Abstract 3417: Pixel by Pixel Comparison of CTP-Defined Infarct Core on Concurrent Non-Contrast CT

Abstract

Listen

Translate article

Introduction Interobserver agreement for the presence of early ischemic changes on non-contrast CT (NCCT) in potential thrombolysis candidates can be poor even among experienced raters. CT perfusion (CTP) may demonstrate infarct core and penumbra according to proposed thresholds but the extent of correlation between CTP defined core and hypodensity on hyperacute NCCT is unclear Methods A pixel by pixel comparison between NCCT and concurrent CTP obtained <6hrs from symptom onset was performed for patients with symptomatic arterial occlusions. Segmentation was performed using time-averaged CTP raw images to compare core and normal tissue pixels for grey and white matter separately. Core was defined as relative Cerebral Blood Flow (CBF) of <45% of normal tissue. NCCT was co-registered to the baseline frames of the concurrent CTP using a rigid body transformation. Core pixels were extracted from CTP and transposed to NCCT. Hounsfield unit (HU) values in CTP-defined core and normal tissue were quantified for grey and white matter. Receiver operating characteristic (ROC) curve analysis was performed to assess the predictive value of NCCT hypodensity and CTP-defined core. Results CT and CTP were compared for 33 patients. Mean age was 72 years (SD 12), median NIHSS was 16 (IQR 11-20) and mean time from symptom onset to CTP was 191 minutes (SD 63). Area under the curve (AUC) for prediction of grey and white matter core by NCCT was 0.641 and 0.601 respectively. The most frequently occurring optimum HU cut-points were 29 and 28 for grey and white matter respectively .These cut-points were associated with low overall specificity for detecting core (Median ± IQR= 0.48, 0.32-0.59 and 0.36, 0.26-0.44) but higher sensitivity (Median ±IQR 0.79, 0.62-0.87 and 0.83, 0.73-0.89) for grey and white matter Conclusion Objectively determined hypodensity on NCCT has low sensitivity and specificity for prediction of ischemic core tissue defined by CT perfusion. Thresholds were similar for both grey and white matter. NCCT scans without established hypodensity may still have irreversibly infarcted pixels which can be detected with CTP. Prospective evaluation of the added value of CTP-defined core on decision making in acute stroke is needed.