Everything brought up is of crucial importance, if this exciting technology is to truly benefit medical care and patients. Not every question found an answer, but the ball is now rolling and future workshops will continue on from here.
Bengt Arne Sjöqvist, Project Manager of PICTA, kicked off the workshop by welcoming participants and presenting Lindholmen Science Park and PICTA. He also mentioned the newly publicized national AI venture, to be located at Lindholmen, in which health and medical care are some of the areas of application that have been identified.
Stefan Candefjord from Chalmers gave an introduction to AI, and Johanna Bergman of PICTA presented the background to the workshop. The work on PreSISe (Prehospital Support for Identifying Sepsis Risk), the new PICTA project, has clearly highlighted several of the general and basic questions that will arise when the AI technology is rolled out on a broad front and is uses in practice in medical care — in this case, prehospital care. These questions formed the basis of the workshop.
There is no doubt about it — things are starting to happen in AI in medical care. As far as image analysis and automatic triage are concerned, there are both Swedish and international examples where AI solutions have been introduced or are close to being introduced into clinical reality. On a more visionary level, ideas such as “AI is my doctor” and “50% self-care by 2030” were presented. The majority were in agreement that development must take place gradually, and near-future solutions dealt with AI as support for existing medical care personnel, for example, in the form of decision-making support.
Mats Artursson, investigator at the Swedish Medical Products Agency, gave a presentation on what regulations say about CE labeling of AI, what that is in Sweden and internationally, and the work under way in the EU. Some of the subjects he addressed were that regulations have a difficult time keeping up with rapid technological advances, and the challenge in handling systems that are so complex that it is no longer possible to explain in detail how they work. A recurring theme was the issue of responsibility: who is responsible for medical decisions and who is responsible for quality assurance of data? Clinical tests and the dilemma between demands for traceability and anonymization of data were also discussed.
After a much-needed coffee break, it was time for the next area of focus: how AI-based solutions will benefit medical care and how medical knowledge is to be disseminated when it is built into the system. Johanna Bergman gave a presentation on how thoughts are running in the PreSISe project, and emphasized the importance of handling these types of solutions nationally and on a joint basis. Douglas Spangler, Project Manager at Uppsala University Hospital, spoke about open source and what role this working method could play in making AI solutions accessible in medical care. Transparency and openness were two key concepts.
Subsequent discussions dealt with ownership and administration of data and algorithms, and what the business model for this type of system should look like. Models in which medical care is responsible for both data and algorithms were discussed. The role of business in a model of this type could be to develop efficient working methods with software, visualization, user interfaces, consulting services and training rather than providing algorithms.
The participants certainly reflected the broad interest in AI. Representatives from medical care, academia and the private sector were on site — all with the ambition of finding shared paths toward how AI-based solutions could circulate and be of real benefit in medical care.
To sum up, it was a successful workshop and a good start to PICTA’s new focus group initiative. There will be more to come!
Meeting notes will be distributed to participants.