CT Perfusion in Acute Lacunar Stroke: Detection Capabilities Based on Infarct Location

Patient Selection

This study was completed at a level 1 trauma center and primary stroke center. Institutional review board approval was obtained for this study. A retrospective review of all patients who presented with symptoms of acute stroke between January 2006 and July 2011 was performed; 1085 patients with CTP examinations were identified. Patients were included if they underwent acute CTP imaging within 12 hours of symptom onset, as well as DWI within 7 days of presentation. One investigator manually measured the longest axial diameter of foci of restricted diffusion on DWI, if present. Patients were included if they had either 1) negative DWI or 2) one or more infarcts <20 mm in maximum diameter on DWI. The size of 20 mm as the uppermost criterion for lacunar infarcts was chosen based on prior studies implicating this as a clinically applicable measurement.^4,16 Exclusion criteria were: 1) infarct >20 mm in diameter, 2) lacunar stroke with coexisting large (>20 mm maximum diameter) infarct, 3) nonoptimal quality of either CTP or DWI (eg, technical problems such as motion), and 4) administration of intravenous tPA between the time of CTP and DWI acquisitions.

DWI and CTP Acquisition and Postprocessing

MR imaging scans were performed on 2 different scanners (Ingenia 1.5T and Ingenia 3T MR scanners; Philips Medical Systems, Best, the Netherlands). On the 1.5T scanner, the sequence parameters were: TR = 3526, TE = 92, flip angle = 90°, EPI factor = 45, 2.4 acceleration (SENSE) factor, 3 directions of measurement, 16 cm FOV, 131 × 131 matrix, 5-mm section thickness. On the 3T scanner, the sequence parameters were: TR = 3832, TE = 84, flip angle = 90°, EPI factor = 51, 2.4 acceleration (SENSE) factor, 3 directions of measurement, 14 cm FOV, 152 × 121 matrix, 4-mm section thickness. Both scanners used 16-channel head coils. Foci of diffusion were measured in the longest axial axis. DWI b-value was b = 1000 for all studies.

Every patient was imaged as part of the same protocol used for emergent strokes at our institution: an NCCT was immediately followed by multidetector CTA; CTP images were obtained <5 minutes later. For NCCT, FOV was 25 cm, with 512 × 512 matrix and 3-mm sections; for CTA, FOV was 22 cm, with 512 × 512 matrix and 0.9-mm sections. CTA is performed before CTP to minimize venous contamination on CTA scans; in several years' experience with this imaging protocol, the venous contamination noted on CTP images is minimal.¹⁷ CTP images were not obtained if hemorrhage was visualized on the NCCT by the daytime hours staff, on-site radiology resident, or neuroradiology fellow; patients with hemorrhage did not meet the study's inclusion criteria.

A 64-section multidetector scanner was used to perform all CT scans (Brilliance CT; Philips Medical Systems). All CTP sequences were completed by using 2 first-pass acquisitions of sequential 55–60 second dynamic injections/scans (1 image/s) with no delay between the acquisitions. Two acquisitions were completed because each included 40 mm of coverage; hence, 1 acquisition was set at the level of the basal ganglia, and the second was set at the level of the superior margin of the ventricles. According to the manufacturer, 55–60 second dynamic acquisitions were used based on several years' experience before this study of dynamic timing that minimized truncation effect. A truncation is applied to exclude the recirculation for the computation of CBV, with only the first-pass contrast taken into account for CBV. Each acquisition yielded 4 contiguous sections (25 cm FOV, 512 × 512 matrix, 10-mm section thickness, 32 × 1.25 mm collimation, 80 kV, 120 mAs), with a total of 40 mm of coverage. A bolus of 36–40 mL of nonionic contrast (iohexol [Omnipaque 350 mg/mL; GE Healthcare, Piscataway, New Jersey]) was intravenously administered for each acquisition via a power injector at a rate of 6.0 mL/s for a total of 72–80 mL of intravenous contrast for each CTP scan. A saline flush was administered between the acquisition of CTA and CTP imaging, but not between the 2 CTP acquisitions. Typically, the inferior margins of each CTP acquisition were the inferior margin of the basal ganglia and the midcentrum semiovale superiorly. The margins were occasionally adjusted if a suspected infarct was identified on the NCCT by the radiology resident or neuroradiology fellow.

CTP postprocessing was completed on a Vitrea workstation (Vital Images, Minnetonka, Minnesota). CTP postprocessing was automated by a delay-corrected singular value decomposition algorithm; the workstation's postprocessing software uses a deconvolution-based delay-corrected singular value decomposition algorithm via interhemispheric comparison (side-to-side). The interhemispheric comparison performed by the postprocessing software is done by using a multistep approach: 1) the midline of the brain is automatically located (but can be manually corrected, if necessary), 2) the affected side is automatically identified by using TTP maps, and 3) MTT, TTP, CBV, and CBF in the affected side are automatically compared with the corresponding parameters in the contralateral side by using relative differences between each voxel. Arterial input functions were selected from the anterior cerebral artery, and venous functions were selected from the posterior segment of the superior sagittal sinus; automatic selection was used, which was verified before the finalization of images.

Image Review and Interpretation

Three staff neuroradiologists (2 with >10 years' experience, 1 with 1.5 years' experience), who had been blinded to the final results, independently reviewed the DWI, NCCT, CTA, and CTP images. Each reviewer was given the patient NIHSS score and symptoms at the time of presentation; the reviewers were otherwise blinded to the patient interpretative reports and the other reviewers' interpretations. The neuroradiologists assessed for the presence of stroke by using 4 different independent CTP parameters: CBF, CBV, MTT, and TTP. Presence or absence of stroke was then assessed on all imaging sequences; NCCT, CTA, and CTP images were reviewed first, and DWI was reviewed afterward to blind the review of CTA and perfusion data to the DWI results. Stroke locations were divided into the subcortical white matter with or without involvement of the overlying cortex (CSWM), periventricular white matter (PVWM), basal ganglia and/or thalamus, cerebellum, and brain stem. Brain stem and cerebellar infarcts (n = 9) were ultimately excluded because the CTP margins did not extend inferiorly enough to encompass strokes in those locations; 1 infarct in the PVWM was also excluded because the CTP images did not fully include the infarct. Interpretations were considered correct if 2 or more reviewers agreed on the results.

NCCT/CTA and CTP findings were subsequently compared with DWI. Using DWI as the “criterion standard,” findings were denoted as true-positive results if: 1) the abnormalities on NCCT/CTA corresponded with the same areas as the infarcts on DWI or 2) regional abnormalities on CTP included the area of lacunar infarction on DWI. Abnormalities on NCCT/CTA and CTP were marked as false-positives if: 1) no infarct was present on DWI or 2) the infarct on DWI was in a different location than that of the CTP. A single neuroradiologist graded the amount of chronic small vessel ischemic disease based on a grading scale: 1 (none-minimal), 2 (mild-moderate), 3 (moderate), 4 (moderate-severe), and 5 (severe).

Statistical Analysis

Intra- and interobserver agreement was calculated by using the Cohen κ. The sensitivity and specificity of the NCCT/CTA and independent CTP parameters (CBF, CBV, MTT, and TTP) in each area of the brain (basal ganglia and/or thalamus, CSWM, PVWM) were calculated. A P value of <.05 was considered statistically significant. Statistical analysis was performed by using Statistical Analysis Software (SAS; SAS Institute, Cary, North Carolina). Charts were created in Microsoft PowerPoint Version 14.6 (Microsoft, Redmond, Washington).