly to the overestimation. A major problem when selecting a VOF for rescaling of the AIF is that the VOF obtained from the sagittal sinus is often distorted at high concentrations of paramagnetic contrast agent due to signal displacement related to the low bandwidth in the phase-encoding direction of single-shot echo planar imaging (EPI) (14) and by signal depletion at the long echo times (TEs) required to obtain adequate signal reduction in tissue (15). In the studies by Knutsson et al (5,10) and Ziegelitz et al (9), a small vein was selected to correct for the distortions in the shape of the VOF obtained from the sagittal sinus. Applying this method to DSC-MRI data requires an experienced user, because the selection of the small vein is somewhat challenging, and this step therefore constitutes a potential source of inaccuracy. Furthermore, the method is hampered by the possibility of an incorrect shape of the small-vein curve and/or of the flanks of the distorted sagittal-sinus curve. Such inaccuracies may, for example, originate from PVEs which can affect shape and/or amplitude, leading to an error in the reconstructed VOF time integral.

996

Prebolus DSC-MRI for Quantification of Perfusion

The purpose of this study was to correct for arterial PVEs by rescaling the AIF using a VOF obtained by injecting a fraction of the total contrast agent dose as a prebolus. During the prebolus passage, a segmented EPI sequence in single-slice mode was used to register the VOF in the sagittal sinus. Using segmented EPI, a higher bandwidth in the phase-encoding direction can be achieved to avoid large-vessel geometric distortion during the bolus passage, and the single-slice mode allows for improved temporal resolution. Furthermore, because the brain tissue is not the target of this part of the examination, a short TE can be used during the prebolus passage to avoid large-vessel signal saturation. Previous studies using prebolus techniques have demonstrated improvement of the AIF determination in the estimation of myocardial blood flow (16–18) and pulmonary blood flow (19,20). For brain, the prebolus technique has been used for retrieval of phase-based AIF in dynamic contrast-enhanced MRI (DCE-MRI) (21). In this study, the prebolus administration design, applied to DSC-MRI data, was implemented in a test–retest study of healthy subjects. The novel prebolus approach was compared with uncorrected data (i.e., without any attempt to remove PVEs) as well as with a previously published PVE correction method, described in the studies by Knutsson et al (5,10) and Ziegelitz et al (9). METHODS Subjects and MR Imaging Twenty healthy volunteers (10 males and 10 females) were examined twice (test–retest) on a 3T MR unit (Philips Achieva, Philips Medical Systems, Best, The Netherlands), with a time interval of 7–20 days between investigations. The volunteers were divided into two equally sized age groups, one with age range 25–34 years (7 males and 3 females) and one with age range 51–84 years (3 males and 7 females). The project was approved by the local ethics committee (The Regional Ethical Review Board in Lund), and written informed consent was obtained from each subject. All volunteers were examined by an experienced neurologist (Y.S.) and were asked to perform two tests, MMSE and Cognistat (22,23), to verify their cognitive ability. The medical history of the volunteers was also examined to exclude the possibility that eventual medication could have altered their blood flow. Morphological images (FLAIR, T2tse, and 3D T1 Look-Locker before and after contrast agent administration) were also evaluated to rule out the presence of any pathology. For proper planning of the DSC-MRI prebolus experiment (for retrieval of the VOF), a sagittal phase-contrast angiography T1-FFE scan was performed. The imaging parameters were as follows: Field of view (FOV) 300  300 mm2, image matrix 256  256, slice thickness 50 mm, in-plane spatial resolution 1.17  1.17 mm2, flip angle 15 , repetition time (TR) 20 ms, echo time (TE) 5.68 ms, and the number of averages was 2. To achieve similar planning for both experiments in the test–retest design, the planning from the first experiment was saved and thereafter used for the planning of the second experiment. The planning of the tissue of interest for

997

DSC-MRI was performed using the automatic planning tool SmartExam (24) to reduce user interaction to a minimum. A prebolus of contrast agent (Dotarem, Guerbet, Paris, France) was administered (0.02 mmol/kg b.w., injection rate 5 mL/s) and a segmented EPI was used to track the prebolus passage through the sagittal sinus in one single slice using a temporal resolution of 0.81 s. The imaging parameters were as follows: FOV 220  220 mm2, image matrix 128  128, slice thickness 5 mm, in-plane spatial resolution 1.72  1.72 mm2, EPI factor 7, flip angle 22 , TR 135 ms, TE 15 ms and acquisition time 1:38 min. After the prebolus scan, a normal single dose of the contrast agent was administered (0.1 mmol/kg) at an injection rate of 5 mL/s, and the actual DSC-MRI experiment was performed using a single-shot gradient-echo EPI (GRE-EPI) with the following imaging parameters: FOV 220  220 mm2, image matrix 128  128, slice thickness 5 mm, in-plane spatial resolution 1.72  1.72 mm2, 20 slices, flip angle 60 , TR 1243 ms, TE 29 ms, 20 slices and acquisition time 1:30 min. Both bolus injections were followed by a 20-mL saline flush injected at a rate of 5 mL/s. Postprocessing For the prebolus as well as the standard DSC-MRI measurement, the signal was converted to concentration using a linear relationship between the transverse relaxation-rate change DR*2(t) and the contrast agent concentration (25). Hence, tissue as well as arterial contrast agent concentrations C, can be obtained from the relationship C(t) ¼ k  (1/TE)  ln[S(t)/S0], where S(t) is the signal at time t, S0 is the baseline signal and k is a constant reflecting the transverse relaxivity. In this study k was arbitrarily set to 1 because the relaxivities of blood and tissue were assumed to be equal and a linear relationship was assumed between DR*2(t) and C(t). The concentration time course in an appropriate brainfeeding artery, i.e., the arterial input function, is denoted AIF(t) below. In all subjects, CBF was calculated using Zierler’s area-to-height relationship and the central volume principle (26):

