Approach to Realizing the DHDP Infrastructure
The DHDP harnesses the talent of experts across the country to develop a state-of-the-art software appliance to help unlock data-driven discoveries for cancer and other diseases. This appliance will be installed at participating sites across Canada and contain specific tools that help ensure privacy, accessibility and traceability of data within a federated data governance framework , allowing researchers to build sophisticated machine learning models and learn collectively from data without ever sharing sensitive patient information.
Federated Learning Ecosystem
To minimize the risks to patient privacy, the DHDP uses a federated model to support pan-Canadian research. Under this model, patient data remains at the site where it is generated, and never crosses any institutional, local or provincial borders. Participating healthcare institutions retain control and autonomy over their own data. Only aggregate statistics or AI models are transferred between sites through our software appliance.
Using the DHDP Federated Learning Ecosystem, members will be able to collaborate on clinical and research projects that use diverse types of data including genomics, imaging, administrative, outcome and resource utilization data.
The DHDP’s Privacy-protecting data governance framework and data science technologies will transform collaborative health research. Our data governance framework is built on the principle of Privacy-by-Design, which means that securing patient data and ensuring the interests of the individual are at the centre of everything we do.
International standards will be adopted and implemented in the areas of data, policy and technology. By leveraging international standards and initiatives, many of which have been informed by our own Canadian experts and thought leaders, the DHDP is ensuring high quality, consistency and interoperability. Key standards and initiatives from organizations such as The Global Alliance for Genomics and Health (GA4GH), The American Society for Clinical Oncology (ASCO), Health Level 7 International (FHIR) and The International Cancer Genome Consortium (ICGC), will inform harmonized policies, ontologies, and interoperability. Over time, and through our experience, the DHDP members may also play a role in influencing international standards and initiatives ensuring Canadian needs continue to be met.
The adoption of FAIR Principles allows data to be Findable, Accessible, Interoperable, and Reusable. FAIR Principles will be built into the DHDP data governance framework and data governance technologies to fully realize the potential of our rich data sources. Through its implementation of the FAIR Principles, the DHDP will build a trusted data source by ensuring data is effectively managed and can be appropriately shared and used across the network.
The DHDP’s Technological Infrastructure
The DHDP technological infrastructure includes three components; Data Consumption Networks, an Open Source Data Lake and a Certification Service. When deployed at local sites, these components make up the DHDP Appliance.
Data Consumption Networks
Through the DHDP local appliance, data consumption networks will access and process the sites’ local data. The two data consumption networks are Imagia’s EVIDENS Platform and CanDIG with an intent to scale over time. Imagia’s EVIDENS Platform provides the capabilities to ingest, index and structure static and live clinical data for federated discoveries using AI and radiomics. CanDIG provides analysis of locally-controlled private genomic data. Interoperability mechanisms will be put in place between the two data consumption networks to leverage their respective capabilities.
Open Source Data Lake (OSDL)
The OSDL will collect, aggregate and make available different health and research data sources following FAIR Principles (Findable, Accessible, Interoperable, Reusable) while ensuring strict adherence to the DHDP Certification Service. The OSDL facilitates local aggregation of data that a site is making available to the DHDP. The OSDL will be interoperable with CanDIG and Imagia's EVIDENS Platform.
The DHDP Certification Service, through automated testing, will ensure that implementation of solutions and technologies complies with the DHDP standards and processes in a secure and interoperable manner. This will provide high-level assurance for the required data protection, privacy, and security mandated by the data governance policies of the DHDP.
Researchers will interact with their local appliance of the DHDP to request access to findings derived from the data. In the case of single-site studies, these queries will be processed locally. In the case of approved Ethics Review Board multi-site studies, the queries will be sent out to the networked nodes.