The focus of this group is the prediction of patient outcome (tumor cure, toxicity) by statistical modeling with the aim to enable patient selection for improved individualized treatment options. The outcome modeling is based on clinical data and in addition includes genomic and imaging data. As these data are of high dimension, machine learning algorithms are applied (i) to select the most important variables (signature) and (ii) to construct prognostic models. The generated models are then validated on external datasets and, if successful, will be used in prospective clinical trials for patient stratification between different treatment options. The combination of all available omics data might furthermore help to reveal the molecular mechanisms responsible for tumor growth and response to irradiation. A specific emphasis of the research is on the analysis of clinical imaging data, i.e. radiomics analyses, as demonstrated in the Image Biomarker Standardization Initiative.
A further objective of the group is to understand the relative biological effectiveness of protons to improve the patient benefit of proton therapy in comparison to conventional photon therapy. For this purpose validated models on normal tissue complication probability (NTCP) are used. Further topics include the development and validation of tumor control probability (TCP) and NTCP models. The group offers statistical and algorithmic support to the other research groups at OncoRay.
