Study Population
This represents a single-center retrospective observational cohort study of patients suffering from AIS due to LVO, consecutively treated at the University Medical Center Hamburg-Eppendorf between June 2015 and July 2020. Patient selection was done based on the following a priori defined inclusion criteria: (1) age >18 years, (2) multimodal CT (non-contrast CT (NCCT), CT angiography (CTA), and CTP) performed at admission; (3) absence of intracranial hemorrhage (4) endovascular treatment with or without prior intravenous (i.v.) recombinant tissue plasminogen activator (rtPA); and (5) follow-up imaging (NCCT) performed within 24 hours of admission. Patient baseline characteristics were extracted from the medical records and clinical outcome parameters (modified Rankin scale (mRS) at discharge and 90 days, mRS90) were documented, if available. The local ethics review board of the University Medical Center Hamburg-Eppendorf approved the use of anonymized patient data for this retrospective analysis and waived the requirement for informed consent. All study protocols and procedures were conducted in accordance with ethical guidelines of the University Medical Center Hamburg-Eppendorf and in compliance with the Declaration of Helsinki.
Image Acquisition
All patients received multimodal stroke imaging at admission with NCCT, CTA, and CTP
performed in equal order on 256 or 384 dual slice scanners (Philips iCT 256, Siemens Somatom Force). NCCT: 120kV, 280 to 340mA, 5.0 mm slice reconstruction, 1-mm increment; CTA: 100kV, 260 to 300mA, 5.0mm slice reconstruction, 1-mm increment, 80mL highly iodinated contrast medium and 50mL NaCl flush at 4mL/second; CTP: 80kV, 200 to 250mA, 5mm slice reconstruction (maximum 10 mm), slice sampling rate 1.50 s (minimum 1.33 s), scan time 45s (maximum 60s), biphasic injection with 30mL (maximum 40mL) of highly iodinated contrast medium with 350mg iodine/mL (maximum 400mg/mL) injected with at least 4mL/s (maximum 6mL/s) followed by 30mL sodium chloride chaser bolus, whole-brain coverage of 12cm. All perfusion datasets underwent quality control and were excluded in case of severe motion artifacts.
Image Analysis
Anonymized CT imaging data was segmented manually using semiautomatic commercially available software (Analyze 11.0, Biomedical Imaging Resource, Mayo Clinic, Rochester, MN). The raters (RM, SE) were blinded for all other imaging data and patient information. CTP-guided delineation of ischemic lesion NWU on admission imaging was determined according to recently published methods11,18. In brief, the early hypoattenuated ischemic core lesion in NCCT was assessed by densitometric measurements (Dischemic). A similarly drawn region of interest was placed on the contralateral, non-affected hemisphere (Dnormal). Region of interest (ROI) histograms were sampled between 20 and 80 Hounsfield units (HU) to exclude voxels that correspond to cerebrospinal fluid or calcifications. The density measurements, Dinfarct and Dnormal, were then used to calculate percent NWU (%NWU, Equation 1), i.e., the proportion of water uptake within the infarct lesion compared to the contralateral side.
%NWU = (1 – Dischemic/Dnormal) x 100% (1)
Raw perfusion data were analyzed on a SiemensÒworkstation using SyngoÒ VPCT Neuro software (Siemens Healthcare, Erlangen, Germany). Quantitative maps of relative CBF (rCBF), CBV, MTT, and Tmax were generated using a delay-insensitive algorithm. For comparability, we selected a predicted core (pCore) threshold of rCBF≤20%, as this has been described to have the highest level of agreement with RAPID-generated rCBF-based volumes19. Hypoperfusion volume was determined to be the volume of tissue with a prolonged Tmax of at least 6 seconds (Tmax>6s). Mismatch (penumbral) volume was defined as the difference between this volume and pCore.
The final infarct lesion volume (FIV) on follow-up imaging (NCCT) was determined by manual segmentation using semiautomatic commercially available software (Analyze 11.0, Biomedical Imaging Resource, Mayo Clinic, Rochester, MN). Finally, lesion growth between admission and follow-up imaging was determined by calculating the difference between the FIV and pCore (Equation 2).
Lesion growth = FIV – pCore (2)
Statistical Analysis and Outcome Measures
Descriptive analyses were used to define population baseline, imaging, and procedural characteristics. Shapiro-Wilk and histograms tested for the normality of distributions. The primary outcome measure was the binarized core overestimation variable; if lesion growth was negative by more than 10 ml (chosen in accordance with a previous study9 and to account for segmentation error), the pCore was scored as being overestimated. The secondary outcomes were good functional outcome at 90 days, mRS90 0-2, and lesion growth. Because the volumetric measurements of lesion growth were heavily skewed, the cubic root was taken for the analysis. Linear (cubic root of lesion growth) and binary logistic regression models (core overestimation and favorable outcome, mRS90 0-2) were used to test the interactions. Significant predictors of outcome variables in univariable regression were used to construct the multivariable models. A sensitivity analysis was performed following stratification of the cohort according to pCore volume ≥50 mL vs. <50 mL. This cut off was chosen to be in line with the inclusion criteria of multiple prospective clinical trials, as well as observations from the HERMES metaanalysis2,20,21.
All analyses were performed using Stata 15.1 (StataCorp LLC). Normally distributed variables are displayed as mean and standard deviation (SD). Non-normally distributed data are displayed as median and interquartile range (IQR). Categorical variables are reported as proportions. Binary logistic regression results are presented as relative risks (RR) with 95% confidence intervals (95%CI). P-values <0.05 were considered significant.