Associate Professor

Thomas Purdie

PhD, University of Western Ontario

Princess Margaret Cancer Centre
700 University Ave, Room 6-412, Toronto, Ontario Canada M5G 1X7
Research Interests
Cancer Diagnosis and Therapy, Data Science and Computational Biology

At a Glance

  • Machine learning-based radiation treatment planning
  • Automated quality assurance for radiation oncology using machine learning methods
  • Radiomics and patient outcomes prediction
  • Complexity and quality assessment for radiation oncology
  • Explainable artificial intelligence for radiation oncology workflow

Short Bio

Dr. Thomas Purdie graduated from McMaster University, Hamilton, Ontario, Canada with a B.Sc. (Hons.) in the Medical and Health Physics program of the Department of Physics and Astronomy in 1997. He then completed his Ph.D. in the Department of Medical Biophysics at the University of Western Ontario in London, Ontario, Canada in 2002. Following graduate school, he completed a medical physics residency and research fellowship at Princess Margaret Cancer Centre (Toronto, Ontario Canada) in 2005. He is currently a Staff Medical Physicist in the Radiation Medicine Program, Princess Margaret Cancer Centre and became a Board Certified Medical Physicist (CCPM – Radiation Therapy Physics) in 2007. He is an Associate Professor in the Departments of Medical Biophysics and Radiation Oncology, University of Toronto and Affiliated Faculty at the Techna Institute,  University Health Network. Dr. Purdie’s research focuses on developing and clinically implementing machine learning based treatment planning and quality assurance (QA) processes in radiation oncology.

Research Synopsis

First automated breast intensity modulated treatment planning algorithm. The work described in the paper was the retrospective evaluation of a completely automated treatment planning method for tangential breast radiation therapy. This work is the basis for clinical automated treatment planning which have been implemented at our institution. Purdie TG, Dinniwell RE, Letourneau D, Hill C and Sharpe MB (2011). Automated Planning of Tangential Breast Intensity-Modulated Radiotherapy Using Heuristic Optimization. Int J Radiat Oncol Biol Phys, 81, 575-583.

Automated breast planning algorithms licensed to RaySearch Laboratories AB. Automated treatment planning algorithms were clinically implemented and enhanced to function within the clinical workflow at our institution, including the development of a user-interface, automatic quality assurance reporting and complete plan documentation. This technology was licensed to RaySearch Laboratories AB, a medical technology company that develops advanced software solutions for improved radiation therapy of cancer. 

  1. Large-scale clinical implementation of first automated treatment planning method. The paper described our large-scale clinical implementation of automated planning for breast RT over a three year period (June 2009-Nov 2012) treating more than 1600 patients. The results demonstrated that we could generate automated plans with fewer errors than more manual approaches. Purdie TG, Robert E Dinniwell, Fyles A, Sharpe MB (2014). Automation and IMRT for individualized high-quality tangent breast treatment plan. Int J Radiat Oncol Biol Phys, 90(3): 688-95.
  2. Patents for automated treatment planning. Two patents detailing automated planning have been filed as worldwide applications. Both patents extend our capacity to disseminate our automated planning developments by being more readily capable to commercialize and engage industry partners. The first patent is focused on automated breast treatment planning:
    1. Purdie TG, Sharpe MB. Method and Systems for Automated Planning of Radiation Therapy 05014971-31PCT, Canada, Ontario. Filing Date: 2011-10-06 The second patent is a more generalizable method also including quality assurance processes:
    2. Purdie TG, McIntosh C, Svistoun I. Method and System for Automated Quality Assurance and Automated Treatment Planning in Radiation Therapy - PCT/CA2014/050551, Canada, Ontario. Filing Date: 2014-06-12.
  3. Novel machine learning methods for automating processes in radiation therapy. The following four publications developed a robust data analysis pipeline that will be  leveraged to faciliate radiomics research: a) McIntosh C , Welch M , McNiven A , Jaffray DA , Purdie TG. (2017). Fully automated treatment planning for head and neck radiotherapy using a voxel-based dose prediction and dose mimicking method. Physics in medicine and biology. 62(15): 5926-5944.
    1.  McIntosh C , Purdie TG. (2017). Voxel-based dose prediction with multi-patient atlas selection for automated radiotherapy treatment planning. Physics in medicine and biology. 62(2): 415-431.
    2. McIntosh C , Purdie TG. (2016). Contextual Atlas Regression Forests: Multiple-Atlas-Based Automated Dose Prediction in Radiation Therapy. IEEE transactions on medical imaging. 35(4): 1000-12.
    3. McIntosh C, Svistoun I and Purdie TG.(2013). Groupwise conditional random forests for automatic shape classification and contour quality assessment in radiotherapy planning.IEEE Trans Med Imaging. 32(6): 1043-57.