Two authors (R.M. and M.P.) received a research grant from Toshiba Medical Systems Corporation (TMSC, Japan). Toshiba Medical Systems Corporation did not have any influence on the concept of the One-Step-Stroke protocol, the execution of this study, the analysis of the data, nor on the writing of this manuscript.
This retrospective study was approved by the ethics committee of our institution, and informed consent was waived. Initially, 34 consecutive subjects were included who underwent CTP scanning at the emergency department of our hospital. Inclusion criteria were: patients with clinical symptoms of acute ischemic stroke, with onset of symptoms within 9 hours, without a history of kidney failure and a minimum age of 18 years. Exclusion criteria were: non-standard CTP acquisition protocol (4), severe patient movement artefacts in CTP (2), incidental finding of a tumour lesion (1), late or poor contrast enhancement (3), intra-arterial contrast injection (1), the presence of a drainage tube (1), clipped or coiled cerebral aneurysm (2). The remaining 20 patients consisted of eight male and 12 female patients (mean age 65 years, median 66 years and age range 36 - 93 years).
In eight out of 20 patients, signs of acute ischemic stroke were seen on the NCCT, CTA and/or CTP images by the attending neuroradiologist.
CT imaging was performed on a 320-row CT scanner (Toshiba, Aquilion ONE, Toshiba Medical Systems Corporation, TMSC, Otawara, Japan). The scan protocol consisted of a cerebral NCCT, cerebral CTP, and head and neck CTA. In all patients two contrast injections were performed, one for CTP and one for the CTA. Only the CTP acquisitions were used in the present study.
For CTP, 50 mL nonionic contrast agent (300 mg iodine/mL Xenetix 300, Guerbet, Villepinte, France) was injected into an antecubital vein with an injection rate of 5 mL/s followed by a 40 mL saline flush at 5 mL/s. Whole brain volumetric acquisitions with 16 cm z-coverage were acquired with 0.5 mm slice thickness, 0.5 s rotation time, and 80 kV tube voltage. A cerebral CTP protocol was used, which started 5 s after contrast agent injection with the first volumetric acquisition at 200 mAs, followed after 4 s by 13 scans at 100 mAs with a 2 s interval, followed after 5 s by five scans at 75 mAs with a 5 s interval. The total number of scans was 19 and total scan duration was 60 s (Fig. 1). Image reconstruction was done using a smooth convolution kernel FC41 and standard AIDR3D (adaptive iterative dose reduction in three-dimensions, TMSC).
A publicly available software program, Perfusion Mismatch Analyzer (PMA) developed by the Acute Stroke Imaging Standardization Group (ASIST), version 22.214.171.124, was used to calculate perfusion maps of cerebral blood flow (CBF), cerebral blood volume (CBV), and mean transit time (MTT). The software automatically selects ten arterial input functions (AIFs) in one slice, which was set at the level of the Circle of Willis. The venous output function (VOF) was automatically chosen in the intracranial veins above the skull base. The perfusion maps were calculated with a delay-insensitive deconvolution algorithm, the block-circulant Singular Value Decomposition (bSVD) . Calculations were performed on a 256 × 256 matrix with smoothing enabled, on 5 mm slabs of the original CTP data; other parameters of PMA were kept at default values.
Simulating the one-step-stroke protocol
The One-Step-Stroke protocol was simulated by eliminating specific acquisitions from a CTP sequence in order to study the optimal time gaps between subsequent acquisitions, similar to a previous study published in the literature . The perfusion values derived from the original sequence served as the reference standard. One observer experienced in CTP analyses (M.O.) selected a region of interest (ROI) in the (unaffected) proximal M1-portion of the middle cerebral artery (M1-MCA) to visually determine the bolus arrival time (BAT). For every patient one volumetric acquisition was deleted, starting from the BAT up to the fifth time point after the arterial peak (see Fig. 1). Because the number of volumetric acquisitions between these two markers may differ per patient, the number of deleted time points and, therefore, the number of simulated One-Step-Stroke protocols differ per patient. Perfusion maps were calculated for the original CTP acquisition and for each One-Step-Stroke protocol. The locations of the automatically chosen AIF and VOF of these simulated One-Step-Stroke protocols were compared to the locations in the original protocol and manually corrected if they were not at identical locations. The ROI in the MCA was also used to estimate arterial enhancement to estimate the enhancement of the carotid arteries.
