A 21st century platform at the intersection of healthcare and artificial intelligence

We have gathered below a list of Frequently Asked Questions about the DHDP. Feel free to also use the search bar at the top of the page, to search for keywords throughout the DHDP Web page.

Learn more about:

 

General DHDP Questions

What is the scope of data you are wishing to bring into this DHDP? Can you provide an overview of the process envisioned (e.g.: managed by TFRI, Imagia or both)?

Each data provider remains the custodian of their own data, determining how and for what purpose the data will be used - (i.e. for data insights, AI model creations, specific studies), based on research ethical committee approvals. Data type support will be added onto the platform based on project selection and availability into standard HL7 or PACS format (i.e. imaging, textual reports, genomic panel, genomic data, virtual pathology, treatment, lab results, outcome).

What safeguard mechanisms are in place for data traceability and data privacy?

Federal and provincial privacy laws safeguard the privacy, security, and authorized access to information. The objective for DHDP is to deploy state-of-the-art solutions with a high level of assurance for the required data protection, privacy, and security.  A certification service will test and ensure that local and third-party implementation of solutions and technologies comply with the DHDP standards and processes in a secure and interoperable manner.

What role are pharma, biotech, medical device and IT companies expected to play in the DHDP?

The role of pharma, biotech, medical device and IT companies is critical to the success of the DHDP as the anticipated research discoveries will require vehicles for commercialization. Industry players can participate in a variety of ways including providing in-kind resources like data as well as cash-contributions in the form of research sponsorships.

How will interested parties be able to tap into the platform for research and commercialization activities?

To assure the highest quality of science and engagement, the DHDP will issue a Request For Applications (RFA) as part of its Science Stack that will be scored based on the combined weighting of the clinical motivation, quality of science, value of the data, the strength of the team, as well as, their collaborative use of the DHDP and commercial potential.

Will DHDP membership expand?

The DHDP is intended to be an inclusive consortium. As the DHDP develops it will look to expand its membership. Initially, the DHDP priority will be to on-board the members who were part of the original application. All partners will be required to contribute actively to the design of the DHDP or to actively influence its roadmap to uphold their membership status. We encourage all members, in particular those who provided Letters of Support without financial or other in-kind commitments, to begin thinking about ways to transition to an active DHDP partner. This can be in the form of research sponsorship, cash contributions, in-kind contributions, data in the network, or other value creators that have yet to be defined.

One of the main objectives of the DHDP is to create and grow the network by connecting cross-disciplinary experts across the country and further accelerate innovation and discovery. 

To become a member, please contact dhdp@dhdp.ca

What are the DHDP strategies?

The DHDP strategies include the following: data, commercialization, education and training, eligible project selection, network and sustainability, technology, membership, project selection, and commercialization. We are currently forming committees and working groups to develop these strategies. Other committees may be formed based on needs, as they are identified.

Will there be an opportunity to use the Digital Platform in other disease areas aside from oncology and the neurosciences (Alzheimer’s)?

Initially, DHDP support, IT resources and funding will be prioritized to focus on oncology and the neurosciences (Alzheimer’s), two therapeutic areas with a high burden of disease and significant unmet needs.

We anticipate entering into other disease areas as the technology matures and the DHDP infrastructure and membership grows. Imagia’s AI platform EvidensTM is indication agnostic and translates to all disease areas involving clinical data to enable AI-driven discovery at scale.

What types of projects does the DHDP fund?

The Digital Health and Discovery Platform (DHDP) seeks to improve health outcomes for Canadians by accelerating precision medicine for cancer and neurological diseases such as Alzheimer’s. To do this, the DHDP funds groundbreaking projects that:

  • Link health data to enable innovation across multiple sites

  • Find novel ways to incorporate AI to clinical settings

  • Mentor a new generation of scientists and clinicians able to problem-solve in complex environments

Learn more about our funding programs and competitions here

How is the DHDP funded?

The Government of Canada has invested $49M to help create the DHDP as part of  the Ministry of Innovation, Science and Economic Development (ISED) Strategic Innovation Fund Stream 4 - Health and Biosciences (SIF-4) competition. Additional commitments from our members bring our total budget to over $200M.