 CBF ¼

 1  Hlarge Rmax

Z1 C ðt Þdt [1]

0

Z1 Z1 rð1  Hsmall Þ RðtÞdt AIF ðtÞdt 0

0

By deconvolving the measured tissue concentration time curve C(t) with an AIF, the tissue impulse response function R(t) can be obtained, and Rmax is the peak value of this function. Hlarge and Hsmall are the hematocrit values in large and small vessels, respectively, and r is the whole-brain mass density and the numerical value (1Hlarge)/[r (1-Hsmall)] ¼ 0.705 cm3/g was used in this study (27). A block-circulant singular value decomposition algorithm (28) with a fixed cutoff of 10% (i.e., singular

998

Knutsson et al.

values smaller than 10% of the maximal singular value were set to zero) was used for deconvolution, and a global AIF was obtained from middle cerebral artery branches in the Sylvian fissure region. Because the mean contrast agent concentration in voxels containing such small vessels is low compared with the sagittal sinus voxels, the risk for signal displacement when using lowbandwidth sequences (such as single-shot EPI) is reduced. On the other hand, small-vessel AIFs are likely to suffer from PVEs, but these effects were assumed to be reduced or removed according to the specific approaches described below. Arterial concentration time curves, to be considered for inclusion in the global AIF, were identified semi-automatically according to certain criteria that resulted in nondistorted peaks with an early rise and a high maximum concentration (29). After visual inspection, arterial concentration curves from typically 4–8 pixels were selected and averaged into one single global AIF. CBV was calculated in all subjects using the following equation (26): 

1  Hlarge

 R1

CBV ¼ rð1  Hsmall Þ

0 R1

FIG. 1. Example of an AIF and a prebolus VOF (before rescaling) from one healthy subject.

C ðt Þdt [2]

AIF ðt Þdt

0

In the prebolus single-slice dataset, a manually placed region of interest (ROI), consisting of 2–3 voxels, was selected in the sagittal sinus to obtain the VOF with maximal time integral (area under curve), without any signs of significant distortion at peak concentration. The prebolus VOF area was then multiplied by five (i.e., 0.1/ 0.02) to obtain the same concentration time integral as would have been observed for a DSC-MRI experiment executed at normal dose. For comparison, a previously described VOF rescaling method (5,9,10), here dubbed “main bolus correction,” was also applied to the data from the standard DSC-MRI measurement. In short, this method uses a distortioncorrected concentration–time curve from the posterior superior sagittal sinus as a VOF. The corrected sagittal sinus concentration–time curve is obtained by selecting an additional concentration–time curve from a small vein and then multiplying this curve by an amplification factor and shifting it in time (if necessary) to match the base and flanks of the distorted sagittal sinus concentration–time curve.

done using the New Segment routine in SPM8 (www.fil.ion.ucl.ac.uk/spm) with default settings (30). Images from the initial time point in the DSC-MRI time series were used for segmentation. GM and WM CBF values were calculated as the mean value of CBF in each segmented ROI. Average whole-brain CBF was based on the voxels remaining after exclusion of voxels with CBF values exceeding 2.5 times the mean CBF over the entire volume, because those voxel values were assumed to originate from large vessels. CBV values were calculated in the corresponding manner. Finally, corrected CBF and CBV values were calculated by multiplying the noncorrected CBF and CBV estimates by the correction factor (CF). Statistical analyses of the test–retest data were accomplished using the Student t-test for two paired samples (merely to rule out the existence of any bias), the intraclass correlation coefficient (ICC) (including 95% confidence intervals) and the Bland-Altman method returning 95% confidence intervals (CI[95%]) of the mean difference, relative repeatability coefficients (RCrel) calculated as 1.96SD/mean and absolute repeatability coefficients (RC) calculated as 1.96SD (31).

RESULTS Data Evaluation To obtain subject-specific correction factors (CFs) in the prebolus approach, the first-pass area under curve of the AIF was divided by the first-pass area of the prebolusbased VOF. The end of the first pass was chosen manually as the point in time at which the minimum concentration occurred after the first pass, but before the recirculation phase. For the main bolus method, CFs was obtained by dividing the area of the AIF with the area of the VOF obtained from the corrected sagittal sinus concentration–time curve (similar as in 5,9,10). Segmentation of gray matter (GM) and white matter (WM) was

