INTRODUCTION
MRI-measured gray-matter (GM) cerebral blood flow (CBF) values are generally larger than PET values, with the underlying reason(s) remaining a topic of discussion 1 . PET literature commonly reports normal GM CBF values of 40–55 ml/100g/min, whereas MRI values as high as 100 ml/100g/min have been published 2 . We demonstrate two main causes of this discrepancy, one relating to GM CBF overestimation by the MRI approach of arterial spin labeling (ASL) and one relating to GM CBF underestimation by PET. First, ASL overestimates perfusion due to insufficient time for in-flowing labeled blood water to leave small arteries and enter tissue. Although this has been previously noticed 3 , such short label inversion times (TI), remain in use at low magnetic field strength where the relaxation time T1 of arterial blood water is short (1350 ms 4 ). We evaluate TI influence on CBF measurements at 3 T, where arterial T1 is much longer (1630 ms 5 ). Second, due to the low resolution at which PET measurements are acquired, partial volume effects between GM, white matter, and cerebrospinal fluid cause GM CBF to be underestimated. We evaluate these effects by analyzing ASL CBF maps as a function of resolution.
Methods
1. ASL studies were performed on 5 subjects at 3 T using the transfer insensitive labeling technique (TILT 6 ), a pulsed ASL technique that labels blood proximal to the imaging slice. Two sets of ASL experiments were performed:
To analyze partial volume effects, multiple-resolution studies were performed with FOV=240 mm, slice thickness=5 mm, and in-plane isotropic matrix sizes=96,80,64,48, and 32 (TR/TI = 2000/1500 ms). Images were interpolated to identical in-plane resolution (128×128).
2. To analyze arterial flow contributions, CBF maps were generated varying TI between 100 ms and 2500 ms; these maps were overlaid on MR angiography (MRA) maps to isolate arterial flow regions.
Results
Multiple resolution CBF maps (in-plane resolution 2.5–7.5 mm2) in Figs. 1a–e illustrate how CBF varies with resolution, with 7.5 mm2 roughly mimicking PET resolution. GM CBF decreases by approximately 20.5 % over this range (Fig. 1f), thereby implicating partial volume effects at PET resolutions. CBF dependence on labeling delay is shown for five different regions of interest in Fig. 1g. At TI<1500 ms, CBF is greatly overestimated by MRI ASL due to arterial contributions, but CBF roughly plateaus at TI1500 ms, once tagged arterial blood water has left the imaging slice and/or entered tissue. At 1.5 T, the T1 of blood water is much shorter than at 3 T, therefore perfusion imaging at TI1500 ms is difficult since most of the tag has decayed. These results largely reconcile the differences between PET and MRI CBF measurements and point toward a true measure of perfusion independent of imaging modality.