What is the DHDP?

The DHDP is a pan-Canadian, digital health data initiative co-led by the Terry Fox Research Institute and Imagia. Its digital platform connects different sources of data to accelerate research and stimulate commercialization of Canadian research discoveries. The DHDP is governed by a 16-member Executive Committee representing a broad range of sectors of the knowledge economy. The Platform consists of a distributed hardware and open-source and proprietary software which will be used under appropriate licences.  Healthcare data will be added to DHDP as part of a distributed digital trust where each contributing member remains the custodian of their respective data

DHDP Infrastructure

How are you planning on accessing the data at the sites? Are you expecting challenges in deploying DHDP across the sites as they will be using different recording systems and languages?

A DHDP Appliance will be installed in participating sites across Canada. This appliance is a digital infrastructure and will contain common DHDP tools including the certification service that will protect patient privacy and data security. Through this local appliance, data consumption networks - Imagia’s EVIDENS Platform and CanDIG - will access and process the sites’ local data. The DHDP has committed to providing an open-source data lake technology (OSDL) that can facilitate local data aggregation (static data or live clinical data). Independently, Imagia’s EVIDENS Platform provides the capabilities to ingest, index and structure static and live clinical data. As we work through the integration of the foundational sites, the DHDP will build an integration playbook that future sites can refer to in preparation for integration of the DHDP. Interoperability mechanisms will be put in place between the two data consumption networks to leverage their respective capabilities.

A DHDP Integration Working Group was formed in late 2020. It includes key individuals at the foundational sites who can socialize the concept of the DHDP and facilitate its integration. The researchers in the Marathon of Hope Cancer Centre Network have already identified multiple cancer cohorts in each of the foundational sites that can be contributed to the DHDP. Similarly, EVIDENS members are already engaging into collaborative multi-institutional research projects through Imagia’s EVIDENS Platform. The DHDP submission to ISED also had the support of researchers in the field of neurodegenerative diseases. Success stories from the federation of these early cohorts will help build additional interest and willingness to contribute data in additional healthcare domains.

What about patient consent? In particular the aggregation of live clinical data vs structure cohorts specifically designed for research? Can this approach be used with retrospective patient data?

Using the DHDP Federated Learning ecosystem, members will be able to collaborate on projects that use both live clinical data (Real-World Data, using the data consumption network Imagia’s EVIDENS Platform) and structured cohort data (such as in the context of the Marathon of Hope Cancer Centers Network - MOHCCN). The MOHCCN data will be an example of deep -omics data from a diverse cancer population that will be leveraged for new discoveries. Similarly, EVIDENS members are already engaging in collaborative multi-institutional research projects through Imagia’s EVIDENS Platform (e.g., including -omics related to immunotherapy). The DHDP will explore different access models. One potential model is Registered Access to data with patient consent for broad secondary use. The other model will be through the traditional Research Ethics Board (REB) approval process to enable discoveries on retrospective data. The DHDP Certification Service will ensure that all data that is contributed to a project has appropriate approval and/or consent in place to allow for the specific use of the data. We are also designing mechanisms to allow patients to have access to information about how their data is being used within the DHDP.

Is it possible to get data across the provinces? What are the processes for investigators to follow to get access to the data?

It is precisely the issue raised by inter-provincial data-sharing that led the DHDP to adopt a federated model to enable pan-Canadian research. Under this model, patient-level data (including personal information) will remain at the site where it is generated, and will therefore not cross any borders, institutional, local or provincial. Under the federated model, only aggregate statistics or model parameters will be transferred between sites, which minimizes the risk to patient privacy.

In order to participate in a project, investigators affiliated with a DHDP Member Institution will need to follow a registration process to be certified to use the DHDP, and agree to abide by DHDP Policies. This process is currently under development.

How far along is this DHDP technology at the moment - when will clinical researchers be able to utilize this developing network?

Both data consumption networks EVIDENS Platform and CanDIG, are established. We are currently building out the DHDP certification service and defining the standards that will enable the operationalization of open innovation across DHDP Member institutions. Timelines will be communicated once development plans have matured further.

Available Resources

Will DHDP be charging researchers for the compute time they use on the nodes?

