Abstract WP39: Perfusion Angiography in TREVO2: Quantitative Reperfusion After Endovascular Therapy in Acute Stroke

Background: Effective reperfusion is the ultimate therapeutic strategy for acute ischemic stroke, yet endovascular trials have been limited by the use of categorical scales for angiographic outcomes. Semi-automated perfusion angiography (PA) software can objectively quantify reperfusion from routine angiography. We studied feasibility and performance of PA software to evaluate angiographic and associated clinical outcomes in the multicenter TREVO2 trial. Methods: Core lab angiography DICOM data in TREVO2 was retrospectively used to identify anterior circulation cases with adequate temporal resolution at both baseline and after revascularization with mechanical thrombectomy. CBF was computed by deconvolution of contrast-intensity data up to 3 sec to correlate with TICI and avoid collaterals. Regions of interest (ROI) outlined areas of hypoperfusion on baseline AP and lateral angiography to map reperfusion onto identical post-procedure angiography runs. A nonlinear cross-validation model was used to map output from the ROI to a single reperfusion metric (PA CBF 3 ). Results: 148/178 (83%) patients with anterior circulation stroke in TREVO2 had DICOM angiography data with complete temporal information on serial runs at baseline and post-procedure that could be processed with the software. Core lab TICI scores in this dataset ranged from 0-3 (0, n=6; 1, n=7; 2a, n=40; 2b, n=83; 3, n=12). The continuous PA CBF 3 metric or reperfusion score ranged from 0-8.6 (mean 3.7±1.2), capturing hyperperfusion and heterogeneity. Overall, PA CBF 3 closely correlated with TICI (ρ=0.69, p 3 results were equivalent in ICA, M1 and M2 occlusions treated in TREVO2. Conclusions: PA can objectively quantify heterogeneity of reperfusion in a multicenter trial, providing a continuous metric that discriminates angiographic outcomes better than TICI.