Get to know T-CAIREM 2025 Trainee Rounds Winner Yuxi Long

The T-CAIREM 2025 Trainee Rounds once again showcased the next generation of innovation at the intersection of AI and health. Every year, these rounds bring together top trainees from across Canada to share their research with peers and leaders in the field.

This year’s winner, Western University PhD candidate Yuxi Long, presented her project “Pre-trained Vision Transformers Enable Robust Thermal Imaging-Based Detection of Rheumatoid Arthritis.” Her research offers a fresh perspective on diagnosing rheumatoid arthritis (RA), a condition that affects roughly 1% of the global population and can lead to permanent joint damage.

We spoke with Yuxi about the inspiration behind her work, the promise of AI-assisted thermal imaging, and her hopes for the future of patient care.

Yuxi’s motivation to focus on rheumatoid arthritis is deeply rooted in its widespread impact and the limitations of current diagnostic approaches. “RA is a chronic disease that can cause irreversible joint damage if not caught early,” she explained. “Traditional diagnostic methods, like blood tests, X-rays, or MRIs, are invasive, costly, and sometimes miss early signs of inflammation. Thermal imaging, on the other hand, is a low-cost, non-invasive tool that can capture subtle heat patterns related to joint inflammation.”

This blend of accessibility and clinical relevance made thermal imaging an ideal candidate for enhancement through AI. Yuxi’s work explores how machine learning could refine the tool’s accuracy, making it a more reliable option for early RA detection.

Her research leverages pre-trained vision transformers—a type of deep learning model originally designed to analyze large-scale image data. “In essence, we take thermal images of patients’ hands and feet, which show heat patterns linked to inflammation. Instead of training a new model from scratch, we use pre-trained vision transformers that have already learned from massive datasets. These models act like expert feature extractors, identifying meaningful patterns in the thermal images. We then feed these patterns into classifiers that distinguish between healthy individuals and patients with RA.”

This approach is especially powerful because it overcomes one of the biggest challenges in healthcare: limited datasets. Building robust models typically requires tens of thousands of examples, which is rare in medical research. By using pre-trained models such as DINOv2 and SigLIP, Yuxi was able to tap into existing knowledge while also employing fusion strategies and dimensionality reduction techniques to boost accuracy and interpretability.

For Yuxi, the most exciting part of her work is its real-world potential. “This research could make RA detection more accessible,” she shared. “For patients, that means earlier, non-invasive diagnosis without the need for costly scans or specialist visits. For clinicians, it provides a reliable support tool to improve diagnostic accuracy and efficiency. Ultimately, it could reduce delays in treatment and prevent long-term disability.”

That patient-centred perspective is what grounds her research. It’s not just about creating cutting-edge models—it’s about addressing tangible challenges in clinical care and improving health outcomes.

Yuxi is already thinking beyond thermal imaging alone. Her next project will integrate multi-modal data, combining thermal images with X-rays to provide a more comprehensive view of rheumatoid arthritis. The aim is not only to detect inflammation but also to predict the Sharp/van der Heijde (SvH) score, a widely used metric for evaluating joint damage.

When asked what advice she’d give to students interested in AI and healthcare, Yuxi emphasized balance and collaboration. “AI in healthcare isn’t just about building accurate models, it’s about deeply understanding the medical problem and working closely with clinicians,” she said. Her encouragement to “start small, embrace interdisciplinary teamwork, and always keep the patient impact in mind” is a timely reminder for anyone entering the field.

By reimagining how we use accessible tools like thermal imaging, her research opens the door to a future where chronic diseases such as RA can be detected earlier, diagnosed more accurately, and managed more effectively.