So far, granting agencies and foundations have provided support to purchase the initial computing infrastructure. The DHDP Sustainability Committee will be responsible for developing strategies to support the goals and sustainability of the DHDP (including maintenance and expansion of the computing nodes). The committee will be responsible for developing a Sustainability Plan for five (5) years of Network operation after Strategic Innovation Fund funding ends with the intention to grow the platform.

What computing power resources are available through the DHDP?

The DHDP budget includes some support for selected sites to deploy infrastructure that will support the DHDP. The planned infrastructure currently consists of CPU and GPU nodes at participating sites, computational requirements will be refined as the development plan for the DHDP technology matures. We will also leverage complementary federal and/or provincial research infrastructure initiatives to further support for existing or additional sites.

Interoperability

What about companies with AI technologies for precision medicine interested in joining the DHDP?

The DHDP is intended to be an inclusive consortium and we encourage participation of both public and private entities. The role of pharma, biotech, medical device and IT companies is critical to the success of the DHDP as the anticipated research discoveries will require vehicles for commercialization. Industry players can participate in a variety of ways including providing in-kind resources like data as well as cash-contributions in the form of industry-sponsored research. Multi-nationals may also consider participation in the pre-competitive space.

Is there a longer-term plan for interoperability with international datasets (e.g., via GA4GH principles and tools)? Is there a plan to introduce a data access application system using GA4GH Passport or a similar approach?

The DHDP Technology Committee will strongly recommend the use of standards developed by GA4GH, as they are state of the art in genomics and data sharing. CanDIG, one of the two identified DHDP data consumption networks, is a GA4GH driver project. This will help us ensure efficient and secure collaboration both across Canada and internationally. Other initiatives to create data interoperability standards in other domains will be taken into consideration by the DHDP Data and Technology committees.

Role of Private Entities

Is the data to be accessed mostly for academic research? Or can for-profit startups working in ML use this data to create models?

For-profit startups as well as other for-profit entities are encouraged to take part in the DHDP. We are pursuing two complementary approaches to DHDP science: (1) Augmenting Research Through Disruptive Technology using DHDP such as:

  • Advances in Cancer Therapy – Personalizing Immunotherapies for Each Cancer Patient
  • Advances in Neurodegenerative Disease – Improving Early Diagnosis and Management

The RFP will encourage multi-stakeholder participation. The project selection will be done through a request for proposal (RFP) process that will be scored based on the combined weighting of the clinical motivation, quality of science, value of the data, and the strength of the team, as well as their collaborative use of the DHDP and commercial potential. Multi-institution collaborations will be preferred, and all proposals will also be evaluated by their engagement in the use of AI methods. We anticipate the launch of the first RFA competition towards the end of Q4, 2021. The date is subject to change given the current pandemic. (2) Building Cross-Functional Training to Enhance the Knowledge Economy with the introduction of the DHDP Catalyst Program to be launched in Spring 2021. Support for several projects to be offered with co-funders such as MITACS and others.


What about companies with AI technologies for precision medicine interested in joining the DHDP?

The DHDP is intended to be an inclusive consortium and we encourage participation of both public and private entities. The role of pharma, biotech, medical device and IT companies is critical to the success of the DHDP as the anticipated research discoveries will require vehicles for commercialization. Industry players can participate in a variety of ways including providing in-kind resources like data as well as cash-contributions in the form of industry-sponsored research. Multi-nationals may also consider participation in the pre-competitive space.

Intellectual Property

Who owns the IP on models developed using federated data?

The Commercialization Committee is tasked with developing a strategy that articulates clear policies on the creation, use and protection of any IP to maximize benefits to Canada. The policy shall further demonstrate value creation/retention, accessibility, and transparency by delineating appropriate licensing and royalty structures for different types of entities and project types (e.g., research projects vs. commercialization projects). We are relying on current mechanisms in place at each institution that have proven effective to manage licenses, royalties and commercialization, including but not limited to approvals by ethics review boards and offices of technology transfers. Underlying IP agreements will need to be in place before projects are funded. One such mechanism in place is the Imagia’s EVIDENS agreements organizing all aspects of data use, IP management and commercialization.

What about companies with AI technologies for precision medicine interested in joining the DHDP?