Figure 1 shows an example of a prebolus VOF curve (before rescaling) and an AIF from one subject. Scatter plots comparing VOF and AIF time integrals from measurement 1 and measurement 2 are displayed in Figure 2. The ICC was 0.52 for the AIF test–retest analysis, 0.44 for the main bolus VOF test–retest analysis and 0.77 for the prebolus VOF test–retest analysis. Prebolus corrected CBF maps for a representative subject are shown in Figure 3 (top). To visualize the general spatial perfusion patterns, all CBF maps were nonlinearly coregistered (spatially normalized) to the MNI152 template brain (ICBM, NIH P-20 project) using SPM8 (30). The group

Prebolus DSC-MRI for Quantification of Perfusion

FIG. 2. Scatter plots showing the AIF and VOF areas from measurement 1 and measurement 2.

mean corrected CBF maps for DSC-MRI are displayed in Figure 2 (bottom). Average CBF and CBV estimates for whole brain, GM and WM, for the noncorrected, main bolus corrected and prebolus corrected DSC-MRI experiments, are listed in Table 1. ICCs with 95% confidence intervals for the test– retest analysis are shown in Tables 2 and 3. Note that the ICCs were generally higher for the prebolus DSC-MRI experiment, indicating that the two measurements resemble each other better. The confidence interval was also smaller for the prebolus corrected DSC-MRI data.

999

Relative and absolute RCs are also shown in Tables 2 and 3. The absolute RCs were smaller for the prebolus corrected experiment, but this is, obviously, related to the fact that the prebolus approach results in lower absolute perfusion estimates. Relative RCs were similar for different methods, except for the noncorrected CBV values which showed lower relative RCs. On the other hand, the ICC confidence interval for noncorrected whole-brain CBV showed a very large range (including a negative lower limit). The Student t-test P-values from the test–retest analysis are listed in Tables 2 and 3. Although none of the paired groups showed significant differences, it can still be observed that P-values corresponding to the noncorrected CBV measurement were noticeably lower than the corresponding P-values for the two corrected CBV experiments. Scatter plots and Bland-Altman plots for noncorrected, main bolus corrected and prebolus corrected whole-brain CBF estimates obtained from the two measurements (n ¼ 20) are displayed in Figures 4a–f. The mean CBF differences between the two measurements (test versus retest) were as follows: 7 6 38 mL/100 g/min (CI[95%] ¼ [68;82]) for noncorrected data, 0.5 6 17 mL/100 g/min (CI[95%] ¼ [34;35]) for the main bolus correction and 2.8 6 8.4 mL/100 g/min (CI[95%] ¼ [19;14]) for the prebolus correction. Similarly, mean whole-brain CBV (displayed in Figs. 5a– f) showed a difference of 1.1 6 2.9 mL/100 g (CI[95%] ¼ [4.5;6.7]) for the noncorrected data, 0.3 6 1.8 mL/100 g (CI[95%] ¼ [3.2;3.9]) for the main bolus correction and 0.14 6 0.97 mL/100 g (CI[95%] ¼ [2.0;1.8]) for the prebolus correction. From the Bland-Altman analyses on GM and WM data, the mean CBF difference between the two prebolus corrected measurements (i.e., test versus retest) was 3.0 6 11 mL/100 g/min (CI[95%] ¼ [24;18]) in GM and 0.9 6 4.6 mL/100 g/min (CI[95%] ¼ [10;8]) in WM. For corrected CBV, the mean difference was 0.14 6 1.2 mL/ 100 g (CI[95%] ¼ [2.2;2.5]) in GM and 0.03 6 0.5 mL/ 100 g (CI[95%] ¼ [1.0;0.99]) in WM. Looking at the two age groups, the elderly showed a mean prebolus-corrected GM CBF of 55 6 19 mL/100 g/ min while the younger group showed a mean GM CBF of 55 6 20 mL/100 g/min. The corresponding main bolus corrected CBF values were 107 6 23 mL/100 g/min and 98629 mL/100 g/min for elderly and younger, respectively, and the noncorrected CBF values where 287 6 56 mL/100 g/min and 270656 mL/100 g/min. Regarding the two gender groups, the females showed a mean GM CBF of 59 6 18 mL/100 g/min while the males showed a mean GM CBF of 51 6 20 mL/100 g/min using the prebolus method. The corresponding main bolus CBF values were 111 6 25 mL/100 g/min and 93 6 24 mL/100 g/min and the noncorrected CBF values where 302 6 59 mL/100 g/min and 255 6 42 mL/100 g/ min, for females and males, respectively. Finally, the mean prebolus correction factor was 0.20 with a SD of 0.06, when including all 40 experiments. The first measurement resulted in a correction factor of 0.2160.06 (n ¼ 20) while the second measurement showed a correction factor of 0.19 6 0.05 (n ¼ 20).

1000

Knutsson et al.

FIG. 3. CBF maps obtained using corrected DSC-MRI (a) in one healthy subject and (b) on group level, by registration to MNI space.

