Visual Scoring Versus Computational Volume Measures of Perfusion Defects In Acute Ischaemic Stroke

Background: Perfusion imaging is used increasingly to assess ischaemic stroke. The Third International Stroke Trial (IST-3) was a large multicentre randomised-controlled trial testing intravenous rt-PA for ischaemic stroke; selected centres collected perfusion imaging. Methods: All IST-3 patients with baseline CT/MR perfusion are included. We assessed perfusion defects on cerebral blood flow (CBF), cerebral blood volume (CBV), mean transit time (MTT) and time to maximum flow (Tmax) perfusion maps both visually with ASPECTS and by a semi-automated computer algorithm that used six predefined thresholds to determine volume. We tested for univariate associations between ASPECTS and computationally measured volumes, and between ASPECTS/computational volumes and age, stroke to scan time, baseline NIHSS. We performed ordinal regression analyses (adjusted for age, NIHSS, stroke to scan time, treatment allocation) to identify associations between ASPECTS and 6-month functional outcome. Results: Amongst 121 patients with baseline perfusion imaging, median (interquartile): age 81(73–86) years; NIHSS 11(6–18); stroke to scan time 169(117–240) minutes; 64(53%) male, 57(47%) treated with rt-PA. ASPECTS did not correlate with computational perfusion volume for any perfusion map or threshold. ASPECTS CBF and CBV were significantly lower (larger defects) among patients scanned <3 versus 3–6 hours (p<0.05). ASPECTS correlated significantly with NIHSS for all perfusion maps (p<0.0001) but not age. Computational perfusion volumes were not associated with age, time to scan or NIHSS. ASPECTS independently predicted poor outcome for all perfusion maps (p<0.01). Conclusions: For assessing ischaemic stroke severity and predicting outcome, visual scoring of perfusion deficits is quick and may be superior to computational analysis regardless of threshold used.