Tom Purdie

My research has focused on developing machine learning algorithms and methods for automating clinical radiation oncology workflow processes. For the last eight years, my research focus has been error/outlier detection, radiation dose prediction, human/machine interactions, and outcomes prediction. A key component of my research has been building clinically applicable tools promoting dissemination. I have also patented and subsequently licensed core machine learning technology to companies that can promote the widespread clinical adoption of machine learning in radiation oncology and healthcare in general.

I have started a number of internal machine learning-based studies at the University Health Network for radiation treatment planning, and decision support tools for quality assurance and peer-review rounds. In particular, our group has deployed machine learning as the standard of care for treating prostate cancer clinically. This particular study represents the first prospective deployment of machine learning for curative intent treatment that comprehensively investigated human preference and clinical decision making, in addition to machine learning performance.