My research focuses on advancing health care with AI along three connected fronts to improve patient care:
- Biomedical image foundation models: general-purpose models that learn from large, diverse datasets to segment, detect, and quantify structures across modalities such as CT, MRI, pathology, and microscopy. These models accelerate the translation of imaging biomarkers into precise cancer quantification and personalized treatment by enabling efficient, consistent, and reproducible measurements across institutions.
- AI agents for biomedical data analysis: combining large multimodal models with specialized tools to integrate imaging, clinical notes, structured EHR, and omics data, supporting high-throughput data analysis, early cancer diagnosis, treatment planning, and outcome prediction.
- Global benchmarks and challenges: build large-scale, clinically realistic datasets and organize international competitions that rigorously evaluate methods on efficiency, robustness, and generalization to under-represented and low-resource settings, exposing the real gap between research and deployment, and catalyzing reproducible, openly available tools that the community can safely bring into clinical practice.
Jun Ma is a Scientist at Princess Margaret Cancer Centre and Machine Learning Lead at UHN AI Hub. His research interests focus on developing cutting-edge algorithms for accurate and efficient biomedical image parsing, with the goal of enabling precise cancer quantification and personalized patient care. His work has been published in top journals, including Nature Methods, Lancet Digital Health, and Nature Communications. He has also won the top three in over 10 international medical image analysis challenges as the first author. His contributions have had a significant impact on the field, with over 17,000 citations. His open-source projects have garnered more than 10,000 stars on GitHub.
- Machine Learning Lead, AI Collaborative Centre, University Health Network