Radiotherapy is one of the major modalities of treatment for cancer, and approximately 50% of all cancer patients are referred for radiotherapy. Modern radiotherapy implies the acquisition of several medical images for initial planning of the treatment - including a CT scan, and/or MR and PET scans. Based on the images, segmentation is performed for identification of the treatment target and of surrounding healthy organs at risk. The patient individual arrangement of radiation beams and beam intensities as well as calculation of the radiation dose is then performed based on inverse optimization. These processes has accumulated a large amount of data including images, segmentations and simulated radiation doses for the population of patients treated over the past ~10 years - however the large amount of data is only to a limited extent used prospectively for prognostic or predictive purposes for future patients. I suggest a variety of projects aimed at prospectively utilizing the databases statistically, including machine learning methods, for assisting in decision making processes in the clinic.