Leveraging FHIR and AI for enhanced interoperability in cancer research: a case study of the Cancer Virtual Lab project

Digital health technologies, including EHRs, AI, and data analytics, are essential for improving healthcare delivery and patient outcomes. To address challenges of interoperability and data privacy, common regulations and standards like FHIR (Fast Healthcare Interoperability Resources) have been established to ensure secure and efficient healthcare data exchange. This thesis explores the integration of FHIR and AI techniques to enhance interoperability in oncology research within the Cancer Virtual Lab (CVL) framework. The project successfully standardized clinical data from IRCCS IRST’s electronic health records using FHIR, while integrating medical ontologies to ensure both structural and semantic interoperability. Although manual mapping proved effective, it was resource-intensive, prompting the exploration of AI-based automation. An experiment with a fine-tuned BERT-based model showed promising results in the task of FHIR resource prediction, suggesting AI’s potential in streamlining the FHIR mapping process.