This project defines new guiding principles named FUTURE-AI and translates these into concrete recommendations and best practices for developing future AI solutions in medical imaging. The proposed guiding principles are (i) Fairness, (ii) Universality, Traceability, Usability, Robustness and Explainability (Figure 1). These principles and the related recommendations and best practices were defined by building on accumulated experiences and results from five large European projects on AI in Health Imaging (the AI4HI Network, comprising the EuCanImage, PRIMAGE, CHAIMELEON, INCISIVE and ProCancer-I projects).
The current recommendations, as detailed in the paper (K.Lekadir et.al), facilitate the application of the FUTURE-AI guiding principles and include a set of 55 checklist points aimed at guiding designers, developers, evaluators, end-users and regulators of AI in medical imaging. In a step-by-step approach, these guidelines will enhance the specification, implementation, evaluation and deployment of imaging AI algorithms that can be trusted, technically, clinically and ethically, within future radiology practice.