Accuracy of dynamic perfusion CT with deconvolution in detecting acute hemispheric stroke

BACKGROUND AND PURPOSE: Dynamic perfusion CT (PCT) with deconvolution produces maps of time-to-peak (TTP), mean transit time (MTT), regional cerebral blood flow (rCBF), and regional cerebral blood volume (rCBV), with a computerized automated map of the infarct and penumbra. We determined the accuracy of these maps in patients with suspected acute hemispheric stroke. METHODS: Forty-six patients underwent nonenhanced CT and dynamic PCT, with follow-up CT or MR imaging. Two observers reviewed the nonenhanced studies for signs of stroke and read the PCT maps for TTP, MTT, rCBF, and rCBV abnormalities. Sensitivity, specificity, accuracy, and interobserver agreement were compared (Wilcoxon tests). Nonenhanced CT and PCT data were reviewed for stroke extent according to previously reported methods. Sensitivity, specificity, and accuracy of the computerized maps in detecting ischemia and its extent were determined. RESULTS: Compared with nonenhanced CT, PCT maps were significantly more accurate in detecting stroke (75.7-86.0% vs. 66.2%; P < .01), MTT maps were significantly more sensitive (77.6% vs. 69.2%; P < .01), and rCBF and rCBV maps were significantly more specific (90.9% and 92.7%, respectively, vs. 65.0%; P < .01). Regarding stroke extent, PCT maps were significantly more sensitive than nonenhanced CT (up to 94.4% vs. 42.9%; P < .01) and had higher interobserver agreement (up to 0.763). For the computerized map, sensitivity, specificity, and accuracy, respectively, were 68.2%, 92.3%, and 88.1% in detecting ischemia and 72.2%, 91.8%, and 87.9% in showing the extent. CONCLUSION: Dynamic PCT maps are more accurate than nonenhanced CT in detecting hemispheric strokes. Despite limited spatial coverage, PCT is highly reliable to assess the stroke extent.