Article Text

Download PDFPDF
E-144 Comparative assessment of different CT perfusion software packages in patient selection for thrombectomy
  1. P Golnari,
  2. P Nazari,
  3. M Darwish,
  4. S Ansari,
  5. M Hurley,
  6. A Shaibani,
  7. M Potts,
  8. B Jahromi
  1. Northwestern Memorial Hospital, Chicago, IL


Introduction Recent thrombectomy trials have utilized CTP to help select patients for endovascular intervention. While most trials utilizing RAPID (iSchemaView Inc, Menlo Park, CA), other commercial software packages for CTP analysis are available. In this study, we assessed differences in CTP summary maps between RAPID and another CTP package (Syngo Volume Perfusion CT Neuro), and examined whether these differences would affect patient selection for thrombectomy.

Materials and methods In this retrospective study, patients admitted between 08/16–11/17 with a diagnosis of acute ischemic stroke and occlusion of ICA, M1 and proximal M2, were analyzed. Two different packages were used to analyze CTP data: package A: RAPID (iSchemaView Inc, Menlo Park, CA) and package B: Syngo Volume Perfusion CT Neuro (Siemens Healthcare, Erlangen, Germany). Both packages use a delay-insensitive algorithm. CTP was performed using default threshold values given by the vendor.

Correlations were assessed using spearman’s rank-order. Wilcoxon signed ranks test was used to assess differences of volumes predicted by the two packages. Bland-Altman plot with the respective linear regression line was drawn to show the infarct volumetric agreement between the packages. To see how the packages affected patient selection for thrombectomy, eligibility criteria of the two most recent trials on thrombectomy (DAWN and DEFUSE 3) were used. Agreement between packages is reported as Cohen’s kappa, and diagnostic value of package 2 is reported as sensitivity, specificity and area under curve (AUC).

Results CT perfusion images of 32 patients [mean (SD) age:63.66 (13.43); male/female:21/11; thrombectomy, yes/no:18/14] were analyzed. The mean difference (SD) in the infarct core volume between the two packages was −22.61 (50.85). Higher infarct volumes were predicted by package B [Mean (SD): 61.88 (64.51) vs 39.26 (67.19), p-value=0.032]. This trend was true in both patients without thrombectomy [69.73 (87.69) vs 44.56 (91.84), pvalue=0.084] as well as patients undergone thrombectomy [55.77 (40.23) vs 35.14 (41.71), pvalue=0.157].

There was a strong correlation between the infarct volumes predicted by the two packages (r=0.584, p-value<0.001), although this was stronger in patients without thrombectomy (r=0.640, pvalue=0.014) than in patients undergone thrombectomy (r=0.421, p-value=0.082). On the other hand, there was no significant correlation between penumbra volumes predicted by the two packages.

When using the DAWN trial’s eligibility criteria for thrombectomy, the two packages had a substantiate agreement [Kappa=0.76 (0.131)]. Package B had a sensitivity of 77.78 (95% CI 39.99–97.19), specificity of 95.65 (95% CI 78.05–99.89) and AUC of 90.62 (95% CI 74.98–98.02).

But when using DEFUSE 3 trial’s criteria, the two packages did not have any agreement at all [Kappa=0.185 (0.178)]. Package B had a sensitivity of 70.83 (95% CI 48.91–87.38), specificity of 50 (95% CI 15.70–84.30) and AUC of 65.62 (95% CI 46.81–81.43).

Conclusion Our study shows that despite quantitative differences in CTP parameters derived by the two software packages, there was substantial overall agreement between them when selecting patients for thrombectomy using DAWN (but not DEFUSE-3) criteria. Further work is required to validate these findings across other vendors and in a larger group of patients.

Disclosures P. Golnari: None. P. Nazari: None. M. Darwish: None. S. Ansari: None. M. Hurley: None. A. Shaibani: None. M. Potts: None. B. Jahromi: None.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.