The potential for using AI applications in clinical healthcare was identified early on and is now becoming an established field of research. The number of publications on medical AI increased fivefold from 2017 to 2022, with nearly 70,000 articles registered in PubMed.
Despite this intense research activity, few AI innovations make the leap from research project to clinical application and are put into use in hospitals. Successfully establishing a clinical AI application is complex and involves many diverse steps.
The first step is the development of the AI model, which requires both clinical and technical expertise to identify a relevant problem and research question, as well as access to the necessary data. The next stage is to take the AI model into a self-developed pilot project and then further through the CE marking process to scale up for broader use.
To successfully achieve this, the following are required:
- A clear need and benefit within clinical operations
- Close collaboration with the hospital’s IT department
- Legal expertise with experience in MDR and CE marking
- A link to commercial actors
CAIR-Lab was started in 2021, but our experience in AI research dates back to 2014 when we began working with AI interpretation of orthopaedic X-ray images and were the first in the world to publish in this area in 2017. Our first pilot project with AI-assisted X-ray image interpretation was launched in 2019. Since then, we have run several projects in various medical fields. CAIR-Lab is part of Clinicum at Karolinska Institutet and is based at Danderyd Hospital and KI DS.