UBC AI Fellows Program: Driving Smarter Cancer Care Across B.C.

Four research projects focused on advancing cancer care have been funded through the inaugural UBC AI and Health Network Fellows Program, which supports researchers harnessing artificial intelligence (AI) to transform health research and improve patient care across B.C.    

The Network’s Fellows Program enables principal investigators (PIs) to recruit, mentor and support postdoctoral and clinical fellows conducting interdisciplinary research at the nexus of AI and health. Each award provides $100,000 towards the hiring of a postdoctoral or clinical fellow.  

In total, 11 research projects, including four in the Cancer Stream and seven in the General Stream, were selected in the 2025 Fellows Program.   

The Cancer Stream awards were fully funded by the BC Cancer Foundation. 

The Fellows Program is designed to support researchers who are developing novel AI methods and tools to address real-world health challenges across B.C. and Canada. The program fosters collaboration across disciplines, from computer science and engineering to population health and clinical practice, and contributes to the responsible, ethical and equitable integration of AI in health care. 

“The Fellows Program will empower researchers to responsibly integrate AI into health care,” says Dr. Raymond Ng, professor in the UBC Department of Computer Science and co-leader of the Network. “The funded projects demonstrate the creative potential of our research community to spark innovation that benefits patients and populations.”  

“By connecting researchers across disciplines, the Fellows Program will cultivate future leaders in AI and health,” says Dr. Anita Palepu, professor and head of the UBC Department of Medicine and Network co-leader. “These collaborations will strengthen B.C.’s capacity for AI innovation, paving the way for better health outcomes across the province.”  

2025 AI and Health Network Fellows Awardees 

The following projects were selected through a competitive peer-review process for their scientific excellence, interdisciplinary approach and potential to advance responsible applications of AI in health research and care delivery. 

Cancer Stream

Adapting and Evaluating a Human-in-the-loop AI System to Enhance Exercise Support for People with Colon Cancer

Principal investigator: Lauren Capozzi, BC Cancer  
Co-investigators: John-Jose Nunez, BC Cancer and UBC Faculty of Medicine, Kristin Campbell, BC Cancer and UBC Faculty of Medicine

A major international study, CHALLENGE, published in 2025 in the New England Journal of Medicine, showed that a structured exercise program can help people with colon cancer live longer and feel better. The program worked because it included regular coaching from qualified exercise professionals to help people stay active safely. However, many cancer programs do not have enough staff to offer this support to everyone who needs it.  

This project will explore how a new type of AI system, called “human-in-the-loop” AI, can help extend this proven exercise program to more patients. This system combines automated coaching with oversight from real clinicians to make sure the advice is safe and personalized. Together with patients and care teams, we will adapt and test this AI-supported approach. Our goal is to make exercise support more accessible and improve recovery and quality of life after colon cancer.  

DiagTrace: Making Cancer Diagnosis Traceable with Knowledge Graphs and Chain-of-Reasoning

Principal investigator: Xiaoxiao Li, UBC Faculty of Applied Science
Co-investigators: Zu-Hua Gao, UBC Faculty of Medicine, Gang Wang, UBC Faculty of Medicine 

 Cancer care depends on information from medical reports and patient updates, but much of it is trapped in free text that requires time-consuming manual review. DiagTrace uses explainable AI to make this information clear, traceable, and actionable. It builds a structured “knowledge graph” linking key details such as tumor site, stage, and biomarkers from BC Cancer reports, with each fact cited to its source.   

Using this foundation, DiagTrace creates concise, auditable summaries that show how conclusions are reached (e.g., lesion → biomarker → therapy eligibility) and flag uncertainties or missing data. The system combines advanced language models with strong privacy and fairness safeguards, integrating patient-reported symptoms and preferences. Co-designed with clinicians, DiagTrace acts as an assistant, helping doctors by triaging reports and providing clear, evidence-based summaries to improve the speed and clarity of cancer care. The project will deliver a validated prototype and a toolkit for scaling AI-enabled diagnostics across cancer centers. 

AI Cancer Care Navigation Assistant

Principal investigator: John-Jose Nunez, BC Cancer and UBC Faculty of Medicine
Co-investigators: Lauren Capozzi, BC Cancer; Srinivas Raman, UBC Faculty of Medicine 

Many people with cancer face challenges accessing support such as counselling, reliable information, or transportation assistance when they need it most. Services are available through BC Cancer and community organizations. However, patients often struggle to find the right resources due to the complexity of the system, as resources can vary by cancer type, location, age, and other factors. These challenges can lead to delays in care, unmet needs, poorer treatment outcomes, and added strain on the healthcare system.   

Our project will bring on a postdoctoral fellow to help develop and test a personalized Cancer Care navigation assistant powered by AI. The assistant will recommend resources tailored to their situation, including those they may not know exist. It will deliver information in formats such as text, automated phone calls, or printable PDFs, empowering patients with varied technological preferences. 

Development of Acceptance Testing and Routine Quality Assurance of AI systems in Breast and Lung screening

Principal investigator: Rasika Rajapakshe, BC Cancer and UBC Faculty of Medicine
 
The integration of AI into breast and lung cancer screening workflows – particularly in mammography and low-dose computed tomography (LDCT) – has shown promise in enhancing detection accuracy, improving workflow efficiency and addressing radiologist shortages. However, the clinical safety and performance of AI systems are contingent upon rigorous acceptance testing prior to deployment and continuous quality assurance (QA) during operation. Unlike traditional imaging equipment, AI systems introduce new layers of variability due to their dependence on training data, model updates, and algorithmic behavior under varying clinical conditions.  

Furthermore, AI systems can detect ethnicity from medical images. There is currently a lack of standardized, evidence-based methodologies for testing, monitoring, and maintaining the performance of these AI tools in real-world screening settings. This project aims to address this gap. 

In addition to cancer-focused research, seven projects in other health domains were awarded through the General Stream. Learn more about these initiatives here. 

This article has been adapted from its original version. Read the UBC Faculty of Medicine story here.