Led by Centre Léon Bérard (CLB) and the Health Data Hub in France, this oncology-focused use case is dedicated to validating AI models for detecting metastasis in lung scans and identifying nodules in mammograms. By establishing a standardized cohort of patient images, SHAIPED provides an environment for validating AI models in cancer imaging, supporting both healthcare providers and AI developers. This approach helps AI developers rigorously test their models under real-world conditions and regulatory requirements, aiding compliance and market readiness. Additionally, healthcare organizations benefit from a consistent evaluation method to select the most effective AI tools available on the market.
To accomplish this, SHAIPED will collaborate with a consortium of healthcare and AI experts across France, the Netherlands, and Italy. Participating cancer centers will manually annotate patient images to create a high-quality dataset for model validation. Approximately 1,000 images per site, covering both lung scans and mammograms, will be collected across Europe. HDABs will run validation tests in secure environments, using 20% of the data for developer testing and 80% for independent evaluation. This standardized approach allows SHAIPED to rigorously test the effectiveness of AI tools for cancer imaging, ensuring they meet clinical, regulatory, and operational standards.