The AI with Purpose series showcases how faculty and staff are thoughtfully using AI to enhance student learning, improve efficiencies and model critically informed use — while clarifying the university’s values, building AI literacy and strengthening U of A leadership in this rapidly evolving space.
AI doesn’t always need to be groundbreaking to make a breakthrough. At the University of Alberta’s Faculty of Medicine & Dentistry, researchers are building on existing artificial intelligence (AI) methods to solve everyday challenges in oral health — from identifying dental anxiety in children to assisting oral pathologists in analyzing cells and predicting jaw development.
Dr. Hollis Lai, professor and associate chair at the Mike Petryk School of Dentistry, says one area of focus within the school’s Oral Health Analytics and Informatics Unit (OHAI-U) is image-based classification. This form of visual AI and machine learning assigns categories or probability scores to images, helping researchers turn visual information into meaningful insights.
One example is a project exploring how AI can help identify, measure and manage dental fear and anxiety in children. The idea addresses a common challenge: young children may struggle to communicate their feelings through traditional Likert scale surveys, which ask respondents to rate their experiences on a scale of five or seven levels.
The project began with Silvia Ortiz, a former master’s student supervised by Hollis, who adapted the Child Drawing: Hospital (CD:H) Scale for use in dental settings. Originally developed and validated to assess the emotional status of school-aged children in hospitals, the scale uses children’s drawings to help evaluate their anxiety.
The process is simple but revealing: children receive a blank sheet of paper and eight crayons in specific colours, with no pencil or eraser, allowing them to freely express themselves. Their drawings are assessed across 23 criteria, each scored from one to 10, with 10 indicating the highest level of anxiety. Criteria include the position and size of the person drawn, as well as the number and predominance of colours used.
Building on Silvia’s work, Hollis’ following master’s student Rojin Adabdokht took the next step: developing and testing machine learning algorithms to automate the anxiety screening process. The goal was to explore how AI could replicate human scoring in a busy dental office, giving dentists insights into their young patients while saving them time.
Hollis clarifies the project focused on comparing AI-generated scores with those scored by human professionals to determine whether the process could be automated. Along the way, the team encountered constraints in training models with a limited sample of images. This led them to consider how generative samples could be used to train detection models — an innovative augmentation approach that helped highlight what made the training data unique.
Currently, a new project leveraging image-based classification has brought together Hollis, oral pathologist Dr. Pallavi Parashar and graduate students, who’re using the Segment Anything Model (SAM) – an AI vision learning model – to identify and quantify cells. The intention is to help oral pathologists analyze samples more efficiently without having to examine details manually under a microscope.
Hollis adds the faculty also has utilized image-based classification to help with more routine dentistry tasks, including training AI to analyze clinical imaging of the palatal suture – the growth line in the roof of the mouth – to estimate whether a patient's jaw is still developing. Other AI applications examine cervical vertebral maturation to predict how much facial development a patient has remaining.
Across these projects, the message is clear: AI’s value isn’t just in inventing something new – it’s in finding practical ways to make existing methods work smarter. But putting AI to work effectively requires more than access to the latest tools. Researchers must first understand the programming and methodology behind them.
“If you don't understand programming, you won't understand AI. That's why we went down the pathway of methodological training,” says Hollis. This methodological training is offered in the Online Graduate Certificate in AI and Health (taught across the College of Health Sciences), where Hollis leads HLTH 503 — The Future of Health: AI Integration and Transformation, a course that includes assisting students in analyzing, applying and integrating AI frameworks.
For flexible, hands-on learning, faculty, staff and students can join the monthly Dental AI Literacy – Virtual Series, which provides interactive discussions, hands-on tool demonstrations and guided literature reviews to explore clinical and research applications of AI in dentistry. The Applied AI in Oral Health Summer Institute also brings together participants from across Canada for AI literacy and methodology training.
“With oral health research, the commercial sector is moving quickly,” says Hollis. “There are tools available, but training is not readily available, and it’s getting harder to adapt to AI research opportunities.”
By combining practical AI applications with accessible, hands-on training, the Faculty of Medicine & Dentistry is helping researchers bridge the gap between emerging technology and everyday clinical challenges — equipping the next generation of oral health leaders to put AI to work where it matters most.