- Email: [email protected]
Correction for vascular artifacts in cerebral blood flow imaging
ABSTRACTS
Correction for Vascular Artifacts in Cerebral Blood Flow Imaging F. Q. Ye, V. S. Mattay, P. Jezzard, J. A. Frank, A. C. McLaughlin CBDB (NIMH), LDRR (OIR), National Institutes of Health, Bethesda, MD, USA BOLD approaches have been very successful in studying regional alterations in brain function during specific activation paradigms. However, the interpretation of BOLD results can be complicated by the dependence on a number of parameters, e.g., cerebral blood flow, cerebral oxygen consumption and cerebral blood volume. One way to circumvent these complications is to measure cerebral blood flow directly during functional activation paradigms. Steady-state arterial spin tagging approaches have been used to image cerebral blood flow in humans (1,2), using the following equation AM/M ~ =
2 ot Q/~ 1/Tlobs [1] where AM/M o is the fractional decrease in longitudinal brain water magnetization when arterial water spins are perturbed, o~ is the degree of arterial inversion, s is the brain-blood partition coefficient for water, and t/Tlobs is the observed longitudinal relaxation rate of brain water spins. This study addresses two problems in the use of Eqn [1] to calculate cerebral blood flow: (1) Eqn [1] applies only to tissue water magnetization, not to vascular water magnetization: tissue water signals must be separated from vascular water signals. (2) Eqn [1] does not take into account relaxation of the "tagged" arterial blood during transit from the "tagging" plane to tissue in the imaging slice~ This effect reduces o~ by a factor exp(-x FFlb), where x is the transit time, and Tlb is the relaxation time of arterial water spins. Experiments were performed on human volunteers in a 1.5 T GE scanner employing a 3-axis local gradient head coil (Medical Advances, WI). Spin echo EPI images were obtained using a 64x64 matrix, fov = 24 cm, slice thickness = 5 mm. TE = 60 ms. TR = 2 s (Figure 1) or 4 s (Figure 2). Bipolar z crusher gradients (3 ms square pulses separated by 3 ms) were placed between the 90 ~ and 180 ~ pulses. Flow induced adiabatic inversion was used to invert arterial water spins in an axial "tagging" plane 3 cm proximal to the axial imaging slice. Figure 1 shows the effect of crusher strength on the value of AM/M o obtained from a whole-brain ROI, using a 1.8 s tagging pulse. This data suggests that strong bipolar gradients can "crush" the signal from moving water spins in the vascular bed, and that values of AM/M o obtained in the presence of strong crushers can be used in Eqn [1] to calculate cerebral blood flow. Figure 2 shows the dependence of AM/M o for grey matter on the length of the tagging period. The top and bottom curves show results obtained with crusher gradients of 0 and 1.7 Gauss/cm, respectively. This data suggests that the arterial transit time from the "tagging" plane to tissue in the imaging slice is - 0.7 s for grey matter. The transit time observed for white matter is substantially longer (- 1.2 s). These results are similar to those found using EPISTAR (3). Assuming that Tib = 1.2 s, O~is reduced by - 44% for grey matter regions and 64% for white matter regions, due to relaxation during arterial transit. Contributions from vascular spins to AM/M o, and relaxation during transit of "tagged" arterial blood can cause substantial errors in cerebral blood flow values calculated using Eqn [1]. Contributions from vascular spins can be minimized using strong crusher gradients, and cerebral blood flow values can be corrected for relaxation of tagged arterial blood using arterial transit times measured with "dynamic" spin tagging approaches. (1) Roberts DA, Detre JA, Bollinger L, Insko EK, Leigh JS., Proc. Nat. Acad. Sci (USA) 91, 33-37 (1994). (2) Pekar J, Jezzard P, Duyn J, Frank JA, McLaughlin AC, SMR, Annual Meeting, Nice, 1995, p. 884. (3) Buxton RB, Frank LR, Siewert B, Warach S, Edelman RR, SMR, Annual Meeting, Nice, 1995, p. 132. 2.5
2 1
2
,-.1.5-
1.5 O
O
1-
1
0.5
<~ 0 . 5 -
o 0
0
I
I
I
0.5
1
1.5
-0.5 2
I
0
Crusher Gradient (G/cm) Figure 1: AM/Mo as a function of crusher strength.
I
I
1000 2000 3000 Tagging Time (ms)
4000
Figure 2: AM/M o as a function of rf tagging period.
