Interventional radiology procedures require extensive cognitive processing from the physician. A set of these cognitive functions are aimed to be replaced by technology in order to reduce the cognitive load. However, limited knowledge is available regarding mental processes in interventional radiology. This research focuses on identifying mental model–related processes, in particular during percutaneous procedures, useful to improve image guidance during interventions. Ethnographic studies and a prototype-based study were conducted in order to perform a task analysis and to identify working strategies and cognitive processes. Data were compared to theories from visual imagery. The results indicate a high level of complexity of mental model construction and manipulation, in particular when mentally comparing mental model knowledge with radiology images on screen (e.g., to steer a needle correctly). Regarding current interface support, most difficult is the interpretation and selection of oblique views. New interface principles are needed to bring cognitive demands within reasonable human range, and also accompanying cognitive work strategies should be developed.