Enabling Privacy-Preserving Analysis of Medical Records from Heterogeneous Data Sources

Despite the improvements in survival rates, neoplasms remain the leading cause of mortality worldwide. Recent technological innovations, particularly in artificial intelligence, are becoming vital in personalized and precision medicine. Indeed, thanks to the increasing availability of medical data, especially from imaging and radiogenomics, artificial intelligence presents opportunities for enhancing diagnosis, therapeutic monitoring, and clinical evaluations of cancer. Data sharing across healthcare centers represents a further potential benefit in this field. However, data privacy regulations, such as the General Data Protection Regulation (GDPR), limit the external processing of patient data. Moreover, data heterogeneity, due to varying formats and standards from multiple sources, complicates integration and analysis. Efforts to standardize data formats and develop interoperability protocols, like HL7 and FHIR, are ongoing but not fully adopted due to their complexity. This paper addresses these challenges and presents a new architecture to improve knowledge sharing across healthcare centers, thereby enhancing the precision and customization of medical treatment. A research project is presented as a case study to illustrate these concepts.