Funded through the AI Knowledge Translation Fund, established with the generous support of donors Joseph and Amy Ip, this program supports innovative research initiatives that strengthen the Faculty of Medicine’s capacity in artificial intelligence. The 2025 competition will support five $10K projects driving AI innovation, including: transforming ultrasound for early fetal health insights, bringing reliable imaging to underserved communities, empowering patients with personalized perioperative education, advancing predictive models for eating disorders through synthetic data, and redefining gait analysis with smartphone-based technology. These projects were reviewed, ranked, and awarded through a competitive process that emphasized research excellence, innovation, knowledge translation and/or end-user engagement, and the potential to address significant real-world challenges with meaningful impact. As part of the application process, applicants provided confirmation of completed SGBA+ (Sex and Gender Based Analysis) or IDEAS (inclusivity, diversity, equity, accessibility, and social justice) training, reflecting the Faculty’s commitment to these principles.
2025 AI Seed Funding Awardees
Dr. Steven Hawken (School of Epidemiology and Public Health; OHRI)
Ultrasound-based Self-Supervised Foundation Models for AI-Assisted Detection of Fetal Anomalies
We aim to develop and refine foundation models to enhance AI-guided fetal anomaly detection from obstetrical ultrasound images. Our team has developed USF-MAE (Ultrasound Self-Supervised Foundation Model with Masked Autoencoding), an ultrasound foundation model that dramatically improves the efficiency and accuracy of AI model training for fetal anomaly detection. This proposal focuses on developing additional enhanced foundation models to support more targeted AI-guided classification tasks, improving performance and clinical applicability.
Dr. Mohamed Hefny (Department of Radiology; OHRI)
Physics-Inspired AI for Organ and Tissue Detection in Ultrasound to Support Health in Rural and Underserved Communities
This project will create and test an AI model that can identify different types of human tissue in ultrasound images. Ultrasound is a safe, low-cost, and widely available imaging tool, but its interpretation can be challenging without specialist expertise. This has critical implications for rural and underserved communities, where access to advanced imaging and trained specialists is often limited. Our approach uses physics-inspired AI, incorporating knowledge of how ultrasound waves behave in tissue to produce results that are more reliable and explainable. The project will also strengthen the Faculty’s AI capacity through development, evaluation, and dissemination of a pilot demonstration.
Dr. Arnaud Mbadjeu Hondjeu (Department of Anesthesiology and Pain Medicine; OHRI)
From Patient Need to Scalable Solution: Implementing Large Language Models for Perioperative Education
Many patients struggle to understand preoperative instructions, leading to anxiety, surgical cancellations, and delayed recovery—particularly among equity-deserving groups. This project implements an AI-powered, bilingual platform delivering personalized, real-time perioperative education, co-designed with clinicians and patient partners. Piloted at The Ottawa Hospital, the platform adapts to patients’ literacy, language, and emotional needs. The project aims to improve comprehension, reduce cancellations, enhance patient outcomes, and support equitable care, with findings informing broader scale-up.
Dr. Nicole Obeid (Department of Psychiatry; CHEO RI)
Evaluation of Synthetic Data Generation to Enhance Predictive Modelling of Eating Disorder Outcomes
Eating disorders are deadly yet underfunded psychiatric illnesses, with research constrained by small sample sizes, limited biopsychosocial data, and underrepresentation of males and gender-diverse populations. Current “one-size-fits-all” treatments fail to account for the complex factors shaping clinical outcomes. This project leverages precision medicine data to evaluate synthetic data generation methods designed to expand sample size and demographic diversity, improving the performance and fairness of machine-learning models predicting eating disorder outcomes.
Dr. Albert Tu (Department of Surgery; CHEO RI)
Evaluating and Improving an Artificial Intelligence–Augmented Automated Observational Gait Analysis Pipeline with Direct User Feedback
This project builds on the Movealytics Motion Lab, which uses consumer smartphones and AI to generate automated, clinically validated assessments of limb function from video data. Using established visual scoring systems, current algorithms achieve over 90% concordance with expert assessment. The project will validate and refine the application through real-world user feedback, iteratively improving usability, accuracy, and end-user satisfaction.
