Overview
aICU is a clinically embedded research programme founded at Karolinska Institutet and Stockholm South General Hospital (Södersjukhuset). It develops and evaluates AI-based decision support for intensive care. Work spans retrospective model development, external validation, human-centred design and evaluation, and prospective clinical studies, supported by large-scale longitudinal ICU data and close integration with the clinical environment in Stockholm.
The programme’s first major clinical application is decision support around mechanical ventilation and extubation, including dynamic prediction of ventilation duration and extubation failure. The methods, infrastructure and evaluation framework are designed to support research across other complex and time-sensitive decisions in intensive care.
Cambridge partnership
International collaboration began with the University of Cambridge. ML@CL works with the aICU team on modelling, systems, and human–computer interaction for clinical decision support — including early clinical integration and co-design, and a systems approach to an ecosystem of models that support decision-making rather than standalone tools. On the Karolinska programme page, Cambridge members listed include Christian Cabrera, Diana Robinson, Radzim Sendyka and Carl Henrik Ek.
On the Cambridge side the work connects to the Interfaces programme and is supported by the Accelerate Programme for Scientific Discovery.
Further details are on the Karolinska Institutet aICU programme page.
