MED Structure

Semantic network of concepts

MED uses a semantic network model that includes a classification hierarchy. Each concept in the terminology is assigned a unique integer as an identifier (the MED code) and a unique name (the MED Name). The concept, "Medical Entity", serves as the topmost node in the classification, All other concepts are nodes in the graph, as immediate descendants of at least one other node. These parent-child relationships between nodes correspond to the classification of the concepts.

Multiple hierarchies

Each concept may have several parents; however, these relationships are acyclic, that is, a concept may not be its own descendant. Thus, the MED hierarchy is defined by a directed acyclic graph. For example, we can see that the term "Plasma Glucose" is in a hierarchy, based on the National Library of Medicine's Unified Medical Language System (UMLS). It has semantic links to other terms, such as its "part of" relation to the panel of tests called "Chem 7". It also is linked to terms which provide definitional information about the test, such as its specimen and the substance it measures. Note that "glucose" appears in an example of a multiple hierarchy. Concepts in the network have named attributes that may or may not have values. An attribute is "introduced" at a single node in the graph and is inherited by all nodes descended from that node.

Frame-based concept representation

Each concept node in the MED graph can be viewed as a frame with slots, in which some of the slots have values.  Click here for slot reference.  Each node can also be viewed as having links to nodes other than parent-child nodes through the semantic attributes. When such an attribute includes a reference to anther MED concept, the reference is a semantic link. A semantic attribute in the MED is always paired with second, inverse, attribute which is introduced into the graph at the node that corresponds to the domain of the first attribute. For example, the inverse of "Has Part" is "Part Of". Both attributes have "Medical Entity" as their domain and are introduced in that concept. "Substance Measured" is paired with the attribute "Measured By". The latter attribute is introduced in the node "Measurable Substance" and has the domain "Diagnostic Procedure". A hierarchical data structure was designed to represent this network of frames, using PC-MUMPS(Datatree, Waltham, MA).