Samenvatting
Purpose:
Relative perfusion values measured with DSC-MRI are frequently used in the initial clinical work-up and the follow up of brain tumors [1-3]. However, absolute quantification may not be feasible due to the unknown difference in relaxivity between artery and tissue [4]. In this study, we compare the cerebral blood flow (CBF) values in glioblastoma multiforme, contralateral white as well as grey matter from T1-DCE-based [CBFT1] and T2*-DSC-based [CBFT2*] perfusion, as well as
the lesion-to-normal-white matter CBF ratios from both techniques.
Methods:
All experiments were approved by the local ethical board. 9 measurements were performed in 5 anatomopathologically proven recurrent glioblastma multiforme patients in the supine position at 1.5T using the quadrature head coil (patient 2 to 5 had a baseline and a follow up scan after antiangiogenic therapy; average time between the two scans in the same patient was 29 days). An IR-prepared-FLASH sequence acquiring 3 slices per second (TR 2.4 ms/ TE 2 ms/ TI 190 ms/ FA
50°/ matrix 146*256/ FOV 219-270 mm/ dynamics 250, temporal resolution 0.3s per slice, slice thickness 5 mm) was performed during the injection of 15 ml of a Gd-DTPA solution (Magnevist®) at 2ml/sec. Ten minutes later, a second measurement using T2*-weighted GE EPI perfusion with three slices at the same imaging position as the DCE-based perfusion (12 slices, slice thickness 5 mm, TR 1590ms, TE 52ms, 50 dynamics, temporal resolution 1.43s per 12 slices) was performed with a second bolus of 20 ml Gd-DTPA solution injected at 4ml/sec. Post-processing was performed offline on a personal computer using the software PMI 0.2 written in-house in IDL (Research Systems, Boulder, CO)[5]. Signals were transformed into concentrations by using a test tube containing 2mM Gadolinium in saline solution placed in the FOV during the easurement. Tracer concentrations were deconvolved with an AIF obtained close to the tumor. In each patient the same AIF region was chosen for DCE- and DSC-MRI. A simple inflow correction was applied to the DCE data. Parametric maps of CBF from T1-DCE-based [CBFT1] and T2*-DSC-based perfusion [CBFT2*] were calculated as the maximum of the impulse response function. In each patient similar regions of interest were manually drawn by a radiologist (MD) on the CBFT1 and CBFT2* maps. CBFT1 and CBFT2* values were calculated for tumor, contralateral white and grey matter. These average CBF
values were compared, as well as lesion-to-normal CBF ratio for DCE-based (CBFLTN,T1) and DSCbased
(CBFLTN,T2*) perfusion.
Results:
Figure 1 illustrates a parametric map of CBFT1 (top) and CBFT2* (bottom) at the same tumor location. Figure 2 shows a scatter plot of CBFT1 and CBFT2* values for tumor as well as for grey and white matter and a scatter plot of CBFLTN,T1 and CBFLTN,T2*. No significant correlation was found between CBFT1 and CBFT2* in tumor (R= -0.6, p=0.07), in white (R= -0.07,
p=0.9) and grey matter (R= -0.5, p=0.2). However, CBFLTN,T1 and CBFLTN,T2* showed a significant
correlation (R=0.8, p=0.008).
Conclusion:
The CBF measures from DSC-based perfusion were systematically higher compared to the CBF from DCEbased
perfusion and CBF measures from both techniques did not show any correlation. Overestimation of DSCMRI
values was reported earlier [6]. The observation is consistent with the overestimation expected by the
difference in relaxivity between artery and tissue [4]. In view of the observed correlations between the relative
values, DSC-MRI CBF values may be reliable when normalized to a reference tissue type. However, since
values of CBFT1 are more in line with the findings for grey and white matter CBF from PET[7], DCE-based
perfusion measures of CBF seem to be preferable in tumor perfusion work-up and follow up.
