• Bruce C V Campbell, Søren Christensen, Christopher R Levi, Patricia M Desmond, Geoffrey A Donnan, Stephen M Davis, and Mark W Parsons.
• Department of Medicine, The Royal Melbourne Hospital, University of Melbourne, Grattan Street, Parkville VIC 3050, Australia. bruce.campbell@mh.org.au
• Stroke. 2011 Dec 1;42(12):3435-40.

Background And PurposeCT perfusion (CTP) is widely and rapidly accessible for imaging acute ischemic stroke but has limited validation. Cerebral blood volume (CBV) has been proposed as the best predictor of infarct core. We tested CBV against other common CTP parameters using contemporaneous diffusion MRI.MethodsPatients with acute ischemic stroke<6 hours after onset had CTP and diffusion MRI<1 hour apart, before any reperfusion therapies. CTP maps of time to peak (TTP), absolute and relative CBV, cerebral blood flow (CBF), mean transit time (MTT), and time to peak of the deconvolved tissue residue function (Tmax) were generated. The diffusion lesion was manually outlined to its maximal visual extent. Receiver operating characteristic (ROC) analysis area under the curve (AUC) was used to quantify the correspondence of each perfusion parameter to the coregistered diffusion-weighted imaging lesion. Optimal thresholds were determined (Youden index).ResultsIn analysis of 98 CTP slabs (54 patients, median onset to CT 190 minutes, median CT to MR 30 minutes), relative CBF performed best (AUC, 0.79; 95% CI, 0.77-81), significantly better than absolute CBV (AUC, 0.74; 95% CI, 0.73-0.76). The optimal threshold was <31% of mean contralateral CBF. Specificity was reduced by low CBF/CBV in noninfarcted white matter in cases with reduced contrast bolus intensity and leukoaraiosis.ConclusionsIn contrast to previous reports, CBF corresponded with the acute diffusion-weighted imaging lesion better than CBV, although no single threshold avoids detection of false-positive regions in unaffected white matter. This relates to low signal-to-noise ratio in CTP maps and emphasizes the need for optimized acquisition and postprocessing.

22 keywords...

hide…