Reach 2025: Novel Networks
By Krista Davidson
Experts from across sectors are joining forces to find responsible AI solutions in health care. Their work could better predict and prevent deadly diseases such as diabetes and heart disease.
Imagine a future where technology helps doctors assess the likelihood of a specific population developing type 2 diabetes, or more quickly identify signs of heart disease in medical images. Advances in artificial intelligence (AI) are steadily enhancing how we approach longstanding challenges in health care, offering new tools to support earlier detection and more personalized care.
Although AI has opened a world of possibilities within health care, much work remains to overcome existing barriers to deployment. CIFAR’s Solution Networks were developed to untangle AI's emerging impacts on society and implement responsible solutions.
Currently, two CIFAR Solution Networks are breaking down barriers and bringing together distinct networks of interdisciplinary experts, researchers and people from the public and private sectors to find answers.
Predicting and preventing type 2 diabetes
According to the National Institutes of Health, type 2 diabetes is one of the fastest-rising chronic health conditions in the world. It’s also seriously expensive – with a projected economic burden of more than $2 trillion (USD) by 2030.
In Canada, the statistics show a worrying upward trend in the rise of diabetes with about $50 million per day spent on health care in Canada to treat diabetes and manage related complications.
These statistics are disheartening, considering type 2 diabetes is preventable, but progress is stalled by a slew of intersecting factors: socioeconomic disparities, poor access to health care, high medication costs and lack of access to healthy and affordable food.
CIFAR’s AI for Diabetes Prediction & Prevention Solution Network is working with partners in Peel Region in Ontario – one of the largest and most diverse municipalities in Canada with a high prevalence of diabetes – to overcome these barriers.
The team is co-designing a socially responsible deployment of a set of already-validated machine learning-based algorithms that can predict, and thereby prevent type 2 diabetes at the population health level. They hope this work will help promote the integration of prevention and management programs to the segments of the population most at risk.
“We know it can be done. We have the technology, but we don’t know how to adapt it into a highly regulated, complex health care system,” says Laura Rosella, a Professor at the Dalla Lana School of Public Health at the University of Toronto and a Canada Research Chair in Population Health Analytics. Rosella co-leads the Solution Network, which consists of six interdisciplinary researchers in areas such as bioethics, computer science, epidemiology and community health.
The team is working directly with providers, public health officials, and other decision-makers to help shape the technology's design. It will combine scientific knowledge of the disease with an AI platform to integrate it at the population level. The team hopes to see their tool implemented by fall 2026.
“The tool bridges the gap for good governance, decision making and a clear systematic process for engaging the right people at the right time when deploying technologies,” explains James Shaw, an Assistant Professor at the University of Toronto and a Canada Research Chair in Responsible Health Innovation.
Given the complexity of the health care system, this needs to be approached carefully.
“A lot of ink has been spilled on the idea of responsible AI,” says Shaw. “It’s easy to think of AI as an artifact, but there’s more to the way we think about deploying and governing these technologies – and that responsibility can and should be embedded throughout the entire process.”
Transforming the future of medical imaging with AI
AI has already transformed medical imaging technology – such as X-rays, MRIs and CT scans – making it easier than ever to detect, prevent and treat the onset of diseases. Yet, in Canada, current restrictions limit the deployment of AI tools at early testing stages – whether for commercial or research models – due to regulatory constraints, confidentiality and safety concerns.
A team of multidisciplinary researchers has come together to form a Solution Network on Integrated AI for Health Imaging. Their new tool, known as the Pictures Archiving Communication System (PACS) - AI, could revolutionize the use of AI for medical imaging.
PACS systems are widely used in hospitals across North America and Europe to store medical images. However, integrating AI models into PACS has proven difficult. PACS-AI addresses this by providing a secure interface for AI models to interact with PACS in real time, enabling testing validation and regulatory approval processes. This accelerates physician access to AI tools, enhances patient care and facilitates impact evaluation, ultimately enhancing diagnostic accuracy and improving clinical workflows.
“Our software is a turnkey solution that seamlessly integrates with a continuously updated repository of imaging data,” explains Robert Avram, Co-Director of the Solution Network. Avram is an interventional cardiologist and an Assistant Clinical Professor at the Faculty of Medicine at Université de Montréal and the Montréal Heart Institute.
“It supports a broad range of modalities – from complex 3D scans and MRIs to standard 2D images like chest X-rays – and facilitates the rapid deployment of advanced computer vision models for enhanced diagnostic support.”
The novel tool has already supported groundbreaking research and garnered positive attention from the medical community, including a publication in the New England Journal of Medicine: Artificial Intelligence. In a study involving over 200 patients in Montréal, San Francisco, and Ottawa, CathEF – an AI algorithm for detecting systolic heart failure in heart attack patients – was successfully integrated into PACS-AI. It proved both fast and accurate, delivering predictions in under five seconds and reliably identifying underlying heart damage.
“With PACS-AI, clinicians will be able to process many modalities and standardize algorithms,” adds Samuel Kadoury, a computer software engineer and a Full Professor at Polytechnique Montréal and researcher at the Centre Hospitalier Université Montréal (CHUM) research centre. “This means that the same prediction AI algorithms can be used irrespective of where images are acquired, or whether it is used for research or in clinical trials.”
The team, which includes experts in cardiology, computer science, health analytics, information technology law, medical imaging and machine learning, is working on both an open- and closed-source platform of PACS-AI. Their interdisciplinary expertise will help translate AI tools into health care and provide other researchers and clinicians with responsible AI mechanisms – for example, a set of fairness metrics that can help physicians assess the data used to train the models – which will help physicians and researchers advance future research.
“Many platforms in existence are closed source and are challenging to modify. We want to support both the research and clinical community. Our aim is to have cutting-edge models that are not yet regulated to be used for experimentation, but we also want to be able to support regulated algorithms that have Health Canada or FDA approval for clinical care,” says Avram.
The goal is to enable physicians with different expertise and specialties, such as radiology and regional oncology, to integrate AI tools into their clinical workflows while also increasing accessibility to advanced medical testing and health care.
At the time of publication, PACS-AI has already been integrated as a research tool at the Montreal Heart Institute, CHUM and the Ottawa Heart Institute, with more to come.
Elissa Strome, Executive Director for the Pan-Canadian AI Strategy at CIFAR, says: “As AI continues to be integrated into our health care systems, the challenge will be to ensure that progress aligns with the values of responsible and ethical innovation. By creating space for collaboration across disciplines and sectors, and including users and patients in the design process, these Networks are shaping a future where leveraging data, AI and real-world expertise, means better care for patients.”
CIFAR’s Solution Networks are setting the stage for a vibrant, equitable and healthy future.
Related Articles
-
CIFAR launches new exploration on “The Future of Food”
July 15, 2025
-
CIFAR Distinguished Fellow W. Ford Doolittle appointed Companion of the Order of Canada
July 08, 2025
-
New and returning AI talent at CIFAR
July 03, 2025
-
In Memoriam: Raymond Laflamme
June 24, 2025