DISCUSSION Absolute quantification of perfusion is warranted when a global reduction in CBF is expected, for example, in patients with cardiac disease, dementia or depressive disorders. DSC-MRI also provides supplementary hemodynamic parameters, such as CBV and mean transit time (MTT), which can be used for diagnosis of certain pathological conditions. In acute ischemic stroke, CBF and CBV maps may be important for tissue outcome and for distinction between infarct core and the ischemic penumbra zone (32,33). The Guidelines and Recommendations for Perfusion Imaging in Cerebral Ischemia states that absolute values of perfusion may be helpful as an aid in determining the risks and benefits of revascularization, including postthrombolysis hemorrhage (34). Therefore, it is encouraging that the prebolus administration design for deriving subject-specific correction factors appears to be highly promising, considering that the quantitative CBF and CBV estimates obtained with the proposed technique approached those typically obtained by gold-standard techniques such as PET and Xe-CT. The mean correction factor of 0.20 leads to the assumption that our proposed prebolus correction method removes a substantial amount of arterial PVEs, and this is also indicated by the obtained corrected mean GM CBF of 55 mL/100 g/min. This value is close to CBF levels obtained by previous MRI (7,35–38), CT, and PET studies (11–13,39). The corrected CBV values of 5.0 mL/100 g in GM and 2.0 mL/100 g in WM are comparable to PET-based values of 5.0 mL/100 g in insular grey matter and 2.6 mL/100 g in WM according to Leenders et al (11). Separation of the investigated study population into sub-groups did not allow for any firm statistical conclusions to be drawn about age-related or gender-related CBF differences, and the separation of results into different subgroups was only meant to serve as informal indications of the reliability of the absolute values. The tendency of a gender difference, with higher CBF in females, is, for example, not inconsistent with previous studies.

With regard to the t-test, applied to rule out any systematic difference between measurement 1 and 2, the Pvalues were higher than 0.05, indicating that a null hypothesis at this significance level cannot be rejected, i.e., data from measurement 1 and 2 were, as expected, not significantly different. However, for CBV, the P-values observed for the noncorrected data (around 0.1) were indeed noticeably lower than for both the corrected datasets. The different measures of repeatability were not entirely conclusive, but several parameters indicated that the prebolus corrected estimates may be somewhat more robust, even for repeated measurements on the same scanner using the same protocol and postprocessing technique, carried out by the same operator. For example, the Pearson correlation coefficients for noncorrected test–retest CBF data were r ¼ 0.68 (whole brain), r ¼ 0.72 (GM), and r ¼ 0.76 (WM), for main bolus corrected test– retest CBF data r ¼ 0.65 (whole brain), r ¼ 0.72 (GM), and r ¼ 0.70 (WM) and for prebolus corrected test–retest CBF data they were r ¼ 0.87 (whole brain), r ¼ 0.87 (GM), and r ¼ 0.84 (WM). Similarly, for noncorrected CBV data, the corresponding correlation coefficients were r ¼ 0.31 (whole brain), r ¼ 0.61 (GM), and r ¼ 0.53 (WM), for main bolus corrected test–retest CBV data they were r ¼ 0.55 (whole brain), r ¼ 0.69 (GM), and r ¼ 0.66 (WM) and, finally, for prebolus corrected CBV they were r ¼ 0.64 (whole brain), r ¼ 0.67 (GM) and r ¼ 0.65 (WM). Although correlation analysis should be treated with caution in this context, the above results were included to serve as a complementary indication of repeatability. Hence, the ICC values, including the confidence intervals, as well as the correlation coefficients suggested that the prebolus correction method shows the overall best repeatability. With regard to the repeatability coefficients, the absolute RCs were lower for the prebolus method (obviously related to the lower absolute levels of the prebolus perfusion estimates), and the relative RCs were similar, with the exception of CBV without correction, for which the relative RCs were lower. However, one should also note

Prebolus DSC-MRI for Quantification of Perfusion

1001

Table 1 Average CBF and CBV Estimates for Whole Brain, GM and WM, for Noncorrected and Corrected (Main bolus and Prebolus) DSC-MRI Experiments Noncorrected CBF [mL/100 g/min] Measurement 1 (mean 6 SD; n ¼ 20) Measurement 2 (mean 6 SD; n ¼ 20) All measurements (mean (n ¼ 40) 6 SD(n ¼ 20))

Whole brain

Gray matter

White matter

217 6 44 224 6 51 220 6 44

274 6 56 284 6 69 279 6 58

107 6 24 113 6 31 110 6 26

Main bolus corrected CBF [mL/100 g/min] Whole brain

Gray matter

White matter

80 6 22 81 6 21 81 6 20

102 6 29 103 6 27 102 6 26

40 6 11 41 6 12 40 6 11

Measurement 1 (mean 6 SD; n ¼ 20) Measurement 2 (mean6 SD; n ¼ 20) All measurements (mean (n ¼ 40) 6 SD(n ¼ 20))

Prebolus corrected CBF [mL/100 g/min] Whole brain

Gray matter

White matter

45 6 17 42 6 14 44 6 15

57 6 21 54 6 19 55 6 19

22 6 9 21 6 8 22 6 8

Measurement 1 (mean 6 SD; n ¼ 20) Measurement 2 (mean6 SD; n ¼ 20) All measurements (mean (n ¼ 40) 6 SD(n ¼ 20))

Noncorrected CBV [mL/ 100g] Measurement 1 (mean 6 SD; n ¼ 20) Measurement 2 (mean 6 SD; n ¼ 20) All measurements (mean (n ¼ 40) 6 SD(n ¼ 20))

