Automated Perfusion Calculations vs. Visual Scoring of Collaterals and CBV-ASPECTS
Marios‐Nikos Psychogios, Peter B. Sporns, Johanna M. Ospel, Aristeidis H. Katsanos, Reza Kabiri, Fabian Flottmann, Bijoy K. Menon, MacKenzie Horn, David S. Liebeskind, T Honda, Marc Ribó, Manuel Requena, Christoph Kabbasch, Thorsten Lichtenstein, Christoph J. Maurer, Ansgar Berlis, Victoria Hellstern, Hans Henkes, Markus Möhlenbruch, Fatih Şeker, Marielle Ernst, Jan Liman, Georgios Tsivgoulis, Alex Brehm
Abstract
PURPOSE: Use of automated perfusion software has gained importance for imaging of stroke patients for mechanical thrombectomy (MT). We aim to compare four perfusion software packages: 1) with respect to their association with 3‑month functional outcome after successful reperfusion with MT in comparison to visual Cerebral Blood Volume - Alberta Stroke Program Early CT Score (CBV-ASPECTS) and collateral scoring and 2) with respect to their agreement in estimation of core and penumbra volume. METHODS: This retrospective, multicenter cohort study (2015-2019) analyzed data from 8 centers. We included patients who were functionally independent before and underwent successful MT of the middle cerebral artery. Primary outcome measurements were the relationship of core and penumbra volume calculated by each software, qualitative assessment of collaterals and CBV-APECTS with 3‑month functional outcome and disability (modified Rankin scale >2). Quantitative differences between perfusion software measurements were also assessed. RESULTS: A total of 215 patients (57% women, median age 77 years) from 8 centers fulfilled the inclusion criteria. Multivariable analyses showed a significant association of RAPID core (common odds ratio, cOR 1.02; p = 0.015), CBV-ASPECTS (cOR 0.78; p = 0.007) and collaterals (cOR 0.78; p = 0.001) with 3‑month functional outcome (shift analysis), while RAPID core (OR 1.02; p = 0.018), CBV-ASPECTS (OR 0.77; p = 0.024), collaterals (OR 0.78; p = 0.007) and OLEA core (OR 1.02; p = 0.029) were significantly associated with 3‑month functional disability. Mean differences on core estimates between VEOcore and RAPID were 13.4 ml, between syngo.via and RAPID 30.0 ml and between OLEA and RAPID -3.2 ml. CONCLUSION: Collateral scoring, CBV-ASPECTS and RAPID were independently associated with functional outcome at 90 days. Core and Penumbra estimates using automated software packages varied significantly and should therefore be used with caution.