Organizing an AI innovation challenge within the European digital innovation hub in Croatia, 2025

Abstract Issue/problem A key barrier to implementing artificial intelligence (AI) in healthcare is the lack of robust and secure infrastructure for developing, testing, and validating innovative AI-based solutions. Without structured environments, it is difficult to translate AI advances into clinical practice, especially in areas such as breast cancer diagnostics. This infrastructure gap hinders the creation of clinically relevant and safe AI tools, particularly in resource-limited settings, and risks the deployment of poorly validated or non-compliant solutions. Description of the problem To address this, the Croatian Institute of Public Health (CIPH) participated in the organization of an innovation challenge within the AI4Health.Cro European Digital Innovation Hub. The challenge targeted breast cancer diagnostics, tasking teams with developing AI models for BI-RADS prediction and lesion localization, ensuring model interpretability, and designing a user interface for clinical use. The project leveraged real, anonymized mammography data and was structured in two phases: initial model development with open data, followed by refinement within a Secure Processing Environment (SPE). The initiative included clear protocols for data security, anonymization, and compliance with GDPR, and fostered collaboration among multiple institutions. Results Approximately 2 terabytes of mammographic images and reports were processed within the SPE. Radiologists contributed annotations to enhance dataset quality for model development. Fourteen teams created AI model prototypes, achieving promising preliminary results. Lessons The project demonstrated that secure, interoperable infrastructure and clear data governance are essential for responsible AI innovation in healthcare. Structured, multi-institutional collaboration and early clinical involvement proved critical for developing trustworthy, clinically relevant AI tools. Key messages • Robust infrastructure enables secure data handling and reliable computing, ensuring safe, effective AI innovation that advances healthcare diagnostics and improves patient outcomes. • Cross-sector partnerships between healthcare, academia and tech innovators, coupled with secure data protocols, drive responsible AI innovation and real-world clinical integration.