Whole brain

Gray matter

White matter

20 6 2.2 21 6 2.7 20 6 2.0

24 6 3.3 26 6 4.1 25 6 3.2

10 6 1.6 11 6 1.9 10 6 1.5

Main bolus corrected CBV [mL/100g] Measurement 1 (mean 6 SD; n ¼ 20) Measurement 2 (mean 6 SD; n ¼ 20) All measurements (mean (n ¼ 40) 6 SD(n ¼ 20))

Whole brain

Gray matter

White matter

7.4 6 1.6 7.7 6 2.1 7.6 6 1.7

9.1 6 2.4 9.4 6 2.7 9.3 6 2.3

3.7 6 0.9 3.9 6 1.1 3.8 6 0.9

Prebolus corrected CBV [mL/100g] Measurement 1 (mean 6 SD; n ¼ 20) Measurement 2 (mean 6 SD; n ¼ 20) All measurements (mean (n ¼ 40) 6 SD(n ¼ 20))

Whole brain

Gray matter

White matter

4.1 6 1.2 4.0 6 1.1 4.0 6 1.0

5.0 6 1.5 4.9 6 1.5 5.0 6 1.3

2.0 6 0.6 2.0 6 0.6 2.0 6 0.6

that for uncorrected CBV, the P-values and correlation coefficients were lower and the ICC 95% confidence interval was very large (e.g., [0.15;0.65] for whole-brain CBV). It may seem somewhat surprising that even without AIF rescaling, the CBF and CBV estimates showed good to fair repeatability. The reason for this is most likely that the AIF selection procedure was kept highly standardized across all subjects, and it was carried out by the same user with equal slice positions on both occasions (assured by the use of SmartExam). One should note that in the case of different experimental setups and operators, noncorrected DSC-MRI estimates would most likely differ more and would not be comparable. Different AIF selection approaches exist within the DSC-MRI community, including registration from the Sylvian fissure region, from the middle cerebral artery and the use of local AIFs. Furthermore, these methods can be manual or automatic and based on various degrees of averaging. All these different procedures for AIF selection are likely to return AIFs with different PVEs. Hence, the use of a prebolus administration design for obtaining quantitative

values can be crucial, for example, in longitudinal studies, when comparing different patient groups or in studies carried out at different sites. Because segmented EPI was used in the measurement of the slice containing the VOF, there was a certain risk of competing T1 effects due to the short repetition time. When designing the protocol, the T1 effects were considered to be limited due to the fresh inflow of blood in the vein. Furthermore, the mean transit time in the brain is around 4–6 s, and the blood that was excited in the arteries within the prebolus slice perpendicular to the sagittal sinus was thus assumed to be fully relaxed when it reached this large vein. However, in some pixels close to the sagittal sinus, a slight signal increase at maximal contrast agent concentration was seen, indicating that a T1 effect could have been present. If a T1 effect was indeed present, the calibration factors would be overestimated due to a smaller VOF area and, consequently, the obtained absolute values of corrected CBF and CBV would also be higher than the true values. In some subjects, distorted VOF peaks were present, suggesting either that the subjects had moved during the

1002

Knutsson et al.

Table 2 ICC with 95% Confidence Interval, RC in Absolute and Relative Values, P Values Obtained from Student’s t-Test for Whole Brain, GM and WM, for Noncorrected and Corrected (Main bolus and Prebolus) DSC-MRI CBF Measurements CBF

ICC

RC [mL/min/100g]

RCrel [%]

p-values

WB GM WM WB GM WM WB GM WM WB GM WM

Noncorrected

Main bolus corrected

Prebolus corrected

0.68 CI ¼ [0.34;0.86] 0.70 CI ¼ [0.39;0.87] 0.76 CI ¼ [0.49;0.90] 75 95 37 34 34 34 0.42 0.34 0.15

0.66 CI ¼ [0.31;0.85] 0.71 CI ¼ [0.41;0.88] 0.70 CI ¼ [0.38;0.87] 35 42 18 43 41 45 0.91 0.86 0.50

0.85 CI ¼ [0.66;0.94] 0.86 CI ¼ [0.68;0.94] 0.84 CI ¼ [0.64;0.93] 16 21 9 38 37 41 0.15 0.23 0.38

scanning or that partial volume effects influenced the shape of the VOF. However, in several subjects, the VOF curve could also be obtained from the internal jugular veins (not used for correction factors in this study). When comparing those VOF curves with the ones obtained from the sagittal sinus, they were very similar both in shape and time integral. The manual step in choosing the first-pass area may be crucial and the influence of selecting a different time point for both the AIF and the prebolus VOF was investigated (detailed data not shown). When the area was expanded to include one more time point of the concentration–time curves, the change in CBF was minor (ranging from 0 to 3%). The retrieval of the VOF was also quite robust from measurement 1 to measurement 2, compared with the corresponding AIF test–retest data (cf. Fig. 1 and the ICCs). Application of a Pearson correlation analysis to the test– retest data resulted in a correlation coefficient of 0.52 for the AIF test–retest data, 0.44 for the main bolus VOF test–retest data and 0.78 for the prebolus VOF test–retest data. In a perfect setting, a change in true AIF would correspond to a similar change in true VOF. However, a larger difference was observed between the AIF test– retest measurements. Although physiological and

