Stroke is one of the leading causes of death and disability worldwide, accounting for approximately 7 million deaths annually, with ischemic stroke representing the majority of cases. Severe types of strokes, particularly large vessel occlusions (LVO), are responsible for a significant portion of stroke-related mortality and disability. Effective treatments such as mechanical thrombectomy significantly improve outcomes for LVO patients, but these interventions are only available at comprehensive stroke centers such as university hospitals in Sweden. Early identification of stroke type and severity in the prehospital phase is therefore critical to enable appropriate triage and direct transport to the right level of care. However, ambulance personnel operate under significant constraints, including limited diagnostic tools, time pressure and the critical nature of stroke management. To support effective decision-making under this circumstance, there is a strong need for advanced clinical decision support systems (CDSS) that can integrate heterogeneous data and provide insights in real time.

The Care@Distance research group at Chalmers University of Technology focuses on improving prehospital care pathways through digital technologies, guided by the principle of increasing precision in all clinical decisions. The group pursues a vision of minimizing errors in assessment, prioritization and management, and emphasizes close collaboration between academia, industry and healthcare providers to ensure clinical relevance and usability. Project development follows the Acute Support Assessment and Prioritizing (ASAP) framework, which combines data fusion, machine learning (ML), telemedicine, clinical decision support and innovative user interfaces.

The ASAP Stroke project investigates how artificial intelligence (AI)-based CDSS can support early stroke assessment and triage in prehospital settings through two research tracks. The first track focuses on traditional AI methods using structured data sources such as registry data, vital signs and clinical observations to identify stroke, differentiate LVO strokes and predict optimal care pathways. The second track focuses on video- and image-based analysis, where patient videos recorded in the ambulance are analyzed using ML techniques to extract clinically relevant features associated with neurological deficits and stroke severity. So far, synthetic video data; AI-generated videos; have been used for model development and evaluation to preserve patient confidentiality. Two studies have been conducted focusing on the assessment of eye movement abnormalities and neglect in stroke patients, using datasets generated through 3D modeling and animation. Promising results were achieved, with over 77% recall and 78% precision in classifying stroke-related eye movement and neglect patterns.

By combining these two tracks, ASAP Stroke aims to create a multimodal decision support system capable of fusing diverse data sources to enhance assessment accuracy and support timely triage decisions. The project is designed to extend existing prehospital stroke care innovations such as Video support in PreHospital Care (ViPHS), in which real-time video streamed from the ambulance is used for remote consultation with stroke specialists to conduct neurological assessments. By integrating AI-based video analysis into this workflow, ASAP Stroke has the potential to enhance specialist consultations and provide additional decision support in situations where real-time consultation is not feasible, contributing to better prehospital stroke care.

This presentation will describe the project’s objectives, technical approaches and ongoing work across both tracks. By strengthening early stroke assessment and supporting informed triage decisions, ASAP Stroke seeks to bridge the gap between prehospital and in-hospital stroke care, reduce time-to-treatment and contribute to more efficient, and accurate stroke management.

Authors: Hoor Jalo¹, Bengt Arne Sjöqvist¹, Stefan Candefjord¹

1) Department of Electrical Engineering, Chalmers University of Technology, Gothenburg, Sweden