ABSTRACT NUMBER: ESOC2026YS131 FACING UP TO STROKE: ANALYSING FACIAL EXPRESSION USING DIGITAL TECHNOLOGIES POST-STROKE

Abstract Background and aims Rapid and accurate assessment is critical in acute stroke care to enable timely treatment and optimise patient outcomes. CognoStroke is a digital, automated cognitive assessment tool that records audio and video while patients respond to brief memory and mood questions, allowing standardised assessment in hospital and home settings. This study aims to investigate whether digitally derived measures of mood and facial behaviours can be used as early, objective biomarkers of post-stroke cognitive and affective changes, with potential to support clinical decision-making across the acute stroke pathway. Methods This is a single centre, prospective study recruiting participants from the hyper-acute stroke unit, acute stroke unit, community rehabilitation centre, outpatient stroke and TIA clinic, and Join Dementia Research. During CognoStroke assessments, participants’ speech and facial video data are recorded. Facial recordings are analysed using an automated computer vision framework to quantify facial movements associated with six basic emotional expressions. Results To date, 107 stroke patients, 12 stroke mimics, and 266 healthy volunteers have been recruited. Stroke patients show higher GAD7 and PHQ9 scores compared to healthy volunteers, suggesting depression and anxiety may influence responses during cognitive assessment. Preliminary emotional expression and facial movement analysis identifies clear and systematic differences between the three groups. These early findings suggest that facial movement markers derived from brief remote assessments may support differentiation between stroke and stroke mimics, while also capturing temporal changes of post-stroke affective and cognitive processing. Ongoing work will expand the sample size and evaluate the robustness and clinical utility of these facial movement markers. Conflict of interest Isabela Ramnarine: Nothing to disclose