Specialised medical databases currently play a major role in archiving and searching for comparable data acquired by different modalities of medical imaging. In the context of storing and searching for data in medical databases, it is of immense importance to find a method of extracting and representing the contents found in the image that would ensure the rapid access and satisfactory results of searches for image information records. What is important, this representation should also be independent of the form of the image. The wide spread of multimedia medical databases that can store not just single images but also video sequences has shown that the problem of effectively searching for images containing specific disease cases that are significant for medical diagnostics is still fraught with great difficulties. This article presents a semantic retrieval methods in medical imaging databases using graph formalisms of syntactic image recognition which contribute to solving these problems. The proposed methods although they are mainly predestined for medical applications can also provide a base for other solutions, particularly for the acquisition and sophisticated semantic analysis of complex image patterns for security and defence reasons.