The aICU project: Developing next-generation decision support tools for intensive care
aICU is a clinically embedded research initiative focused on developing and evaluating AI for intensive care. We aim to create robust, clinically relevant decision-support tools and to study how they can be tested, integrated, and used safely in real-world healthcare settings.
About the project
This project aims to develop artificial intelligence (AI) tools to support clinical decision-making in intensive care. The focus is on making better use of the large volumes of data generated in ICU environments to improve decision support for healthcare professionals.
Intensive care units treat the most critically ill patients, where decisions must often be made rapidly under uncertainty. At the same time, vast amounts of data—such as monitoring signals, laboratory tests and treatment information—are generated but not fully utilised in clinical practice.
Aims and objectives
The overall aim is to develop AI-based models that can support clinicians in complex decision-making situations. Examples include:
- predicting ICU length of stay
- identifying risk of clinical deterioration
- supporting treatment decisions, such as when to safely discontinue mechanical ventilation
A key focus is also to ensure that the developed systems are safe, interpretable and clinically meaningful.
Research approach
The project, part of the aICU programme at Karolinska Institutet, integrates clinical research with AI development. Key components include:
- establishing secure infrastructure for large-scale data analysis
- building high-resolution ICU datasets
- linking national health registries for long-term outcome studies
- developing systems for real-time data use and clinical testing
The work is conducted in collaboration with the University of Cambridge AI Laboratory and involves multidisciplinary teams.
Challenges and potential
Despite rapid advances, relatively few AI models are currently used in routine ICU care, with many remaining at the research stage.
This project addresses that gap by focusing on the full pipeline—from development to evaluation and implementation in real clinical settings.
The goal is for AI to act as a supportive tool for clinicians, contributing to more informed decisions, better use of resources, and improved patient outcomes.