This paper describes an approach to representing cases as nested graph-structures, i.e., as hierarchically, spatially, temporally and causally interconnected nodes (case nodes), which may be themselves recursively described by other sets of interconnected nodes. Each case node represents a case piece (sub-case). An adjacency matrix may represent these nested graph-structured cases. Within our approach, new cases are constructed using an iterative context-guided retrieval of case nodes from multiple cases. In order to illustrate the expressiveness of this case representation approach, we discuss its application to the diagnosis and therapeutics of neurological diseases, to architectural design and to storytelling. Some issues that come out of this approach, like its contribution to the representation of cases of CBR and to integrate ordinary and creative reasoning, are discussed.