Accuracy of perfusion-CT in predicting malignant middle cerebral artery brain infarction

Background We performed a prospective study on patients with middle cerebral artery (MCA) ischemic stroke to evaluate the accuracy of perfusion-CT imaging (PCT) to predict the development of malignant brain infarction (MBI). Methods 106 patients (women 37%, mean age 65 years) underwent native cranial computed tomography (CCT), CT angiography (CTA) and PCT after a median of 2 h after stroke onset. We assessed the patency of the MCA and the area of tissue ischemia (AIT) according to cerebral blood flow (CBF), cerebral blood volume (CBV) and time-to-peak (TTP) maps. Optimum sensitivity, specificity, positive (PPV) and negative predictive values (NPV) were calculated for the end-point MBI (=midline shift > 5 mm or decompressive surgery) by means of receiver operating characteristics (ROC). Results 20 patients (19 %) developed a MBI. In these patients, a larger AIT was found in all perfusion maps as compared to the remaining patients (p < 0.001). All perfusion maps had a very high NPV (95.4-98.4%), a high sensitivity (85-95 %) and specificity (71.6-77.9 %) and only a moderate PPV (44-47.4 %). Best prediction was found for CBF maps with AIT of > 27.9 % of the hemisphere. Conclusion PCT allows the discrimination of patients without a relevant risk for MBI from those having a 50 % risk of MBI development. Due to the high sensitivity and specificity, PCT is a reliable tool in detecting MBI. Because of PCT's better availability, it is the method of choice at present for an early risk stratification of acute stroke patients.