International Experts Establish FUTURE-AI Guidelines for Trustworthy Healthcare AI

A Global Effort to Shape the Future of AI in Medicine

Artificial Intelligence (AI) offers exciting possibilities to improve healthcare, yet trust remains a challenge due to concerns about safety and ethics. To address these issues, a diverse group of 117 experts from 50 countries collaborated over three years to develop the FUTURE-AI framework. Among them are representatives from leading universities and research institutions, including Imperial College London, the University of Oxford, the Technical University of Munich, Stanford University School of Medicine, Harvard Medical School, Helmholtz Munich and Macquarie University of Sydney.

Recently accepted by BMJ, this consensus paper provides guidelines for the development and deployment of trustworthy AI tools in healthcare. The framework includes best practices and recommendations covering the entire AI lifecycle, from design, development, and validation to regulation, deployment, and monitoring.

Helmholtz Munich’s Role

At Helmholtz Munich, Prof. Julia Schnabel, Director of the Institute of Machine Learning in Biomedical Imaging, and Dr. Georgios Kaissis, Group Leader of the Research Group “Reliable AI”, work on this international project.

"The FUTURE-AI guidelines will by design result in trustworthy, transparent and deployable medical AI tools, thereby providing a competitive advantage for regulatory approval," says Julia Schnabel.

"Using advanced deep learning techniques, we work at the intersection of privacy-preserving artificial intelligence and AI safety. My team and I are developing the next generation of AI algorithms with a focus on data privacy, robustness, and safety for medical applications,” states Georgios Kaissis.

Six Key Principles of FUTURE-AI

In short, the FUTURE-AI framework serves as a code of practice for AI in healthcare, built around six fundamental principles:

• Fairness: AI tools should work equally well for everyone, no matter their age, gender, or background.
• Universality: AI tools should be adaptable to different healthcare systems and settings around the world.
• Traceability: AI tools should be closely monitored to ensure they work as expected and can be fixed if problems arise.
• Usability: AI tools should be easy to use and fit well into the daily routines of doctors and healthcare workers.
• Robustness: AI tools should be trained with real world variations to be robust against real world variations. To remain accurate, the tools should be evaluated and optimized accordingly.
• Explainability: AI tools should be able to explain their decisions clearly so doctors and patients can understand them.