The DHDP is intended to be an inclusive consortium and we encourage participation of both public and private entities. The role of pharma, biotech, medical device and IT companies is critical to the success of the DHDP as the anticipated research discoveries will require vehicles for commercialization. Industry players can participate in a variety of ways including providing in-kind resources like data as well as cash-contributions in the form of industry-sponsored research. Multi-nationals may also consider participation in the pre-competitive space.

Data sharing incentives

It is well documented that well-informed patients readily give their consent to data sharing for research purposes. What is not well understood is the willingness of hospitals to share data. Any comment on this?

As custodians of patient information, healthcare institutions have the responsibility to ensure patient privacy and data security is protected. In the context of research projects, Institutional Ethics Review Boards are involved in ensuring that the data used for research has appropriate consent, and that the use of this data is reasonable. Complexities arise when data consented for research leads to IP generation then managed outside the scope of ERB, by offices of technology transfer. The DHDP’s approach to address this challenge is to design and implement a federated data governance framework. This framework will ensure the security and traceability of all of the data within the ecosystem, and is built on three key principles: Autonomy, Diversity and Privacy by Design. Custodians will always retain control, and autonomy over their own data, which will never leave their local servers. For instance, custodians can decide which data to use for which projects. Through the platform, clinicians and researchers will have access to easy to use, analytic tools that have been developed by data scientists and AI experts, allowing them to generate robust clinical evidence by leveraging diverse datasets from across multiple sites, going beyond what is available to them locally. By taking a privacy by design approach, patient data and other sensitive data will be respected through all aspects of the initiative: The DHDP federated data governance framework allows experts to build robust sophisticated machine learning models, and learn collectively from the data without ever sharing any sensitive private information.

How can the DHDP help policy makers draft good policies on responsible clinical data sharing, and data access? Do you anticipate this will lead to a culture-change on these issues?

The DHDP is founded on three pillars: Policy, Technology, and Science. It is designed to create world-leading policy and technology stacks while being driven by expert-led use cases from Canada’s outstanding science community. The key principles we are abiding to, like data privacy and FAIR data (findable, Accessible, Interoperable and Reusable) will be implemented by design in the platform. Further innovations in policy, including advancing enabling technologies, and harmonizing policies between nodes within the Network will be needed to realize the full vision of the Network. The learnings and outputs will be a key driver in policy discussions by showcasing how multi-institutional collaborative projects can lead to groundbreaking science, stimulate commercialization and ultimately improve patient outcomes. It is our hope that this leads to an openness to responsible clinical data access.

When the TFRI MoH project finishes, what incentives will there be for health care institutions to entrust data to the DHDP appliance?

With respect to incentives in general, although the DHDP IP Policy framework has yet to be finalized, an option that will be considered is revenue flow back to the sites that provide data and results in a commercial product, aligned with the Imagia’s EVIDENS agreements. Furthermore, the DHDP provides a platform on which technologies that automate data curation can be developed and proven out - one such example is the Imagia’s EVIDENS Platform that facilitates an increasing set of features relating to AI-assisted real-world data curation capabilities. These technologies could potentially reduce the cost of data curation and collection in the future. The DHDP will facilitate multi-centric collaboration by deploying appliances that safely ingest live clinical data, which does not entail any preparation of the data, beyond establishing an interconnection with the appliance. 

Education and Training

Getting the mindset alignment in the academic centers is key. Is the DHDP team working on education materials on this topic? Do you see ways we can have the technology educate?

One of the key objectives of the DHDP is to mentor a new generation of interdisciplinary Canadian scientists & clinicians in healthcare AI, who are able to solve problems in complex environments. Working with partners, the Education and Training Committee will utilize DHDP to enrich and innovate training of highly qualified personnel (STEM and healthcare professionals). The Education and Training Committee will be charged with developing strategies to design innovative interdisciplinary training programs in the application of cutting-edge technologies (AI and experimental) applied to healthcare, short-term internships and inter-laboratory training visits, skills upgrading, mentorship programs and making recommendations on educational and training material. The overall objective being to produce skilled STEM workers and healthcare professionals and attract and retain top talent to grow Canada’s knowledge economy.

