Canada is investing in a national approach to one of healthcare AI's most persistent problems: how to use data from multiple organisations without moving sensitive patient information into a central database. Fourteen projects involving 55 healthcare, research and industry partners have been selected for the country's first Digital Health Innovation Fund. Together, the projects represent C$40.9 million in activity and will use the Digital Health & Discovery Platform (DHDP), a pan-Canadian infrastructure built around federated learning.
Instead of transferring health data between institutions, algorithms can analyse information where it is held and return insights to researchers. The selected projects range from AI-supported cancer detection and personalised diabetes care to mental health, ophthalmology and neurodegenerative disease. The common denominator is therefore less the individual AI application than the infrastructure being created underneath it.
Data stays where it is
Healthcare AI increasingly depends on access to large and diverse datasets, yet clinical information remains distributed across hospitals, research organisations and regional systems. Privacy, governance and data-sovereignty requirements make simply bringing all that information together in one central environment difficult and, in many cases, undesirable. Federated learning offers another model: participating organisations retain their data locally while computational models travel between participating environments, allowing researchers to learn from larger populations without creating a single central repository containing all patient information.
The Canadian programme will apply this approach across organisations throughout the country. Participants include healthcare providers and research centres such as Princess Margaret Cancer Centre and the Centre de recherche du CHU de Québec-Université Laval, alongside universities, technology companies and other partners. Interest was substantial: more than 200 expressions of interest were submitted after the funding programme opened in March 2025, from which fourteen multidisciplinary teams were ultimately selected.
Several projects demonstrate what becomes possible when information can be analysed across institutional boundaries. Teams will investigate AI-powered imaging for real-time cancer detection during surgery, personalised approaches to diabetes and mental health, improved skin cancer diagnostics incorporating Indigenous data and technologies intended to improve eye care in remote and underserved communities. Other work focuses specifically on strengthening Canada's privacy-preserving health-data infrastructure.
Infrastructure before scale
That makes the initiative different from simply funding another collection of unrelated healthcare AI pilots. Canada is attempting to create a common technical foundation on which multiple research and clinical applications can operate. The DHDP itself has received C$49 million in federal funding and is led by the Terry Fox Research Institute, while the new Digital Health Innovation Fund is supported by Innovation, Science and Economic Development Canada.
Whether the individual projects ultimately improve patient outcomes will still require clinical validation. Federated infrastructure also does not remove difficult questions around data quality, governance, interoperability, bias and accountability. It can, however, address one of the structural barriers that appears repeatedly as healthcare organisations try to move AI beyond individual research projects: gaining access to sufficiently large and representative datasets without requiring every participating organisation to surrender control over its data.
That challenge extends well beyond Canada. Europe is confronting similar questions as implementation of the European Health Data Space progresses and healthcare systems seek ways to make data more usable for research, innovation and secondary use while maintaining strong safeguards. The regulatory and technical models differ, but the underlying challenge is increasingly universal.
Canada's programme therefore provides a useful test of a broader proposition. Scaling healthcare AI may depend less on developing yet another model and more on creating the infrastructure that allows existing and future models to learn from healthcare data safely across organisational boundaries. Instead of moving the data, Canada is investing in ways to move the analysis.
Add ICT&health on Google
Show more content from ICT&health in Google Search.
