Patient models have been established as a useful tool for medical imaging. A number of methods haven shown that important applications such as effective dose, skin dose, and scatter kernel estimation can be improved using such a model. A patient model can, for example,  be generated from pre-operative CT/MRI data or by adapting parameters of a computational phantom, e.g., according to demographic patient data comprising gender, age, height, and weight. Commonly the patient model is generated before intervention. Furthermore, the model is considered static during the intervention, and it may not have been registered exactly to the actual patient position. These assumptions may introduce additional errors for interventional applications, e.g., with respect to skin dose estimation. Better results are expected by using a more shape-adaptive and better registered patient model. The current research focus is on patient modelling and model based skin and scatter estimation. This project is in cooperation with Siemens Healthcare GmbH, Forchheim.