As part of the EU Horizon 2020-funded project PANCAIM, the team at Karolinska Institutet has built a decision-analytic prediction model to test whether earlier detection of pancreatic cancer with AI is worth the investment. The model evaluates one of PANCAIM AI models, which was developed by Karolinska Institutet in close collaboration with partners at Radboud UMC, Oslo University Hospital and University of Glasgow.
The AI model is trained to flag potential pancreatic cancer on everyday CT scans in clinical routine. Its larger-scale implementation was theoretically evaluated with survival and cost data drawn from Sweden’s nationwide cancer registries, taking into account data from two million patients with up to 60 years of follow-up.
Preliminary results are clear: In the Swedish healthcare system, AI-assisted scans add quality-adjusted life-years and still fall within accepted cost-effectiveness thresholds compared to today’s diagnostic pathway!
This further confirms PANCAIM’s significant achievements in collecting data, exploring the latest algorithms that show level 2 evidence of assistive use benefits. PANCAIM now shows that the AI is not only clinically viable, but also economically sound!
The Karolinska AI model will be released shortly so that researchers, clinicians, and policy-makers can test alternative early detection strategies. PANCAIM is now running this AI in “shadow-mode” deployment in four European hospitals.
Further findings will be presented later this year – stay tuned for details and join us at the workshop on 4th of September to learn more!
