Background and purpose: Advanced CT perfusion (CTP) imaging can estimate the size of the ischemic core, which can be used for selection of patients for endovascular therapy. The relative cerebral blood volume (rCBV) and cerebral blood flow (rCBF) thresholds chosen to identify ischemic core influence the accuracy of prediction. The purpose of this study was to analyze the accuracy of various rCBV and rCBF thresholds for predicting 27 hour infarct volume using RAPID automated analysis software.
Methods: Patients from the SWIFT PRIME study with baseline and 27 hour follow-up CT perfusion scans were included if they had complete reperfusion based on Tmax>6s perfusion maps obtained at 27 hours. Patients from both the tPA and endovascular groups were included. Infarct volume was determined on MRI (FLAIR images) or CT scans obtained 27 hours after symptom onset. The predicted ischemic core volume on rCBV and rCBF maps using thresholds ranging between 0.2 and 0.8 was compared with the actual infarct volume to determine the most accurate thresholds.
Results: Among the 47 subjects, the following baseline CTP thresholds most accurately predicted the actual 27 hour infarct volume: rCBV=0.34 (Median absolute error (MAE)=11.5 ml); rCBV=0.36, MAE=9.9 ml; rCBV=0.4, MAE=12.5 ml; rCBF=0.3, MAE=8.8 ml; rCBF=0.32, MAE=7.3; and rCBF=0.34, MAE=7.8.
Conclusions: Brain regions with rCBF ≤ .32 or rCBV ≤ .36 provided the most accurate prediction of infarct volume in patients who achieved complete reperfusion with median absolute errors less than 10 ml. Our data support the value of automated image analysis software as a tool for accurate prediction of ischemic core lesion volume.