Abstract Background and aims Stroke remains a leading cause of disability and mortality worldwide. Despite successful mechanical thrombectomy (MT), up to 40–50% of patients do not achieve good functional recovery. Accurate early prediction of functional outcome, measured by the modified Rankin Scale (mRS), is therefore essential. The RCP-SEVILLA project (Revival and Consolidation of a database Platform - Stroke Evaluation Via Intelligent Learning for Life-Saving Algorithms) aims to validate a machine learning model to identify patients with good functional prognosis (mRS 0–2) 90 days after stroke. This project is supported by the FIS2026 grant (N°PI25/01957). Methods We developed a supervised classification model using real-world data from the ARTISTA registry (Seville–Huelva Thrombectomy Network), including 3,703 patients treated since 2017. Functional outcome at 90 days was dichotomised as good (mRS 0–2) versus poor (mRS 3–6). Model performance was assessed using ROC analysis in training, internal validation, and external validation cohorts, reporting AUC values. Results The final model was based on logistic regression using 7 clinically relevant variables (age, admission mRS, ASPECTS, NIHSS at 24h, TICA, TSR and blood glucose), achieving an accuracy of 0.82 and an AUC of 0.897 in internal validation. External validation was performed across four independent spanish hospitals, including a total of 2895 samples and, yielding an AUC of 0.911, demonstrating robust generalisability and supporting its potential clinical utility. Future work will explore more advanced AI methods, including neural networks, and extend prediction to three outcome categories (mRS 0–2, 3–5, and 6). Conflict of interest Victor Toscano-Duran: nothing to disclose; Pablo Baena Palomino: nothing to disclose; Asier de Albóniga-Chindurza: nothing to disclose; Marta Aguilar: nothing to disclose; Elena Zapata Arriaza: nothing to disclose; Henry Andrade: nothing to disclose; Angela Gonzalez-Diaz:nothing to disclose; Nazaret Nieto: nothing to disclose; Alejandro González: nothing to disclose.
Toscano-Durán et al. (Fri,) studied this question.