The research group develops statistical and artificial intelligence (AI)-based methods for precision radiation oncology. Our goal is to predict how individual patients will respond to treatment – whether a tumour will be controlled and which side effects may occur – and to translate these predictions into clinically useful tools for treatment personalization and decision support. We integrate clinical and treatment information with quantitative imaging and molecular data, including CT, MRI, PET, radiation dose distributions and genomic information. Depending on the research question, we use classical biostatistics, machine learning and modern deep-learning approaches. A major focus is radiomics and radiogenomics, including longitudinal and multimodal imaging biomarkers and their integration with clinical and molecular information.
An important prerequisite for clinical AI is that models and biomarkers are not only accurate, but also reproducible, robust and reliable across institutions and patient populations. The group therefore develops methods for model validation and uncertainty quantification and contributes to international standardization activities such as the Image Biomarker Standardization Initiative (IBSI) and to open-source software for standardized radiomics. Ultimately, promising models are externally and prospectively validated with the aim of translating them into biomarker-driven clinical studies and individualized treatment strategies.
A second major research area is biologically informed proton therapy. We combine radiation dose, linear energy transfer (LET), imaging and clinical outcome data to better understand the variable relative biological effectiveness (RBE) of protons and radiation-induced side effects. Tumour control probability (TCP) and normal tissue complication probability (NTCP) models, together with AI-based approaches, are used to identify patients who may benefit most from proton therapy and from biologically adapted treatment planning.
More broadly, the group works at the interface of medical physics, radiation oncology, imaging, radiobiology and data science, linking preclinical and clinical evidence. Methodological research includes survival and outcome modelling, machine learning, uncertainty analysis and study design. The group also provides biostatistical and algorithmic expertise to other research groups at OncoRay.