$44
Correction for Vascular Artifacts in Cerebral Blood Flow Imaging F. Q. Ye, V. S. Mattay, P. Jezzard, J. A. Frank, A. C. McLaughlin CBDB (NIMH), LDRR (OIR), National Institutes of Health, Bethesda, MD, USA BOLD approaches have been very successful in studying regional alterations in brain function during specific activation paradigms. However, the interpretation of BOLD results can be complicated by the dependence on a number of parameters, e.g., cerebral blood flow, cerebral oxygen consumption and cerebral blood volume. One way to circumvent these complications is to measure cerebral blood flow directly during functional activation paradigms. Steady-state arterial spin tagging approaches have been used to image cerebral blood flow in humans (1,2), using the following equation AM/M ~ =
2 ot Q/~ 1/Tlobs [1] where AM/M o is the fractional decrease in longitudinal brain water magnetization when arterial water spins are perturbed, o~ is the degree of arterial inversion, s is the brain-blood partition coefficient for water, and t/Tlobs is the observed longitudinal relaxation rate of brain water spins. This study addresses two problems in the use of Eqn [1] to calculate cerebral blood flow: (1) Eqn [1] applies only to tissue water magnetization, not to vascular water magnetization: tissue water signals must be separated from vascular water signals. (2) Eqn [1] does not take into account relaxation of the "tagged" arterial blood during transit from the "tagging" plane to tissue in the imaging slice~ This effect reduces o~ by a factor exp(-x FFlb), where x is the transit time, and Tlb is the relaxation time of arterial water spins. Experiments were performed on human volunteers in a 1.5 T GE scanner employing a 3-axis local gradient head coil (Medical Advances, WI). Spin echo EPI images were obtained using a 64x64 matrix, fov = 24 cm, slice thickness = 5 mm. TE = 60 ms. TR = 2 s (Figure 1) or 4 s (Figure 2). Bipolar z crusher gradients (3 ms square pulses separated by 3 ms) were placed between the 90 ~ and 180 ~ pulses. Flow induced adiabatic inversion was used to invert arterial water spins in an axial "tagging" plane 3 cm proximal to the axial imaging slice. Figure 1 shows the effect of crusher strength on the value of AM/M o obtained from a whole-brain ROI, using a 1.8 s tagging pulse. This data suggests that strong bipolar gradients can "crush" the signal from moving water spins in the vascular bed, and that values of AM/M o obtained in the presence of strong crushers can be used in Eqn [1] to calculate cerebral blood flow. Figure 2 shows the dependence of AM/M o for grey matter on the length of the tagging period. The top and bottom curves show results obtained with crusher gradients of 0 and 1.7 Gauss/cm, respectively. This data suggests that the arterial transit time from the "tagging" plane to tissue in the imaging slice is - 0.7 s for grey matter. The transit time observed for white matter is substantially longer (- 1.2 s). These results are similar to those found using EPISTAR (3). Assuming that Tib = 1.2 s, O~is reduced by - 44% for grey matter regions and 64% for white matter regions, due to relaxation during arterial transit. Contributions from vascular spins to AM/M o, and relaxation during transit of "tagged" arterial blood can cause substantial errors in cerebral blood flow values calculated using Eqn [1]. Contributions from vascular spins can be minimized using strong crusher gradients, and cerebral blood flow values can be corrected for relaxation of tagged arterial blood using arterial transit times measured with "dynamic" spin tagging approaches. (1) Roberts DA, Detre JA, Bollinger L, Insko EK, Leigh JS., Proc. Nat. Acad. Sci (USA) 91, 33-37 (1994). (2) Pekar J, Jezzard P, Duyn J, Frank JA, McLaughlin AC, SMR, Annual Meeting, Nice, 1995, p. 884. (3) Buxton RB, Frank LR, Siewert B, Warach S, Edelman RR, SMR, Annual Meeting, Nice, 1995, p. 132. 2.5
2 1
2
,-.1.5-
1.5 O
O
1-
1
0.5
<~ 0 . 5 -
o 0
0
I
I
I
0.5
1
1.5
-0.5 2
I
0
Crusher Gradient (G/cm) Figure 1: AM/Mo as a function of crusher strength.
I
I
1000 2000 3000 Tagging Time (ms)
4000
Figure 2: AM/M o as a function of rf tagging period.
$44