Abstract Number: Esoc2026a1182 Predicting Reduced Ejection Fraction in Acute Ischemic Stroke Patients: A Prediction Model From the Mind the Heart Study

Abstract Background and aims Cardiac CT is increasingly used in the acute diagnostic work-up of acute ischemic stroke (AIS), but commonly used scan protocols do not allow assessment of left ventricular ejection fraction (LVEF), a known predictor of functional outcome. We aimed to develop a prediction model for reduced LVEF using clinical and imaging data. Methods This post-hoc analysis of the Mind the Heart study included AIS patients who underwent acute cardiac CT (acquired during diastole) and routine diagnostic work-up, including transthoracic echocardiography (TTE), which served as the reference standard for LVEF. Multivariable logistic regression models incorporating baseline variables, arterial input function (AIF) derived from CT-perfusion, and left atrial and ventricular cardiac CT volumes were developed using the training set (80%) to predict reduced LVEF (50%). Performance was evaluated on the testing set (20%) using receiver operating characteristic analysis, reporting the area under the curve (AUC), sensitivity, and specificity. Non-linear associations were modeled using restricted cubic splines, and Least Absolute Shrinkage and Selection Operator regression was used for variable selection. We evaluated whether cardiac CT information improved model fit. Results We included 298 AIS patients (61% male, median age 71 IQR 61-79 years), of whom 87 (29%) had reduced LVEF. The final model showed moderate discrimination (AUC 0.71), sensitivity (0.67) and specificity (0.70). Including cardiac CT volumes in addition to baseline and AIF variables improved model fit (p=0.04). Conclusions Our model may aid risk stratification for reduced LVEF in AIS patients undergoing acute cardiac CT and could support identification of patients in whom additional TTE is warranted. Conflict of interest MMW received the Arthur Fonville Award for Stroke Research in 2025 for this work. SSN received travel grants from the Cultuurfondsbeurs, Remmert Adriaan Laan Fonds, CONTRAST consortium and Dr. Jan Meerwaldt Stichting outside the submitted work, JMC reports grants from Medtronic, Siemens, AstraZeneca and Bayer outside the submitted work (all paid to institution). JMC is shareholder and co-founder of TrianecT. CBM reports grants from Stryker and Boeringher-Ingelheim outside the submitted work (paid to institution) and is shareholder of Nicolab (minority interest). LAR was supported by a personal Dekker Junior Clinical Scientist Grant from the Dutch Heart Foundation.The other authors have no financial conflicts of interest.