Abstract Number: Esoc2026ys178 Improving Post-Stroke Cognitive Prognostication: Can an Infarct Location-Based Tool Support Clinical Judgment?

Abstract Background and aims Post-stroke cognitive impairment (PSCI) is common after ischemic stroke but difficult to predict early due to variable cognitive trajectories. Timely identification is crucial, particularly for patients discharged home, who often receive limited follow-up. A prediction tool based on acute infarct location (location impact score; https://metavcimap.org/features/software-tools/location-impact-score/) has been validated in a large international dataset, but its added value for PSCI risk assessment in clinical practice remains unknown. This study evaluates whether this tool improves prediction of 6-month PSCI beyond neurologists’ clinical judgment in patients with acute ischemic stroke. Methods In this prospective observational pilot study, 50 patients with acute ischemic stroke will be recruited (2026-2028). Inclusion criteria are being discharged home, ability to undergo MRI and neuropsychological assessment, and an acute infarct visible on MRI. Expected cognitive prognosis (i.e. from unlikely to likely) will be judged by the attending neurologist at discharge, using a 7-point Likert scale. The location impact score will be derived from MRI-based lesion maps. PSCI is defined as impairment on one or more cognitive domains at 6 months. Multivariable logistic regression will assess predictive performance of the location impact score on top of clinical judgment alone, using model calibration and discrimination measures. Results Ethical approval has been obtained, and inclusion will start in February 2026. We are currently designing follow-up studies to build on this pilot. Feedback from the YSPR session could guide subsequent study phases, such as further validation in a larger cohort or use of the prediction tool for patient selection for a cognitive intervention. Conflict of interest Céline Didderen, Joris de Graaf, and Anne Visser-Meily: nothing to disclose. Geert Jan Biessels: supported by Vici Grant 918.16.616 from ZonMW. Nick Weaver: supported by Dekker Clinical Scientist grant 3-005-2024-0111 from Dutch Heart Foundation. Figure 1 - belongs to Methods