EUVIMEDEUVIMED entdecken
Crossref

Abstract 1518: Automated Core and Penumbra Estimation in Ischemic Stroke: Accuracy and Eligibility Impact

M. Elsayed, M. Tolba, S. Labony, N. Ali

Stroke: Vascular and Interventional Neurology · 18.10.2026 · 6 (Suppl_1)

DOI: 10.1161/svi270000_1518 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Stroke: Vascular and Interventional Neurology, erschienen am 18.10.2026. Verfasst von M. Elsayed und 3 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Introduction/Purpose: Automated perfusion and non-contrast imaging platforms are now central to selecting patients with acute ischemic stroke for endovascular thrombectomy, particularly in extended and unknown time windows. However, these platforms rely on vendor-specific deconvolution algorithms, thresholds, and artifact-correction methods that are not interchangeable, raising concern about the reliability of automatically generated core and penumbra volumes. We synthesized the published evidence comparing automated imaging outputs with reference standards and quantifying how often automated estimates change eligibility for thrombectomy. Materials/Methods: We reviewed comparative, validation, and outcome studies evaluating automated CT perfusion (CTP), MR perfusion, CT angiography-based, and non-contrast CT deep-learning platforms (including RAPID, Viz.ai, e-CTP/Brainomix, Olea, iStroke, syngo.via, and UGuard) against reference standards of follow-up diffusion-weighted MRI or final infarct volume. Studies reporting volumetric correlation, Dice similarity, intraclass correlation, and downstream effects on DAWN/DEFUSE-3 eligibility or mismatch classification were prioritized, alongside multicenter software-comparison and meta-analytic data.

Results

Across platforms, correlation between automated core volume and reference infarct volume was generally moderate to substantial (Pearson/Spearman r or ICC commonly 0.4-0.85), but spatial overlap was consistently poor, with median Dice coefficients frequently below 0.5, indicating that volumetric agreement does not imply accurate lesion localization. Head-to-head comparisons of commercial packages (e.g., RAPID vs. Viz.ai, RAPID vs. Olea, RAPID vs. e-CTP) showed statistically significant differences in absolute core and penumbra volumes—sometimes exceeding 8-10 mL on average—driven by differing rCBF/Tmax thresholds, deconvolution methods (SVD versus Bayesian), and scanner-specific artifact. Despite these volumetric discrepancies, most multicenter analyses found that final DAWN/DEFUSE-3 thrombectomy eligibility was concordant between software packages in the large majority of cases, with discordance concentrated in a small subset of patients (roughly 5-15%) and disproportionately linked to specific CT scanner models or borderline mismatch profiles. Novel non-contrast CT and CTA-based deep-learning tools approached CTP-level accuracy for core estimation and eligibility classification, suggesting a path toward reducing dependence on perfusion imaging. Systematic review-level data confirmed insufficient high-quality evidence to firmly establish CTP as non-inferior to DWI/FLAIR for infarct core detection, and artifact-related overestimation of hypoperfused tissue was identified in a substantial minority of automated outputs, though this rarely altered final treatment decisions.

Conclusion

Automated core and penumbra imaging tools show reasonable volumetric correlation but limited spatial precision relative to reference imaging, and meaningful vendor-to-vendor variability in absolute volumes exists. Despite this, the clinical impact on thrombectomy eligibility is more limited than volumetric discordance alone would suggest, with most patients classified concordantly across platforms; eligibility-altering discordance clusters around borderline mismatch cases and specific scanner-software combinations. These findings support continued integration of automated imaging with clinical and conventional imaging context rather than isolated threshold-based decision-making, and highlight the need for platform standardization and prospective validation of emerging non-perfusion-based automated core-estimation tools.

Zitieren

APA

Elsayed, M., Tolba, M., Labony, S., & Ali, N. (2026). Abstract 1518: Automated Core and Penumbra Estimation in Ischemic Stroke: Accuracy and Eligibility Impact. Stroke: Vascular and Interventional Neurology, 6(Suppl_1). https://doi.org/10.1161/svi270000_1518

Vancouver

Elsayed M, Tolba M, Labony S, Ali N. Abstract 1518: Automated Core and Penumbra Estimation in Ischemic Stroke: Accuracy and Eligibility Impact. Stroke: Vascular and Interventional Neurology. 2026;6(Suppl_1). doi:10.1161/svi270000_1518

BibTeX
@article{Elsayed2026,
  author = {Elsayed, M. and Tolba, M. and Labony, S. and Ali, N.},
  title = {Abstract 1518: Automated Core and Penumbra Estimation in Ischemic Stroke: Accuracy and Eligibility Impact},
  journal = {Stroke: Vascular and Interventional Neurology},
  year = {2026},
  volume = {6},
  number = {Suppl_1},
  doi = {10.1161/svi270000_1518}
}
Kostenlos starten →