Specification and Detection of Periodic Patterns in Clinical Data

We specify and detect periodic patterns in clinical data using a constraint-based pattern specification language. The language has been implemented within both a graphical knowledge acquisition tool and a pattern interpreter, both of which are implemented within a larger temporal abstraction architecture. We provide several examples of how to use the syntax of the pattern language to specify complex periodic patterns. Working in the domains of diabetes and oncology, we have begun an evaluation of the pattern language, the knowledge acquisition process, and the pattern interpreter. Using clinical data from the University of Chicago bone marrow transplantation center, we have tested the interpreter’s ability to detect patterns. Preliminary results show that domain experts can use our language to specify most patterns of interest in the two domains.