Healthcare Systems

Objectives

Define the key elements required to develop and implement multimodal foundational AI models that are robust and accountable, i.e., feasible, efficient, equitable and relevant from the point of view of intended users and the universal healthcare system.

• Identify promising use cases to test multimodal foundation models for the universal healthcare system in Quebec at different points along the care continuum.
• Set up the governance and infrastructure needed to implement robust and responsible multimodal foundation models in a real-world context, regardless of the data types and needs the models aim to address.
• Co-construct a quality assessment framework to evaluate these infrastructures, models and their added value from the perspective of users, organizations and the healthcare system.
• Analyze the factors and mechanisms that promote and limit the development and implementation of these algorithms at organizational and systemic levels, in order to develop guidelines and tools for decision-makers and managers.

Research Axes

Use cases are at the heart of this regroupement’s research plan, and will focus on the following key aspects:

• Identification of the gaps between the promises of AI and the reality on the field;
• Identification of barriers and risks;
• Identification of promising strategies for robust and responsible implementation;
• Integration of an interdisciplinary team (including: user, developer, decision-maker, facilitator).
• Development of frameworks and guidelines through expert consensus and consultation with the ecosystem.

The use case selection process is currently underway, and R8 members have been invited to submit their use case scenarios. Selection will be based on the following criteria:

• Alignment with the vision, research objectives and approach proposed by the R8 co-leaders;
• Structuring effect for the Quebec healthcare system;
• Feasibility and anticipated impact;
• Diversity of projects in the care continuum, domains and target populations;
• Potential for collaboration and synergies with other IVADO R³AI regroupements.

Challenges

The main challenge is to overcome the gap between the promises of AI and the reality of its use in the Quebec healthcare system. There is a potential to improve prevention, early detection and treatment of diseases, as well as to improve the efficiency of the healthcare system. Experts emphasize that the specific challenges of implementing AI solutions in the healthcare system are notably due to the nature of the data, the nature of the work and the cognitive workflows involved, and the particular context of professional organizations in a fragmented healthcare system.

While the implementation of multimodal foundation models is promising, the novelty of the approaches and the scarcity of evidence on its concrete use in healthcare and the associated risks create several unknowns.  Finally, access to the health data needed to develop and implement these models will have to comply with current laws and ethical standards, and patient buy-in will have to be obtained.

Anticipated Impact

• Improved clinical decision-making and quality of care, owing to more accurate diagnosis, more personalized care and more effective treatment.
• Optimized organization of healthcare services: automation and workflow optimization will result in increased efficiency and reduced operational costs, while freeing clinicians from low value-added clerical tasks.
• Strengthening health surveillance and promotion: the implementation of these models will facilitate early detection of diseases and the protection of public health.
• Capacity building and talent attraction: organizational and systematic capacity building to guide robust and responsible AI integration, while fostering the development of a dynamic and innovative AI ecosystem in healthcare.
• Reducing health inequalities by ensuring equitable access to cutting-edge technologies for the entire population, with a particular focus on under-represented or vulnerable groups.

Researchers

• Hassane Alami – Université de Montréal
• Carl-Éric Aubin – Polytechnique Montréal et CHU Sainte-Justine
• Robert Avram – Université de Montréal et Institut de Cardiologie de Montréal
• Danilo Bzdok – Université McGill
• Julien Cohen-Adad – Polytechnique Montréal et CHU Sainte-Justine
• Jean-Louis Denis – Université de Toronto
• Philippe Després – Université Laval
• Guillaume Dumas – Université de Montréal et CHU Sainte-Justine
• Julie Hussin – Université de Montréal et Institut de cardiologie de Montréal
• Marc-André Legault – Université de Montréal
• Cristina Longo – Université de Montréal
• Venkata Manem – Université Laval et CHU de Québec
• Erica Moodie – Université McGill
• Bouchra Nasri – Université de Montréal
• Jean Noel Nikiema – Université de Montréal
• Esli Osmanlliu – Université McGill et Centre universitaire de santé McGill
• Samira Rahimi – Université McGill
• Catherine Régis – Université de Montréal
• Louis-Martin Rousseau – Polytechnique Montréal
• Angel Ruiz – Université Laval
• Mireille Schnitzer – Université de Montréal
• Pablo Valdes Donoso – Université de Montréal

Research Advisor

Audrée Janelle-Montcalm: audree.janelle-montcalm@ivado.ca