Abstract Number: Esoc2026a423 Hypodensity Derived Automated Non-Contrast Computed Tomography Infarct Core Volume Predicts Poor Functional Outcome

Abstract Background and aims Ischemic injury appears as a hypodensity on non-contrast CT (NCCT). The relative NCCT map (rNCCT) enables automated infarct core segmentation based on relative voxel-wise attenuation differences (Figure1). We evaluated the association between rNCCT-segmented core metrics with poor functional outcome. Methods We retrospectively analyzed baseline NCCT of consecutive stroke patients with an anterior circulation large vessel occlusion, presentation 3 hours from last-seen-well, and pre-stroke modified Rankin Scale (mRS) 3. Infarct core was automatically segmented using rNCCT thresholded at ≥4.9% relative hypodensity. Core volume and mean relative hypodensity were assessed for association with poor functional outcome (90-day mRS ≥3). Results We included 185 patients with median rNCCT core volume of 23 mL (IQR 9–62) and median relative hypodensity within this core of 7.04% (IQR 6.34–8.30; Table1). In univariate analysis, both rNCCT core volume (OR 1.02, 95% CI 1.01–1.03) and hypodensity (OR 1.46, 95% CI 1.17–1.82) were associated with poor outcome (P 0.001). The hypodensity in the core was more pronounced in patients with poor vs good outcome (7.4 vs 6.6%; P 0.001). In multivariable analysis hypodensity was excluded due to multicollinearity, but rNCCT core volume independently predicted poor outcome (OR 1.01, 95% CI 1.00–1.03, P = 0.02; Table2). Conclusions rNCCT-based core volume independently predicted poor functional outcome. The hypodensity within the core was more severe in patients with poor outcome, suggesting prognostic value of hypodensity beyond core volume. These findings support the use of the rNCCT as a standalone tool for core segmentation and outcome prediction in patients admitted 3 hours. Conflict of interest Sources of Funding LV gratefully acknowledges financial support for this publication by the Fulbright U.S. Student Program, which is sponsored by the U.S. Department of State and Fulbright Belgium/Luxembourg/Schuman. Its contents are solely the responsibility of the author and do not necessarily represent the official views of the Fulbright Program, the Government of the United States, or Fulbright Belgium/Luxembourg/Schuman. Furthermore, LV discloses receipt of financial support for the research and authorship of this article by Research Foundation Flanders, PhD fellowship fundamental research 1113925 N; and by the Belgian American Educational Foundation. ATS received grants from the Fulbright Program, the Monahan Foundation, the Philippe Foundation, the France-Stanford Center for Interdisciplinary Studies and the Rotary International. PK receives funds from 2025-26 NIH StrokeNet Training Program (U10NS086487) and has received funds from NIH/NINDS (grant nr R01NS075209). RL is senior clinical investigator (1841923 N) at the Research Foundation Flanders (Fonds Wetenschappelijk Onderzoek, FWO). The CRISP-2 study (PI MGL) was funded by the NIH and by the Research Foundation Flanders (Fonds Wetenschappelijk Onderzoek, FWO), project number G049620N. Disclosures LV reports no disclosures relevant to the manuscript. PK is a cofounder and shareholder of inSteps BV.. JD, SC, ATS, PS, NY, MM, and SK report no disclosures relevant to the manuscript. J.J. Heit reports consulting fees from Medtronic and MicroVention, and is a member of the medical and scientific advisory board for iSchemaView. AW reports consulting fees from Bayer. GWA reports stock holdings in iSchemaView; and compensation from Biogen, iSchemaView and Genentech for consultant services. RL reports institutional fees paid to KU Leuven for consultancy by iSchemaview. Figure 1 - belongs to Background and aims Table 1 - belongs to Results Table 2 - belongs to Conclusions