This PhD thesis focuses on the low adoption of image-based models in radiation oncology for treatment outcome prediction. The thesis examines the reasons for this issue, including poor reproducibility of image data and the confounding influence of other factors, such as tumor volume, on image results. The author offers solutions, including methods to estimate uncertainty in image data and to remove confounding influences. The thesis also addresses technical challenges in implementing these solutions, including integrating radiomics into clinical imaging workflows and conducting federated radiomics studies between cancer centers. The goal of the thesis is to improve the use of imaging data in radiation oncology for better treatment outcomes.
Ivan Zhovannik is a biomedical engineering graduate who worked as data scientist and then joined Radboud University in 2018 for a PhD program in data mining and machine learning in radiation oncology. He has researched image biomarker analysis and has collaborated with multiple institutions worldwide. Ivan is currently a researcher in the Adaptive Radiotherapy group of the Netherlands Cancer Institute.