Influence of the arterial input function on absolute and relative perfusion-weighted imaging penumbral flow detection: a validation with ¹⁵O-water positron emission tomography.

BACKGROUND AND PURPOSE Perfusion-weighted imaging maps are used to identify critical hypoperfusion in acute stroke. However, quantification of perfusion may depend on the choice of the arterial input function (AIF). Using quantitative positron emission tomography we evaluated the influence of the AIF location on maps of absolute and relative perfusion-weighted imaging to detect penumbral flow (PF; <20 mL/100 g/min on positron emission tomography(CBF)) in acute stroke. METHODS In 22 patients with acute stroke the AIF was placed at 7 sites (M1, M2, M3 ipsi- and contralateral and internal carotid artery-M1 contralateral to the infarct). Comparative (15)O-water positron emission tomography and AIF-dependent perfusion-weighted imaging (cerebral blood flow, cerebral blood volume, mean transit time, and time to maximum) were performed. A receiver operating characteristic curve analysis described the threshold independent performance (area under the curve) of the perfusion-weighted maps for all 7 AIF locations and identified the best AIF-dependent absolute and relative thresholds to identify PF. These results were compared with AIF-independent time-to-peak maps. RESULTS Quantitative perfusion-weighted imaging maps of cerebral blood flow and time to maximum performed best. For PF detection, AIF placement did significantly influence absolute PF thresholds. However, AIF placement did not influence (1) the threshold independent performance; and (2) the relative PF thresholds. AIF placement in the proximal segment of the contralateral middle cerebral artery (cM1) was preferable for quantification. CONCLUSIONS AIF-based maps of cerebral blood flow and time to maximum were most accurate to detect the PF threshold. The AIF placement significantly altered absolute PF thresholds and showed best agreement with positron emission tomography for the cM1 segment. The performance of relative PF thresholds, however, was not AIF location-dependent and might be along with AIF-independent time-to-peak maps, more suitable than absolute PF thresholds in acute stroke if detailed postprocessing is not feasible.