Research on the medical applications of artificial intelligence has increased the knowledge of logical and methodological principles of clinical reasoning. Thus, computer-based diagnostic systems are developed on the basis of progress in this field, because thorough knowledge is necessary to obtain efficient simulation. This work was aimed at analyzing the structure of medical and radiological reasoning and at discussing the modalities to simulate it with computer-based diagnostic systems. The diagnostic process includes two steps: data collection and data interpretation; radiological reasoning involves the following 5 steps: procedural, executive, observative, interpretative and communicative. Each of them needs a different approach to simulation, considering, in its development, the different characteristics of each kind of reasoning. The expert system shells on the market are necessary tools to develop expert systems, but they cannot cover the whole of processes taking place during radiological work. Therefore, a particular, radiology-aimed shell should be developed to help the radiologist.