Data and System Traceability for Transparent AI in Medical Imaging

Artificial intelligence holds the promise to revolutionize medical practices, particularly the realm of image-based diagnostics. Nonetheless, the integration of artificial intelligence technologies brings forth a range of immediate and future challenges that are the focus of almost all related discussions. A responsible approach to the development and use of artificial intelligence is essential to effectively address and mitigate these challenges, via strong scientific foundations, technical reliability, thorough testing and validation procedures, risk assessment and alignment with ethical principles. Central to this is the principle of transparency, as a key ingredient to foster trust and reliability. Transparency can be upheld through measures such as disclosing data sources and their use, as well as demonstrating transparent system development, operation and use. In this respect, it is strictly interconnected with the traceability of data and AI systems. This discussion paper briefly outlines the most relevant issues related to transparency and the methods used in the EU H2020 ProCAncer-I project to fulfill its mandates, in terms of data and system traceability, also linked to other projects, such as the Tuscany Region's NAVIGATOR project, and in compliance with the requirements of the FUTURE-AI guidelines.