Automated estimation of ischemic core prior to thrombectomy: comparison of two current algorithms.

Abstract

Endovascular thrombectomy (EVT) improves clinical outcomes in ischemic stroke with large vessel occlusion. Clinical benefits are inversely proportional to size of the pre-treatment ischemic core. This study compared estimated ischemic core volumes by two different CT perfusion (CTP) automated algorithms to the gold standard follow-up infarct volume using diffusion-weighted imaging (DWI) to assess for congruence, and thus eligibility for EVT. Retrospective, single-center cohort study of 102 patients presenting to a comprehensive stroke center between 2012 and 2018. Inclusion criteria were CT perfusion prior to EVT, successful EVT with mTIBI 2b-3 reperfusion, and DWI post-EVT. CTP data were retrospectively processed by two algorithms: "delay and dispersion insensitive deconvolution" (DISD, RAPID software) versus "delay and dispersion corrected single value decomposition" (ddSVD, Mistar software), using commercially available software. Core volumes were compared to follow up DWI using independent software (MRIcron). Agreement between each algorithm and DWI was estimated using Lin's concordance coefficient and analyzed using reduced major axis regression. We included 102 patients. Both algorithms had excellent agreement with DWI (Lin's concordance coefficients: DISD 0.8 (95% CI: 0.73; 0.87), ddSVD 0.92 (95% CI: 0.89; 0.95). Compared to ddSVD (reduced major axis slope = 0.95), DISD exhibited a larger extent of proportional bias (slope = 1.12). The ddSVD algorithm better correlates with DWI follow-up infarct volume than DISD processing. The DISD algorithm overestimated larger ischemic cores which may lead to patient exclusion from thrombectomy based on selection by core volume.