What about companies with AI technologies for precision medicine interested in joining the DHDP?

The DHDP is intended to be an inclusive consortium and we encourage participation of both public and private entities. The role of pharma, biotech, medical device and IT companies is critical to the success of the DHDP as the anticipated research discoveries will require vehicles for commercialization. Industry players can participate in a variety of ways including providing in-kind resources like data as well as cash-contributions in the form of industry-sponsored research. Multi-nationals may also consider participation in the pre-competitive space.

Marathon of Hope Cancer Centres Network (MOHCCN)

Can you discuss the relationship between the MOHCCN and the DHDP?

The MOHCCN is concentrating on generating and organizing different types of experimental digital health data. The DHDP will provide the distributed technology platform and tools required to house, aggregate and analyze cancer case data from the MOHCCN, in an entirely complementary manner. The DHDP will integrate AI approaches, and will support validation exercises to prove the utility of the Platform. Cancer centers will have the opportunity to become members of the DHDP.

Will the data generated from the Marathon of Hope studies be available to all DHDP members to do their own research on?

In principle yes! To be designated as a Marathon of Hope Cancer Centre will require centers to make their MOHCCN data available for analysis under the FAIR principles. In practice, dashboard-level data will be available to all DHDP members. Research that requires access to controlled or complete datasets will require approval from the Data Access Committee (DAC) which will require Research Ethics Board (REB) approval.

How are complete cases defined for inclusion into the 15.000 cases?

A complete case for MOHCCN will consist of clinical, omic, imaging, treatment and outcome data and available biospecimens.

What is the procedure to be part of the MOHCCN?

Cancer Centers need to be invited to undergo designation to be part of the MOHCCN. Designation will require an application to be submitted by a cancer center, and an international peer review process to evaluate the capabilities for the center and its contribution to the Network. There will be at least two rounds of designation. The first round is currently underway with applications invited from BC Cancer, Princess Margaret Cancer Centre and the Montreal Cancer Consortium institutions. Learnings from the first round will be applied to the second round of designations later in 2021/2022. Organizations that are not cancer centers can also be included in designation, but will need to collaborate within consortia that include centers that treat cancer patients. Please visit www.marathonofhopecancercentres.ca for information related to the Marathon of Hope.

Imagia's EVIDENS Platform

Who gets to see the data: clinician only or patient and clinician?

Within an institution, researchers and clinicians under the appropriate Ethics Review Board approvals are allowed to work on the data. As part of the DHDP, ultimately, we want patients to have access to information about how their data is used, and to have the possibility to give or revoke consent depending on data uses. At the moment we are working within the realm of clinical research processes and the EVIDENS Platform facilitates data access and cross-institutional data analysis via a federated learning platform.

Will clinicians have the ability to provide feedback to change search fields/query terms as they start using? How will this feedback be incorporated?

Yes. Within EVIDENS there are various levels of administrative controls, and feedback is collected at various steps in the process. Within a single institute, collaborators working together on a project go through the annotation process and can leave comments to each other. For instance, a medical resident is able to flag an annotation for review and approval by a senior doctor. This commenting and approval capability allows us to ensure annotations (segmentations, localizations) that are submitted are appropriate, and quality controlled. Across different sites, when a query is shared, it needs to get adapted to the local setting. For instance, text can be recorded in English or French, depending on the institution, and elements don’t always translate 1 to 1, requiring a certain level of data reprocessing or NLP to perform ontology matching. In general, differences between institutional datasets may require clinicians to make adjustments to enable cross-site data analysis. 

The explainability must be by design at the outset ... how do you approach the capture of metadata/specific domain knowledge after the fact?

Domain specific knowledge is gained through the data maturation process. By using technologies like AI-assisted human-in-the-loop annotation, specialists help provide the necessary context as part of the process. This supports continuous knowledge capture throughout the ecosystem. 

Do researchers and clinicians want to see which feature(s) of the image was the most important for the AI algorithm to make a decision?

Yes. Explainability and understanding of what is activated in the model is imperative. This feature is not currently available on the EVIDENS Platform. In healthcare, understanding what the features mean, what the model is looking at, and breaking open this blackbox is a key necessity both during research and clinical practice.