The Performance of MRI-Based Cerebral Blood Flow Measurements in Acute and Subacute Stroke Compared With 15O-Water Positron Emission Tomography: Identification of Penumbral Flow

Background and Purpose— Perfusion-weighted MRI-based maps of cerebral blood flow (CBFMRI) are considered a good MRI measure of penumbral flow in acute ischemic stroke but are seldom used in clinical routine due to methodical issues. We validated CBFMRI on quantitative CBF measurement by 15O-water positron emission tomography (CBFPET). Material and Methods— Comparative CBFMRI and CBFPET were performed in patients with acute and subacute stroke. In a voxel-based seed-growing technique, predefined CBFMRI thresholds (<40, <30, <20, <10 mL/100 g/min) were applied and the resulting volumes were compared with the hypoperfusion volume detected by the penumbral threshold (<20 mL/100 g/min) on CBFPET. The volumetric comparison was expressed as the C-ratio (volume CBFMRI/volume CBFPET) to identify the best MRI threshold. The influence of vessel pathology, hypoperfusion size, and time point of imaging was described. The proportion of voxels correctly classified as hypoperfused and the proportion of voxel correctly classified as nonhypoperfused of the best CBFMRI threshold was calculated and a Bland-Altman plot illustrated the method-specific differences. Results— In 24 patients (median time MRI to PET: 68 minutes; 16 patients imaged within 24 hours after stroke), the median volume of hypoperfusion <20 mL/100 g/min (CBFPET) was 78.5 cm3. Median hypoperfusion volume on CBFMRI ranged from 245.9 cm3 (<40 mL/100 g/min) to 35.5 cm3 (<10 mL/10 g/min). On visual inspection, an excellent qualitative congruence was found. The quantitative congruence was best for the MRI-CBF threshold <20 mL/100 g/min (median C-ratio: 1.0), reaching a proportion of voxels correctly classified as hypoperfused of 76% and a proportion of voxel correctly classified as nonhypoperfused of 96%, but a wide interindividual range (C-ratio 0.3 to 3.5) was found. Ipsilateral vessel pathology, time point of imaging, and size of hypoperfusion did not significantly influence the C-ratio. The Bland-Altman analysis for the volumetric difference of CBFMRI and CBFPET found a good overall agreement but a large SD. Conclusion— Hypoperfusion areas below the CBFPET penumbral threshold can be well identified by the CBFMRI threshold <20 mL/10 g/min at a group level, but a large individual variance (exceeding 20% of volume in nearly half of the patients) could not be explained. Our results support a prudent use of MRI-based quantitative CBF measurement in clinical routine.