planning-related differences can influence the result from the two experiments, these findings together with the intra-individual similarities to the VOFs observed in the internal jugular veins indicate that the use of prebolus VOFs is generally a robust approach. The ROIs for CBF and CBV estimation in GM and WM used in this study were based on segmented tissue. This might, to some extent, influence the results because segmentation is not optimal in all aspects. For example, a small fraction of CSF might be included in the ROIs which could have influenced the mean values of CBF and CBV in these areas. However, this effect is not expected to be of considerable importance because the segmented ROIs included a large number of voxels. In the literature, a nonlinear relationship between DR2* and contrast agent concentration in whole blood has been reported (40). When applying the nonlinear model to AIF and VOF data, the transverse relaxivity and relaxivity-related constants for tissue and blood must be set by the user, and the absolute values will thus depend on the numerical values of these constants. A transverse relaxivity of 87 mM1s1 in tissue at 3 Tesla (T) has been extracted from a theoretical model by Kjïlby et al (41) and the use of this value in combination

Table 3 ICC with 95% Confidence Interval, RC in Absolute and Relative Values, P Values Obtained from Student’s t-Test for Whole Brain, GM and WM, for Noncorrected and Corrected (Main Bolus and Prebolus) DSC-MRI CBV Measurements CBV

ICC

RC [mL/100g]

RCrel [%]

p-values

WB GM WM WB GM WM WB GM WM WB GM WM

Noncorrected

Main bolus corrected

Prebolus corrected

0.30 CI ¼ [0.15;0.65] 0.52 CI ¼ [0.11;0.78] 0.60 CI ¼ [0.23;0.82] 5.6 7.1 3.0 27 27 29 0.10 0.12 0.08

0.54 CI ¼ [0.13;0.79] 0.69 CI ¼ [0.36;0.86] 0.64 CI ¼ [0.29;0.84] 3.5 4.0 1.7 47 43 45 0.42 0.49 0.32

0.64 CI ¼ [0.29;0.84] 0.67 CI ¼ [0.34;0.86] 0.65 CI ¼ [0.31;0.85] 1.9 2.3 1.0 46 47 49 0.56 0.61 0.82

Prebolus DSC-MRI for Quantification of Perfusion

1003

FIG. 4. Scatter plot (a–c) and Bland-Altman plot (d–f) of whole-brain CBF estimates obtained from the two noncorrected, main bolus corrected and prebolus corrected DSC-MRI measurements (n ¼ 20). In (a), the black solid line corresponds to the identity line, and in (b) the blue line, dashed blue line and dashed red line correspond to the mean difference (bias), 95% confidence interval for the difference of the means and 1.96 times the standard deviation of the differences (limits of agreement), respectively.

with the nonlinear model resulted in underestimated and unrealistic absolute levels when applied to the data obtained in the present study . The tissue T2* relaxivity

by Kjïlby et al could very well be too high, because it was assumed that the magnetic field disturbance of each vessel does not interact with magnetic field disturbances

1004

Knutsson et al.

FIG. 5. Scatter plot (a–c) and Bland-Altman plot (d–f) of the whole-brain CBV estimates obtained from the two noncorrected, main bolus corrected and prebolus corrected DSC-MRI measurements (n ¼ 20). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

of neighboring vessels. This assumption is probably not true for the maximum concentration of contrast agent in grey matter, as apparent from the large signal drop (>50%) predicted by data from Kjïlby et al The difficulty in assessing the accurate T2* relaxivity in tissue was acknowledged by Calamante et al, (42) who applied

and compared two significantly different transverse tissue relaxivities, 44 mM1s1 (given by Kjïlby et al for 1.5 T) and 5.9 mM1s1, when studying the effects of a nonlinear DR2* relationship at 1.5T. It was in this context emphasized that the transverse relaxivity value in tissue is a scaling factor and thus will not influence

