Abstract Background and aims Perfusion imaging is routinely used to estimate infarct core and penumbra in patients with large vessel occlusion (LVOs), thus having an important role in reperfusion strategies. Common perfusion-derived threshold-based maps (Tmax, CBF, CBV) have been validated mainly in LVOs and in trials that shaped thrombectomy indications. However, medium/distal vessel occlusions (MDVOs) differ from LVOs in occlusion size, collateral patterns, hemodynamic impact, and often produce smaller, more heterogeneous perfusion abnormalities. These differences lead to the hypothesis of reduced accuracy of standard perfusion thresholds and automated software outputs when used to predict final infarct in MDVOs . We investigated the effectiveness of LVO-perfusion-derived maps in prediction of final infarct in MDVO. Methods This retrospective single center observational cohort study included 71 MDVO-patients. Patients were stratified into fully recanalized (TICI2b ) and non-recanalized (TICI2a) groups. Using RapidAI-CTP and an automated pipeline adapted from our previous work, we segmented four perfusion maps: Tmax4 s , Tmax6 s, cerebral blood flow (CBF), and cerebral blood volume (CBV). Final infarct was manually segmented on diffusionweighted MRI performed ~24h postpresentation. Perfusion–DWI spatial correspondence and volumetric agreement were assessed with different metrics (Absolute and relative volume difference, Dice score, Hausdorff-distance and distance between the centers of mass). We exploratively analyzed the data and correlated them with anatomical (occlusion-site, vessel-size) and clinical parameters (NIHSS, mRS). Results Currently we are at the result-interpretation-phase. Our early results suggest that the infarct-core is poorly represented by perfusion maps. Conflict of interest No disclosures for the authors regarding the carrying-out of the project, no specific funding sources were required for this project
Ntoulias et al. (Fri,) studied this question.