CAIR Lab

Clinical Artificial Intelligence Laboratory (CAIR-Lab) is a world unique institution, an innovation laboratory dedicated to developing clinical applications based on artificial intelligence (AI).

Foto: Pixabay.
Photo: Pixabay

Clinical Artificial Intelligence Research Laboratory – CAIR-Lab 

Clinical Artificial Intelligence Laboratory (CAIR-Lab) is a world unique institution, an innovation laboratory dedicated to developing clinical applications based on artificial intelligence (AI). CAIR-Lab will be located at Danderyd University Hospital (DUH) and will be in a close collaboration with Karolinska Institutet, the Institution for Clinical Sciences at Danderyd Hospital (KI DS). Our vision is to create an environment that combines competence in medical science, technical development, and innovation to aid clinicians to develop products that will be used in the health care of the future. 

Background 

AI with the intention for clinical use is a recent concept, and despite progress in AI use in a research setting, few AI products are available to be used in healthcare. CAIR-Lab gained the experience of developing an idea into a clinically implemented product when we developed an AI application for evaluation of skeletal radiographs that was launched for clinical use at DUH. Through this we identified key elements that are needed to develop clinically relevant AI: 

  1. Description of the clinical problem 
  2. Relevant patient data 
  3. Technical knowledge 

Worldwide, there are few institutions that have access to all three elements. With our gained knowledge and experience we hope that CAIR-Lab will enable that more AI innovations can benefit patients. The laboratory will be an environment open to researchers interested in developing clinical AI innovations, regardless of medical specialty. 

Strategy 

The heart of CAIR-Lab is the competence and equipment to develop AI models and technical infrastructure (named CORE). CORE will be available to all CAIR-Lab projects, that are divided into two categories: 

  1. Incubators: The projects/ideas that need to develop an AI motor that can be evaluated in the clinical setting. 
  2. Accelerators: The projects that has graduated from an incubator (successfully launched as a clinical pilot project) can continue as an accelerator project with the aim of being further developed into an AI innovation that will be used in clinical healthcare. 

The incubator environment will act as an early support for projects with the aim to start a pilot project and developing a “minimal viable product” (MVP). One example of our planned projects is to create an application that will use AI to predict myocardial infarction by analyzing atherosclerotic plaques on coronary angiography x-rays. The incubator support will consist of collecting radiographs from the hospital’s picture archiving and communication system (PACS), selection and development of suitable algorithms and development of MVP, that will be evaluated in a controlled and limited clinical setting. This environment is a close collaboration between CORE and the hospital environment. 

The accelerator environment will act as a springboard for successful incubator projects with the aim of creating a commercially viable product. The MVP will be further developed and later tested in a larger scale both at DUH and at collaboration partners, such as other hospitals and quality registries. The accelerator will also reach out to commercial partners and other incubators. 

Early prototypes 

In CAIR-lab, we intend to quickly develop prototypes that can be safely tested in clinical environments to find out which ideas and algorithms work. The location of CAIR-Lab at DUH will enable us to quickly sift out the best ideas as the hospital is one of Sweden's largest with a focus on common conditions. The environment gives us unique conditions for finding the areas where AI has the greatest opportunity to influence and develop healthcare. 

Contact

E mail to CAIR-lab: cair@kids.ki.se  

Max Gordon

Affiliated to research

Martin Magnéli

Affiliated to research