ONCOlogy-targeted NLP-powered Federated Hyper-archItecture and Data Sharing

ONCO-FIRE proposes to build a novel hyper-architecture and a common data model (CDM) for oncology, as well as a rich, modular toolset enabling significantly increased interoperability, exploitability, use and reuse of diverse, multi-modal health data available in electronic Health Records (EHR) and cancer big data repositories to the benefit of health professionals, healthcare providers and researchers; this will eventually lead to more efficient and cost-effective health care procedures and workflows that support improved care delivery to cancer patients encompassing support for cancer early prediction, diagnosis, and follow-up. The applicability, usefulness and usability of the proposed hyper-architecture, CDM and toolset for oncology and the high exploitability of health data will be demonstrated in diverse data exploitation scenarios related to breast and prostate cancer involving a number of Virtual Assistants (VAs) and advanced services offering to health care professionals (HCPs), hospital administration/healthcare providers and researchers data-driven decision-support and easy navigation across large amounts of cancer-related information. Through the above mentioned outcomes and the (meta)data interoperability achieved, ONCO-FIRE contributes to the exploitation of large volumes, highly heterogeneous (meta)data in EHR and data repositories including imaging data, structured data (e.g. demographics, laboratory, pathological data), as well as diverse formats of unstructured clinical reports and notes (e.g. text, pdf), including (but not limited to) temporal information related to the patient care pathway and genomics data currently "hidden" in unstructured medical reports, and more. Importantly, ONCO-FIRE interconnects, following a federated approach, large, distributed cancer imaging repositories, currently used for AI tools training and validation, with patient registries and EHRs of cancer-related data and supports exploitation of relevant unstructured data through novel Natural Language Processing (NLP) tools. The ultimate goal is to establish a patient-centric, federated multi-source and interoperable data-sharing ecosystem, where healthcare providers, clinical experts, citizens and researchers contribute, access and reuse multimodal health data, thereby making a significant contribution to the creation of the European Health Data Space.