Abstract TP101: Defining Ischemic Core In Acute Ischemic Stroke Using CT Perfusion: A Multi-center Validation Study

Background: Estimation of infarction based on computed tomographic perfusion (CTP) remains challenging, mainly due to noise associated with CTP data and variability in reported thresholds ischemic core esstimation. Commercial software companies have attempted to establish CTP thresholds; for example, a relative cerebral blood flow (rCBF) of <30% for RAPID™ (iSchemaView). In this retrospective multi-center study, we sought to establish and validate CTP threshold values for estimation of ischemic core using Olea Sphere® software (Olea Medical Solutions). Methods: Under an approved multi-institutional protocol, patients with acute ischemic stroke of the anterior circulation from three comprehensive stroke centers were included if they met the following criteria: baseline CTP, recanalization achieved (Thrombolysis in cerebral infarction scale ≥ IIb), and follow up MRI (DWI) performed. Perfusion maps were generated from the CTP data by the Bayesian probabilistic model and oscillation singular value decomposition (oSVD). CTP parameters, including time to peak (TTP), rCBF, time to max (Tmax), and Delay (equivalent of Tmax in Bayesian model) were generated. Brain was extracted from CTP and coregistered with the follow up MRI. An infarction mask and non-infarcted masks were drawn on MRI. We used a robust logistic regression to assess the binary outcome of our voxel-based analysis (infarcted vs non-infarcted) by adjusting for intra-subject correlations. Results: Eighty-one patients (42 Male; mean age 67.4 ±17.3) were included. All CTP parameters for the Bayesian and oSVD methods showed significant differences between infarcted and non-infarcted areas (p < 0.001). In the Bayesian model, a combination of rCBF at threshold of <45% and delay at threshold of 0.82 sec provided the highest diagnostic accuracy; AUC/sensitivity/specificity (0.79/0.69/0.79). In the oSVD method, a combination of rCBF at threshold of <16% and Tmax at threshold of 4.00 sec provided the highest diagnostic accuracy; AUC/sensitivity/specificity (0.71/0.64/0.73). Conclusion: Using heterogenous CTP data from three stroke centers, we established thresholds for estimation of ischemic core for Olea software that can be used for both Bayesian and oSVD postprocessing models.