Prebolus DSC-MRI for Quantification of Perfusion

repeatability. In another study, comparing absolute CBF values from DSC-MRI and Xe-133 SPECT, using both the linear and nonlinear relationship in blood, no significant difference in the correlation between SPECT-based and MRI-based CBF was seen when applying the two different DR2*-versus-concentration relationships (43). Although this result could be explained by AIF signal data originating from a combination of blood and surrounding tissue, the findings still indicate that it is not straightforward to determine whether or not the nonlinear DR2*-versus-concentration relationship is optimal in practical DSC-MRI implementations. The quite reasonable quantitative values, in absolute terms, obtained using the prebolus concept in DSC-MRI, together with the repeatability of the method, is encouraging, for example, because these qualities enable the method to be used in pathological conditions where a global reduction of perfusion is expected or when acute ischemic tissue viability CBF threshold values are warranted. Future prebolus DSC-MRI studies involves comparison of corrected values to a “gold-standard” perfusion technique such as PET, and other PVE correction single acquisition methods. Furthermore, application to a patient population with neurodegenerative disease and a comparison of the quantitative perfusion results of this patient group with the results from the healthy volunteers in this study would be of considerable interest. CONCLUSIONS By applying the correction scheme given by the proposed prebolus concept, absolute quantitative CBF and CBV values obtained by DSC-MRI approached those typically obtained by other perfusion techniques such as ASL, PET, and Xe-CT. The estimates from the DSC-MRI method also showed good repeatability. This leads us to expect that the prebolus DSC-MRI concept can be a powerful tool if repeatable absolute quantitative perfusion levels are warranted. REFERENCES 1. Calamante F. Bolus dispersion issues related to the quantification of perfusion MRI data. J Magn Reson Imaging 2005;22:718–722. 2. Kjïlby BF, Østergaard L, Kiselev VG. Theoretical model of intravascular paramagnetic tracers effect on tissue relaxation. Magn Reson Med 2006;56:187–197. 3. Kiselev VG. On the theoretical basis of perfusion measurements by dynamic susceptibility contrast MRI. Magn Reson Med 2001;46:1113– 1122. 4. Thijs VN, Somford DM, Bammer R, Robberecht W, Moseley ME, Albers GW. Influence of arterial input function on hypoperfusion volumes measured with perfusion-weighted imaging. Stroke 2004;35:94– 98. 5. Knutsson L, B€ orjesson S, Larsson EM, et al. Absolute quantification of cerebral blood flow in normal volunteers: correlation between Xe133-SPECT and dynamic susceptibility contrast MRI. J Magn Reson Imaging 2007;26:913–920. 6. Zaharchuk G, Bammer R, Straka M, et al. Improving dynamic susceptibility contrast MRI measurement of quantitative cerebral blood flow using corrections for partial volume and nonlinear contrast relaxivity: a xenon computed tomographic comparative study. J Magn Reson Imaging 2009;30:743–752. 7. Bjïrnerud A, Emblem KE. A fully automated method for quantitative cerebral hemodynamic analysis using DSC-MRI. J Cereb Blood Flow Metab 2010;30:1066–1078.

1005 8. Crane DE, Donahue MJ, Chappell MA, Sideso E, Handa A, Kennedy J, Jezzard P, MacIntosh BJ. Evaluating quantitative approaches to dynamic susceptibility contrast MRI among carotid endarterectomy patients. J Magn Reson Imaging 2013;37:936–943. 9. Ziegelitz D, Starck G, Mikkelsen IK, Tullberg M, Edsbagge M, Wikkesl€ o C, Forsell-Aronson E, Holta˚s S, Knutsson L. Absolute quantification of cerebral blood flow in normal volunteers: dynamic susceptibility contrast MRI compared with CT-perfusion. Magn Reson Med 2009;62:56–65. 10. Knutsson L, van Westen D, Petersen ET, Bloch KM, Holta˚s S, Sta˚hlberg F, Wirestam R. Absolute quantification of cerebral blood flow using dynamic susceptibility contrast MRI: correlation with model-free arterial spin labeling. Magn Reson Imaging 2010;28:1–7. 11. Leenders KL, Perani D, Lammerstma AA, et al. Cerebral blood flow,blood volume and oxygen utilization: normal values and effect of age. Brain 1990;113:27–47. 12. Ito H, Kanno IH, Kato C, et al. Database of normal cerebral bloodflow, cerebral blood volume, cerebral oxygen extraction fractionand cerebral metabolic rate of oxygen measured by position emission tomography with 15O-labelled dioxide or water, carbon monoxide and oxygen: a multicentre study in Japan. Eur J Nucl MedMol Imaging 2004;31:635–643. 13. Rostrup E, Law I, Pott F, Ide K, Knudsen GM. Cerebral hemodynamics measured with simultaneous PET and near-infrared spectroscopy in humans. Brain Res 2002;954:183–193. 14. Rausch M, Scheffler K, Rudin M, Radu EW. Analysis of input functions from different arterial branches with gamma variate functions and cluster analysis for quantitative blood volume measurements. Magn Reson Imaging 2000;18:1235–1243. 15. Ellinger R, Kremser C, Schocke MF, et al. The impact of peak saturation of the arterial input function on quantitative evaluation of dynamic susceptibility contrast-enhanced MR studies. J Comput Assist Tomogr 2000;24:942–948. 16. K€ ostler H, Ritter C, Lipp M, Beer M, Hahn D, Sandstede J. Prebolus quantitative MR heart perfusion imaging. Magn Reson Med 2004;52: 296–299. 17. Utz W, Greiser A, Niendorf T, Dietz R, Schulz-Menger J. Single- or dual-bolus approach for the assessment of myocardial perfusion reserve in quantitative MR perfusion imaging. Magn Reson Med 2008;59:1373–1377. 18. Christian TF, Rettmann DW, Aletras AH, Liao SL, Taylor JL, Balaban RS, Arai AE. Absolute myocardial perfusion in canines measured by using dual-bolus first-pass MR imaging. Radiology 2004;232:677–684. 19. Oechsner M, M€ uhlh€ ausler M, Ritter CO, Weininger M, Beissert M, Jakob PM, Beer M, Hahn D, K€ ostler H. Quantitative contrastenhanced perfusion measurements of the human lung using the prebolus approach. J Magn Reson Imaging 2009;30:104–111. 20. Risse F, Semmler W, Kauczor HU, Fink C. Dual-bolus approach to quantitative measurement of pulmonary perfusion by contrast enhanced MRI. J Magn Reson Imaging 2006;24:1284–1290. 21. Garpebring A, Karlsson M, Wirestam R, Yu J, Asklund T. Phase-based arterial input functions in humans applied to dynamic contrastenhanced MRI: potential usefulness and limitations. MAGMA 2011; 24:233–245. 22. Folstein MF, Folstein SE, McHugh PR. "Mini-mental state". A practical method 2 grading the cognitive state of patients for the clinician". J Psychiatr Res 1975;12:189–198. 23. Kiernan RJ, Mueller J, Langston JW, Van Dyke C. The Neurobehavioral Cognitive Status Examination: a brief but differentiated approach to cognitive assessment. Ann Intern Med 1987;107:481–485. 24. Young S, Bystrov D, Netsch T, Bergmans R, van Muiswinkel A, Visser F, Springorum R, Hoogenraad F, Gieseke J. Automated planning of MRI neuro scans. In Proceedings of the SPIE 6144 on Medical Imaging (Image Processing), San Diego, California, USA, 2006. 25. Fisel CR, Ackerman JL, Buxton RB, Garrido L, Belliveau JW, Rosen BR, Brady TJ. MR contrast due to microscopically heterogeneous magnetic susceptibility: numerical simulations and applications to cerebral physiology. Magn Reson Med 1991;17:336–347. 26. Knutsson L, Sta˚hlberg F, Wirestam R. Absolute quantification of perfusion using dynamic susceptibility contrast MRI: pitfalls and possibilities. MAGMA 2010;23:1–21. 27. Rempp KA, Brix G, Wenz F, Becker CR, G€ uckel F, Lorenz WJ. Quantification of regional cerebral blood flow and volume with dynamic susceptibility contrast-enhanced MR imaging. Radiology 1994;193: 637–641.