Perfusion values in normal-appearing white matter (WM) and normal-appearing gray matter in the basal ganglia (GM) were estimated by drawing multiple ROIs (M.O.) and averaging perfusion values across these ROIs for each patient and tissue type. White matter ROI’s included the centrum semiovale and cortical spinal tract. Subcortical gray matter ROI’s included the caudate nucleus, putamen and globus pallidus. The size and location of the ROIs were kept constant within the patient, while the size of the ROI’s differed between patients (freehand ROI’s were applied). The total number of voxels included was 6,922 ± 3,674 (mean ± standard deviation) for NAWM and 547 ± 244 for NABG.
For each skipped time point, percentage errors of the perfusion values were calculated per tissue type and patient. The percentage error was calculated by taking the difference of the perfusion values between the original CTP protocol and the simulated One-Step-Stroke protocol divided by the perfusion value of the original CTP protocol multiplied by 100. Since these percentage errors can be positive or negative, we used the absolute percentage error |% error| for further analysis.
First, to estimate the magnitude of the absolute percentage errors, we determined the mean and standard deviation of all patients and all simulated One-Step-Stroke protocols, for CBV, CBF, and MTT in WM and GM. Thus, in total six mean absolute percentage errors and corresponding standard deviations were calculated. Furthermore, we report the maximum absolute percentage errors.
Next, we determined the optimal timing of the neck CTA. Since we aimed to evaluate which time point could be deleted without having a major influence on any of the three perfusion parameters, for each patient we selected the maximum percentage error across the three perfusion parameters CBV, CBF, and MTT in WM and GM per deleted time point. Thus, per patient and per simulated One-Step-Stroke protocol (i.e., per deleted time point), one maximum percentage error is selected for further analyses.
In order to determine the optimal timing of the neck CTA, we simulated bolus tracking by determining the first time point T0 in which the enhancement in the M1-MCA exceeded a given relative threshold value. Relative thresholds, i.e., enhancement above the baseline pre-contrast scan, were varied between 40 HU and 100 HU in steps of 10 HU. Every deleted time point was reported in seconds relative to T0, for up to 10 s after T0.
Therefore, each threshold defines a T0 (which may differ per patient in absolute time) and each simulated One-Step-Stroke protocol results in a maximum percentage error per patient. The mean (maximum percentage error) of all patients was then calculated. We calculated the mean maximum percentage error as a function of the relative threshold (a trigger for bolus tracking) and as a function of the deleted time points from T0.
To estimate how often the selected timing would lead to larger errors, we reported the number of patients in whom the absolute percentage error exceeded 10 % in at least one perfusion parameter. The time point and threshold with the lowest average maximum percentage error and the lowest number of patients with more than 10 % errors in combination with the highest enhancement in the MCA (which served as proxy for carotid enhancement) was chosen as the optimal timing.
Finally, after determining the optimal time point for neck CTA the analysis was repeated but with two time points deleted to simulate a 6 s gap at the optimal timing only.
Visual assessment of the perfusion maps
A subsequent observer study was performed in order to evaluate the influence of skipping two CT perfusion time points on the qualitative evaluation of perfusion maps. An experienced observer (F.J.A.M.) evaluated the original perfusion maps and perfusion maps with a 6 s time gap randomly. The CT perfusion maps were scored for: 1) The presence or absence of a perfusion deficit. 2) Absence or presence of infarct core. 3) Absence or presence of penumbra. and 4) Size of infarct core or penumbra. The sizes were described as either small or large relative to the vascular territory affected. Infarct core was defined by a perfusion deficit with increased MTT, decreased CBF, and decreased CBV. Penumbra was defined as the presence of a perfusion deficit in relation with normal or slight elevated CBV. The observer was blinded to perfusion maps shown (original or with a 6 s gap), clinical information and diagnoses.
Statistical analyses were performed using the Statistical Package of Social Sciences version 20.0 for Windows (SPSS Inc., Chicago, USA). A Wilcoxon signed rank test was used to show significant differences between the original CTP protocol and the One-Step-Stroke protocol at the optimal time point with 4 s and 6 s gaps. A P value < 0.05 was considered significant. By examining the slope of a linear fit intersecting the origin, the linear relationship between the original CTP protocol and the One-Step-Stroke protocol with 4 s and 6 s gaps were assessed. A linear fit of 1 was considered ideal. Spearman correlation coefficients were reported. Bland-Altman analyses were performed to compare the original CTP and the simulated One-Step-Stroke with 4 s and 6 s time gaps.