References:
[1] Akella et al (2004) JMRI 20 (6): 913-22
[2] Hakyemez et al (2005) Clin Radiol 60(4): 493-502
[3] Hakyemez et al (2006) JMRI 24(4): 817-24
[4] Kiselev (2001) MRM 46: 1113-1122
[5] Sourbron et al (2007) PMB 52:429-447
[6] Smith et al (2000) JMRI 12: 400-410
[7] Grandin et al (2005) Neuroimage 26: 525-535
Relative perfusion values measured with DSC-MRI are frequently used in the initial clinical work-up and the follow up of brain tumors [1-3]. However, absolute quantification may not be feasible due to the unknown difference in relaxivity between artery and tissue [4]. In this study, we compare the cerebral blood flow (CBF) values in glioblastoma multiforme, contralateral white as well as grey matter from T1-DCE-based [CBFT1] and T2*-DSC-based [CBFT2*] perfusion, as well as
the lesion-to-normal-white matter CBF ratios from both techniques.
Methods:
All experiments were approved by the local ethical board. 9 measurements were performed in 5 anatomopathologically proven recurrent glioblastma multiforme patients in the supine position at 1.5T using the quadrature head coil (patient 2 to 5 had a baseline and a follow up scan after antiangiogenic therapy; average time between the two scans in the same patient was 29 days). An IR-prepared-FLASH sequence acquiring 3 slices per second (TR 2.4 ms/ TE 2 ms/ TI 190 ms/ FA
50°/ matrix 146*256/ FOV 219-270 mm/ dynamics 250, temporal resolution 0.3s per slice, slice thickness 5 mm) was performed during the injection of 15 ml of a Gd-DTPA solution (Magnevist®) at 2ml/sec. Ten minutes later, a second measurement using T2*-weighted GE EPI perfusion with three slices at the same imaging position as the DCE-based perfusion (12 slices, slice thickness 5 mm, TR 1590ms, TE 52ms, 50 dynamics, temporal resolution 1.43s per 12 slices) was performed with a second bolus of 20 ml Gd-DTPA solution injected at 4ml/sec. Post-processing was performed offline on a personal computer using the software PMI 0.2 written in-house in IDL (Research Systems, Boulder, CO)[5]. Signals were transformed into concentrations by using a test tube containing 2mM Gadolinium in saline solution placed in the FOV during the easurement. Tracer concentrations were deconvolved with an AIF obtained close to the tumor. In each patient the same AIF region was chosen for DCE- and DSC-MRI. A simple inflow correction was applied to the DCE data. Parametric maps of CBF from T1-DCE-based [CBFT1] and T2*-DSC-based perfusion [CBFT2*] were calculated as the maximum of the impulse response function. In each patient similar regions of interest were manually drawn by a radiologist (MD) on the CBFT1 and CBFT2* maps. CBFT1 and CBFT2* values were calculated for tumor, contralateral white and grey matter. These average CBF
values were compared, as well as lesion-to-normal CBF ratio for DCE-based (CBFLTN,T1) and DSCbased
(CBFLTN,T2*) perfusion.
Results:
Figure 1 illustrates a parametric map of CBFT1 (top) and CBFT2* (bottom) at the same tumor location. Figure 2 shows a scatter plot of CBFT1 and CBFT2* values for tumor as well as for grey and white matter and a scatter plot of CBFLTN,T1 and CBFLTN,T2*. No significant correlation was found between CBFT1 and CBFT2* in tumor (R= -0.6, p=0.07), in white (R= -0.07,
p=0.9) and grey matter (R= -0.5, p=0.2). However, CBFLTN,T1 and CBFLTN,T2* showed a significant
correlation (R=0.8, p=0.008).
Conclusion:
The CBF measures from DSC-based perfusion were systematically higher compared to the CBF from DCEbased
perfusion and CBF measures from both techniques did not show any correlation. Overestimation of DSCMRI
values was reported earlier [6]. The observation is consistent with the overestimation expected by the
difference in relaxivity between artery and tissue [4]. In view of the observed correlations between the relative
values, DSC-MRI CBF values may be reliable when normalized to a reference tissue type. However, since
values of CBFT1 are more in line with the findings for grey and white matter CBF from PET[7], DCE-based
perfusion measures of CBF seem to be preferable in tumor perfusion work-up and follow up.
References:
[1] Akella et al (2004) JMRI 20 (6): 913-22
[2] Hakyemez et al (2005) Clin Radiol 60(4): 493-502
[3] Hakyemez et al (2006) JMRI 24(4): 817-24
[4] Kiselev (2001) MRM 46: 1113-1122
[5] Sourbron et al (2007) PMB 52:429-447
[6] Smith et al (2000) JMRI 12: 400-410
[7] Grandin et al (2005) Neuroimage 26: 525-535