1006 28. Wu O, Østergaard L, Weisskoff RM, Benner T, Rosen BR, Sorensen AG. Tracer arrival timing-insensitive technique for estimating flow in MR perfusion-weighted imaging using singular value decomposition with a block-circulant deconvolution matrix. Magn Reson Med 2003; 50:164–174. 29. Thilmann O. LUPE: an extensible modular framework for evaluation of DSC-acquired perfusion images. Magn Reson Mater Phys 2004; 16(Suppl 1):537. 30. Friston KJ, Ashburner J, Frith CD, Poline JB, Heather JD, Frackowiak RSJ. Spatial registration and normalization of images. Hum Brain Mapp 1995;3:165–189. 31. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307– 310. 32. Hatazawa J, Shimosegawa E, Toyoshima H, Ardekani BA, Suzuki A, Okudera T, Miura Y. Cerebral blood volume in acute brain infarction: a combined study with dynamic susceptibility contrast MRI and 99mTC-HMPAO-SPECT. Stroke 1999;30:800–806. 33. Koenig M, Kraus M, Theek C, Klotz E, Gehlen W, Heuser L. Quantitative assessment of the ischemic brain by means of perfusion-related parameters derived from perfusion CT. Stroke 2001;32:431–437. 34. Latchaw RE, Yonas H, Hunter GJ, et al; Council on Cardiovascular Radiology of the American Heart Association. Guidelines and recommendations for perfusion imaging in cerebral ischemia: a scientific statement for healthcare professionals by the writing group on perfusion imaging, from the Council on Cardiovascular Radiology of the American Heart Association. Stroke 2003;34:1084–1104. 35. Donahue MJ, Lu H, Jones CK, Pekar JJ, van Zijl PC. An account of the discrepancy between MRI and PET cerebral blood flow measures. A high-field MRI investigation. NMR Biomed 2006;19:1043–1054.

Knutsson et al. 36. Gunther M, Bock M, Schad LR. Arterial spin labeling in combination with a look-locker sampling strategy: inflow turbo-sampling EPI-FAIR (ITS-FAIR). Magn Reson Med 2001;46:974–984. 37. Shin W, Horowitz S, Ragin A, Chen Y, Walker M, Carroll TJ. Quantitative cerebral perfusion using dynamic susceptibility contrast MRI: evaluation of reproducibility and age- and gender-dependence with fully automatic image postprocessing algorithm. Magn Reson Med 2007;58:1232–1241. 38. Petersen ET, Mouridsen K, Golay X, on behalf of all named coauthors of the QUASAR test-retest. The QUASAR reproducibility study, Part II: results from a multi center Arterial Spin Labeling test-retest Study. Neuroimage 2010;49:104–113. 39. Kudo K, Terae S, Katoh C, Oka M, Shiga T, Tamaki N, Miyasaka K. Quantitative cerebral blood flow measurement with dynamic perfusion CT using the vascular-pixel elimination method: comparison with H2(15)O positron emission tomography. AJNR Am J Neuroradiol 2003:24;419–426. 40. van Osch MJP, Vonken EPA, Viergever MA, van der Grond J, Bakker CJG. Measuring the arterial input function with gradient echo sequences. Magn Reson Med 2003;49:1067–1076. 41. Kjïlby 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;61:1300–1309. 42. Calamante F, Connelly A, van Osch MJ. Nonlinear DeltaR*2 effects in perfusion quantification using bolus-tracking MRI. Magn Reson Med 2009;61:486–489. 43. Knutsson L, Sta˚hlberg F, Wirestam R, van Osch MJ. Effects of blood DR(2)* non-linearity on absolute perfusion quantification using DSCMRI: comparison with Xe-133 SPECT. Magn Reson Imaging 2013;31: 651–655.